Systematic Review of Acoustic Monitoring in Livestock Farming: Vocalization Patterns and Sound Source Analysis

Featured Application

An evidence map for study design in acoustic welfare monitoring. By synthesizing 36 experimental studies across cattle, poultry, and swine, this review collates recording setups, SPL scales, and frequency-based descriptors, and identifies gaps in calibration and reporting. These summaries support metric and device selection, data-acquisition planning, and interpretation of vocalization/noise under farm conditions, enabling comparable designs and gradual integration into precision-livestock workflows.

Abstract

Environmental sound and animal vocalizations provide non-invasive information for welfare assessment in livestock systems. This systematic review surveys their application in beef and dairy cattle, poultry, and swine, with a focus on environmental noise, vocalizations and the characterization of acoustic sources. Searches in Scopus and Web of Science followed PRISMA guidance and the PICO framework. After applying strict criteria that required peer-reviewed experimental studies in English, quantifiable acoustic data, and clear descriptions of measurement procedures, the review included 36 studies. Four approaches recur: vocalizations as welfare indicators; characterization of acoustic sources; combined analyses of vocalizations and sources; and evaluation of animal responses to acoustic stimuli. Recent work reports advances in recording equipment, signal processing, and precision livestock tools. Important challenges remain, including heterogeneous acoustic metrics, limited physiological validation, and difficulties applying models under commercial conditions. Overall, the evidence supports sound as a candidate for real-time monitoring and highlights the need for accessible, standardized methods. The findings provide a basis for future research and practical applications in welfare assessment.

1. Introduction

Estimates suggest that by the year 2080, the global population is projected to reach 10.3 billion people [1], which will result in a substantial increase in the demand for animal-based food products. In this context, it is essential to intensify production systems while ensuring productive efficiency and addressing sustainability challenges. The adoption of innovative technologies plays a fundamental role in optimizing productivity, preserving natural resources, and improving animal welfare conditions [2,3].

Several aspects that promote animal welfare are closely linked to recent technological developments. As highlighted by [4], non-invasive monitoring technologies are becoming increasingly important in ensuring appropriate welfare standards. Real-time data collection systems have proven valuable in supporting decision-making processes in production settings by providing measurable indicators related to productivity, efficiency, and welfare [5]. An example of this is the development of precision livestock farming (PLF) tools that integrate environmental and physiological variables. In swine farming, for instance, such technologies have demonstrated great potential for supporting the development of future comfort and welfare indices [6].

The variables of interest in production facilities, which are mentioned by Baêta and Souza [7], include air temperature, relative humidity, air velocity, radiation, light intensity and noise. The latter has been pointed out by [8] as an emerging factor in the evaluation of thermal comfort and animal behavior. Its influence on animals’ emotional state and health can ultimately compromise their welfare [9].

Recent studies have shown that environmental stimuli, including acoustic stimuli, influence physiological and behavioral well-being. Noise is considered a relevant stressor that can impair neuroendocrine, cardiovascular, and reproductive functions [10,11]. The advancement of monitoring technologies has also improved the precision with which these variables can be managed, reinforcing the importance of incorporating noise as a quantifiable parameter in environmental improvement programs [12].

Nevertheless, sound in animal production systems does not function exclusively as a stressor. Several studies have shown that sound can be a useful and non-invasive indicator of the physiological and emotional state of animals. Within the field of bioacoustics, vocalizations, acoustic signals, and environmental sound patterns have been analyzed as indicators of health, behavior, and thermal comfort.

In the case of swine production, prolonged vocalizations during the weaning phase were used to indicate the existence of stress [13]. Other studies have used acoustic analysis as a tool to evaluate welfare and thermal comfort in slaughterhouses and pig farms [14,15]. Likewise, poultry production systems have used vocal analysis, as demonstrated in [16], which found relationships between sound signals of laying hens with their welfare in enriched environments. Similarly, ref. [17] found acoustic patterns related to the feeding behaviors of broilers, generating a relevant contribution to the management and welfare in the facility.

Despite the growing interest in using bioacoustics as a control tool in animal production systems, its implementation still encounters some methodological challenges that have been pointed out in research papers and literature reviews. These include the lack of technical standardization for the use and analysis of acoustic variables [18], logistical barriers that limit access to specialized equipment [19], the absence of methodological standardization and variability in relating acoustic parameters to other parameters of interest in animals such as age, body size, and health status [20].

Although interest in acoustic monitoring within livestock production has increased in recent years, the available literature is still rather fragmented. Coutant et al. [20] reviewed advances in bioacoustics, with an emphasis on animal welfare indicators across different species, whereas [21] concentrated on specific applications such as the study of pig vocalizations. These reviews, however, stop short of offering a systematic comparison of methodologies, acoustic parameters, and outcomes across production systems. A further challenge that has been repeatedly highlighted is the absence of standardized protocols for sound recording, analysis, and interpretation [20], which complicates reproducibility and limits cross-study comparisons. Meanwhile, Precision Livestock Farming (PLF) has emerged as an important approach to promote the sustainable intensification of livestock production [22]. Berckmans [23] pointed out that acoustic information can serve as a valuable resource for tracking animal health and welfare in real time. Even so, its incorporation into PLF systems is still in its early stages.

Thus, the aim of the systematic review was to provide a comprehensive overview of studies that have evaluated sound in animal production systems, particularly in swine farming, poultry farming (broilers and laying hens) and cattle farming (dairy and beef). This review seeks to identify current trends, methodologies, and challenges, and to propose recommendations to guide future research and the technological application of acoustic monitoring for animal welfare assessment in intensive production systems.

2. Materials and Methods

2.1. Search Strategy

This systematic review was conducted exclusively using electronic formats from two databases: Web of Science and Scopus, selected for their broad coverage and scientific relevance. Grey literature sources, industry reports, and other academic databases were not included in the search strategy, which represents a limitation of the review but ensures methodological consistency and comparability of the selected studies. Data collection followed the PICO framework (Population, Intervention, Comparison, and Outcome), in accordance with the PRISMA methodology (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) [24]. This systematic review was registered in the Open Science Framework (OSF) under the identifier https://osf.io/e5ybj (accessed on 7 September 2025). Based on these criteria, key terms were defined and organized according to the acronyms presented in

Table 1. Keywords used to search for papers in the Web of Science and Scopus databases, based on the PICO structure (Population, Intervention, Comparison and Outcome).

Acronym	Palabras Clave
Population	(((cattle OR cow* OR calves OR calf OR heifer*) AND (dairy OR lactating OR milking OR beef)) OR (chicken* OR “laying hen*” OR broiler* OR poultry) OR (pig* OR swine* OR sow* OR piglet* OR pork* OR “domestic pig*”))
Intervention	(facility OR facilities OR installation* OR farm* OR livestock OR “animal husbandry” OR “Precision Livestock Farming” OR “precision animal production” OR confinement* OR handling OR parlor* OR slaughter* OR nursing)
Comparison	((noise OR sound* OR acoustic* OR vocal* OR audio*) AND (sensor* OR sensing OR measurement OR quantification OR monitoring))
Outcome	(“acoustic characteristics” OR dB OR decibel* OR Hz OR frequency OR “sound intensity” OR “sound level*” OR “sound pressure” OR “sound analysis”)

*—employed to include alternative spellings of the words and/or expressions of interest, such as variations between US English and British English. Source: Authors.

.

2.2. Study Selection

Studies indexed in both databases from the first record in 1965 up to the most recent available date in October 2024 were included. However, the earliest article meeting the eligibility requirements was published in 1989, so the effective coverage of the final sample spans 1989–2024. In the preliminary stage, results from both sources were merged, and duplicate entries were removed using the R software environment [25] and the RStudio interface, version 4.4.2 [26]. However, due to inconsistencies in title formatting, a manual review was necessary to identify duplicate references not detected by the software. At this stage, J.N.R.N. (Author 1) and C.E.A.O. (Author 3) conducted the initial screening of records, followed by verification of categorization by J.N.R.N. (Author 1) and F.C.S. (Author 2). Any discrepancies were discussed collectively and resolved by consensus among all authors, who also contributed to finalizing the classification into thematic categories.

As an initial filter, only articles written in English were accepted. Non-experimental publications, such as literature reviews, theses, conference abstracts, and book chapters, were excluded. Subsequently, only studies meeting the following criteria, formulated as guiding questions, were selected:

• Was noise and/or sound measured in facilities housing dairy cattle, beef cattle, laying hens, broiler chicken, or swine?
• Were the measurements conducted within animal production facilities, slaughterhouses, or controlled research environments?
• Was the methodology and the equipment used for sound measurement in animal production described in detail?
• Were quantitative criteria applied to assess sound within animal production systems?

It is important to highlight that the last question served as a key inclusion criterion for filtering out articles that, although meeting the previous parameters, focused primarily on variables other than sound. This included studies that used sound data as input for machine learning (ML) models without providing a detailed description or evaluation of the sound measurement itself. The sequence of search and selection of studies is presented in flowchart format (Figure 1).

2.3. Data Extraction and Data Processing

As a final step, relevant information was extracted from the 36 selected papers with the aim of addressing the research question posed in this systematic review. During this process, several aspects were analyzed, including general information for each article (such as title and objectives), characteristics of the animals studied (species or production category, age, weight, and so forth), and details regarding the location where the study was conducted (type of facility, country, and specific setting).

In addition, methodological aspects directly related to sound evaluation were reviewed, including the type of equipment used, the positioning of the devices, and the techniques applied. The most relevant findings reported by the authors were identified, distinguishing between variables directly related to sound and those that were non-acoustic, as well as the conclusions and future research perspectives indicated in each study.

Each article was then classified into one of the following thematic categories: (1) vocalizations as indicators of animal welfare, (2) characterization of sound sources, (3) combined analysis of sound sources and vocalizations, or (4) animal responses to acoustic stimuli. In cases where a study addressed more than one theme, classification was based on the predominant research aim described by the authors.

For example, Andrade et al. [27], who analyzed the amount of noise and gas concentrations within swine housing systems, were placed under (2) “characterization of sound sources”. Similarly, Ferrari et al. [28], who analyzed the potential for characterizing sounds of cough from sounds of the background, were placed under (1) “vocalizations as welfare indicators for animals”. Another example is Ferrari et al. [29], who presented acoustic features of sounds of cough based on pulmonary infection and compared them with coughs caused by citric acid, placed into (3) “combined analysis of sound sources and vocalizations”. Finally, Lemcke et al. [30], who analyzed the effect of some genres of music on dairy cows that are milked from automatic systems, were placed under (4) “animal responses to acoustic stimuli”.

The variables identified were categorized as acoustic (

Table 2. Classification of sound-related (acoustic) variables.

Acoustic Category	Direct Examples	Purpose in the Studies
Sound-pressure level (SPL)	LAeq, Lmax, L10–L90 percentiles, dB (Lin), dB(A), dB(C), etc.	Quantify overall noise load.
Duration/time	Call duration, inter-call interval.	Analyze emission rhythms and temporal patterns.
Vocalization type/emission	Alarm, contact, estrus sounds, tractor noise.	Link the sound to its function or technical source.
Waveform/acoustic contour	Amplitude variation in the time domain, temporal envelope, number of pulses.	Segment signals; detect artefacts.
Overall acoustic profile	Spectro-temporal characteristics: timbre, entropy, roughness, global power spectrum.	Summarize the global sonic signature of a habitat or species.
Frequency	Fundamental f0, formants, main resonance f*.	Used in the detection of sounds and stressful situations or animal health

Note: The asterisk (*) indicates the main resonance frequency (f*). Source: Authors.

) and non-acoustic (

Table 3. Classification of non-acoustic variables.

Category	What It Groups	Typical Examples Found in the Papers
Environmental conditions	Physical factors of the housing environment.	Relative humidity, air velocity, CO2 or NH3 concentration, wet bulb temperature, dust, etc.
Animal behavior	Observable actions or postures that signal welfare or stress.	Wing flapping in poultry, tail biting in piglets, pain related vocalizations, lying/standing time, cubicle occupancy.
Weight/development	Growth and body condition metrics.	Live weight, average daily gain, feed conversion ratio, heart girth circumference, body condition score.
Stress/physiology/health	Biochemical or physiological indicators of the internal state.	Salivary or serum cortisol, heart rate, respiration rate, plasma lactate, rectal temperature, lesion scoring.
Phenotype/classification	Genetic or morphological traits that differentiate groups.	Breed, genetic line.
Management/routine	Human imposed housing or experimental practices.	Number of milkings per day, feeding regime, fan/sprinkler cycles.

Source: Authors.

). These tables include both the purposes or definitions of each category and specific examples extracted directly from the 36 papers analyzed in this study.

To facilitate the analysis and interpretation of the results, graphical resources and visual representations were employed in this review. Initially, a word cloud was generated using the R Bibliometrix package, version 0.3.0 [31], based on the metadata (titles, abstracts, and keywords) of the 36 articles selected after the full screening process. This approach enabled the identification of the most frequently occurring terms in the final sample, offering an overview of the main thematic trends addressed in the reviewed literature. The collected data were analyzed to identify temporal and thematic patterns across studies, as well as associations between animal categories, production phases, and acoustic variables. Additionally, co-occurrence analysis was performed to explore the relationships between acoustic and non-acoustic parameters.

As part of the data extraction process, information regarding the production stage of the animals studied was standardized. In the case of swine, terms used across studies varied, requiring harmonization to ensure consistent analysis. The review keeps terminology consistent by assigning each swine category to one production stage, guided by physiological traits and production goals. This approach recognizes four stages: nursery, growing, finishing, farrowing, and breeding and gestation. For instance, the terms “nursery pigs” and “nursery-to-finishing pigs” were placed under the nursery phase, while reproductive categories such as “sows,” “gilts,” and “gestating or breeding sows” were included in the breeding and gestation phase. This standardized approach facilitated more accurate comparisons across production contexts and research settings.

Risk of bias was evaluated using the Joanna Briggs Institute (JBI) eight-item checklist for analytical cross-sectional designs [32,33]. Two independent reviewers (J.N.R.N. and C.E.A.O.) performed the appraisal and reconciled discrepancies through consensus. The overall risk was categorized as low when ≥70% of applicable items were marked “Yes”; high when <70% were “Yes” and the share of “Unclear” judgments was ≤50%; and unclear when “Unclear” predominated (≥50% of applicable items or a tie). Plots and summary tables were created in RStudio interface, version 4.2.2 [26].

The statistical analyses and graphical outputs were performed using the following packages in the R statistical software: bibliometrix [31], tidyverse [34], readxl [35], dplyr [36], tidyr [37], ggplot2 [38], forcats [39], writexl [40], scales [41], RColorBrewer [42], stringr [43], terra [44], sf [45], ggrepel [46], viridis [47], ComplexUpset [48], igraph [49] and ggraph [50].

3. Results

As a first step, the metadata of the 36 articles (abstracts, titles and keywords) were used to generate a word cloud (Figure 2), which represented the most frequently occurring terms and provided an overview of the predominant themes in the literature review.

3.1. General Overview of Studies

The 36 studies included in Table 4 encompass a diverse range of research on the use of acoustics in animal production systems. Each study was classified under one of four thematic axes defined during the methodological phase: (1) vocalizations as indicators of animal welfare, (2) characterization of sound sources, (3) combined analysis of sound sources and vocalizations, or (4) animal responses to acoustic stimuli. This classification, indicated in the last column of Table 4, provides a structured basis for the subsequent analysis.

Table 4. Summary of the objectives of the 36 papers selected for this study.

Paper	Summary of Objectives	Thematic Axis *
[27]	Analyze noise levels and gas concentrations in swine housing systems.	(II)
[51]	Assess gas concentrations and sound pressure levels across three types of facilities for swine growing and finishing phases.	(II)
[52]	Assess the impact of acute noise exposure on stress responses in broiler chicken.	(IV)
[53]	Extract and classify acoustic features to develop a model for the identification of vocalizations in laying hens.	(III)
[29]	Describe the acoustic features of cough sounds associated with pulmonary infection and compare them with coughing induced by citric acid inhalation.	(III)
[28]	Investigate the feasibility of distinguishing cough sounds from background noise within a livestock facility.	(I)
[54]	Investigate the relationship between animal vocalizations and body weight.	(I)
[55]	Analyze and characterize chick vocalizations in relation to social behaviors and age progression.	(I)
[56]	Identify and validate a model linking broiler growth rate with peak vocalization frequencies.	(I)
[57]	Explore how vocalizations and phonatory behaviors of Holstein-Friesian cows during peripartum events (e.g., dystocia and calf separation) convey contextual and sensory information relevant to animal welfare assessment.	(I)
[58]	Determine whether individual swine can be identified through scream analysis in audio recordings.	(I)
[59]	Evaluate whether swine vocalizations are useful indicators of thermal adaptability.	(I)
[60]	Measure noise levels in three slaughterhouses (bovine and swine) using a smartphone application.	(II)
[61]	Update current knowledge regarding noise levels in swine farms in Germany.	(II)
[62]	Propose a robust system for behavioral characterization of laying hens to enhance monitoring efficiency.	(I)
[30]	Explore the effects of selected music genres on dairy cows milked by automated milking systems (AMS).	(IV)
[63]	Evaluate the immediate physiological and behavioral responses of piglets subjected to ear notching, ear tagging, and intraperitoneal transponder injection.	(I)
[64]	Investigate the relationship between vocal behavior and other behavioral traits in cattle.	(I)
[65]	Analyze the relationship between sneeze frequency and various strains of influenza virus in domestic swine.	(I)
[66]	Develop software to monitor piglet vocalizations based on stress-related sound patterns.	(I)
[67]	Identify and analyze the vocal responses of sows and piglets following short-term separation and assess the potential of such calls to indicate levels of arousal.	(I)
[68]	Identify changes in swine vocalizations during different stages of the castration process.	(I)
[69]	Evaluate the impact of multiple daily feedings on body weight, backfat thickness, aggressiveness, and locomotor issues in sows.	(I)
[70]	Determine whether the energy envelope dynamics of swine coughing sounds are associated with respiratory pathologies.	(I)
[71]	Test the hypothesis that noise levels in swine housing vary with time of day and season.	(II)
[72]	Evaluate the potential impact of sound measurements in swine facilities based on distance, time of day, building orientation, and season.	(II)
[73]	Examine the effects of two injection methods and two local anesthetics on piglet escape behaviors, including vocalizations and resistance movements, as well as procedure duration.	(I)
[74]	Investigate the effects of tartaric acid nebulization in swine finishing units.	(II)
[75]	Characterize the types of sounds swine are exposed to during housing, transport, and slaughter stages.	(II)
[76]	Assess the effectiveness of buprenorphine in mitigating pain during piglet castration using behavioral and vocalization indicators.	(I)
[77]	Understand vocal behavior in cattle under different stress conditions.	(I)
[78]	Conduct a situational analysis of noise levels in swine finishing farms and evaluate the adequacy of noise as a welfare indicator.	(II)
[79]	Assess the feasibility of segmenting animal vocalizations using electronic acoustic analysis tools.	(I)
[80]	Assess whether significant differences exist among acoustic characteristics of various vocal signals within a herd of crossbred cows.	(I)
[81]	Determine the frequency ranges of six common sounds in commercial broiler houses, including fans, heaters, feeding systems, vocalizations, wing flapping, and dust bathing.	(III)
[82]	Evaluate the effectiveness of low-cost, passive strategies for significant noise reduction.	(II)

* Each study was classified into one of the following thematic axes: (I) Vocalizations as indicators of animal welfare; (II) Characterization of sound sources; (III) Combined analysis of sound sources and vocalizations; (IV) Animal responses to acoustic stimuli. Source: Authors.

Table 4. Summary of the objectives of the 36 papers selected for this study.

Paper	Summary of Objectives	Thematic Axis *
[27]	Analyze noise levels and gas concentrations in swine housing systems.	(II)
[51]	Assess gas concentrations and sound pressure levels across three types of facilities for swine growing and finishing phases.	(II)
[52]	Assess the impact of acute noise exposure on stress responses in broiler chicken.	(IV)
[53]	Extract and classify acoustic features to develop a model for the identification of vocalizations in laying hens.	(III)
[29]	Describe the acoustic features of cough sounds associated with pulmonary infection and compare them with coughing induced by citric acid inhalation.	(III)
[28]	Investigate the feasibility of distinguishing cough sounds from background noise within a livestock facility.	(I)
[54]	Investigate the relationship between animal vocalizations and body weight.	(I)
[55]	Analyze and characterize chick vocalizations in relation to social behaviors and age progression.	(I)
[56]	Identify and validate a model linking broiler growth rate with peak vocalization frequencies.	(I)
[57]	Explore how vocalizations and phonatory behaviors of Holstein-Friesian cows during peripartum events (e.g., dystocia and calf separation) convey contextual and sensory information relevant to animal welfare assessment.	(I)
[58]	Determine whether individual swine can be identified through scream analysis in audio recordings.	(I)
[59]	Evaluate whether swine vocalizations are useful indicators of thermal adaptability.	(I)
[60]	Measure noise levels in three slaughterhouses (bovine and swine) using a smartphone application.	(II)
[61]	Update current knowledge regarding noise levels in swine farms in Germany.	(II)
[62]	Propose a robust system for behavioral characterization of laying hens to enhance monitoring efficiency.	(I)
[30]	Explore the effects of selected music genres on dairy cows milked by automated milking systems (AMS).	(IV)
[63]	Evaluate the immediate physiological and behavioral responses of piglets subjected to ear notching, ear tagging, and intraperitoneal transponder injection.	(I)
[64]	Investigate the relationship between vocal behavior and other behavioral traits in cattle.	(I)
[65]	Analyze the relationship between sneeze frequency and various strains of influenza virus in domestic swine.	(I)
[66]	Develop software to monitor piglet vocalizations based on stress-related sound patterns.	(I)
[67]	Identify and analyze the vocal responses of sows and piglets following short-term separation and assess the potential of such calls to indicate levels of arousal.	(I)
[68]	Identify changes in swine vocalizations during different stages of the castration process.	(I)
[69]	Evaluate the impact of multiple daily feedings on body weight, backfat thickness, aggressiveness, and locomotor issues in sows.	(I)
[70]	Determine whether the energy envelope dynamics of swine coughing sounds are associated with respiratory pathologies.	(I)
[71]	Test the hypothesis that noise levels in swine housing vary with time of day and season.	(II)
[72]	Evaluate the potential impact of sound measurements in swine facilities based on distance, time of day, building orientation, and season.	(II)
[73]	Examine the effects of two injection methods and two local anesthetics on piglet escape behaviors, including vocalizations and resistance movements, as well as procedure duration.	(I)
[74]	Investigate the effects of tartaric acid nebulization in swine finishing units.	(II)
[75]	Characterize the types of sounds swine are exposed to during housing, transport, and slaughter stages.	(II)
[76]	Assess the effectiveness of buprenorphine in mitigating pain during piglet castration using behavioral and vocalization indicators.	(I)
[77]	Understand vocal behavior in cattle under different stress conditions.	(I)
[78]	Conduct a situational analysis of noise levels in swine finishing farms and evaluate the adequacy of noise as a welfare indicator.	(II)
[79]	Assess the feasibility of segmenting animal vocalizations using electronic acoustic analysis tools.	(I)
[80]	Assess whether significant differences exist among acoustic characteristics of various vocal signals within a herd of crossbred cows.	(I)
[81]	Determine the frequency ranges of six common sounds in commercial broiler houses, including fans, heaters, feeding systems, vocalizations, wing flapping, and dust bathing.	(III)
[82]	Evaluate the effectiveness of low-cost, passive strategies for significant noise reduction.	(II)

* Each study was classified into one of the following thematic axes: (I) Vocalizations as indicators of animal welfare; (II) Characterization of sound sources; (III) Combined analysis of sound sources and vocalizations; (IV) Animal responses to acoustic stimuli. Source: Authors.

Figure 3 illustrates the distribution of studies across thematic categories. The most prominent theme is the use of vocalizations as indicators of animal welfare, representing 58% of the studies.

In this regard, several automated models have demonstrated the ability to detect specific vocal and behavioral patterns in laying hens, to detect behavioral patterns [53,62] to recognize individual swine screams [58], and to associate the peak frequency of broiler vocalizations with growth rate [56]. Studies found in swine production have used sneezing and coughing sounds as biomarkers of respiratory disease [29,65,70], and the intensity and duration of piglet squeals during procedures involving buprenorphine application as an indicator of pain [68,76].

Likewise, other research conducted on Holstein-Friesian cows has analyzed vocal modulation across peripartum phases [57] and as an individual personality trait [64]. Examples include the study by [54] that uses vocalizations in the estimation of body weight, and the study of sound emissions by [67], who evaluated the anxiety of gilts as a consequence of isolation.

The second most frequent category is the characterization of sound sources (28% of the studies). Some of these studies only measure sound intensities, as in the case of [61,78], who characterized the acoustic environment in commercial facilities in Germany. Others have presented other environmental variables as well as sound-related variables. This is the case of [27,51], who measured both sound intensity levels and concentrations of gases or dust. Other aspects related to the characteristics of the installations have also been studied, as in the case of [71,72] who analyzed the spatial and temporal variability of sound taking into consideration aspects such as distance to the source, orientation of the installation and seasonality. Ref. [82] also analyzed the effectiveness of low-cost barriers in reducing sound intensity. From the perspective of sound sources in facilities, ref. [81] identified, associated and mapped sound sources in broiler facilities, including fans, feeders and some animal behaviors such as wing flapping. Along the same lines, ref. [75] recorded and characterized the acoustic environment during the transport and slaughter of pigs.

A smaller part of the articles, corresponding to 8%, focus on a subject that combines the study of acoustic effects in animal production, both vocalizations and those generated by sound sources. The least represented theme (6%) includes studies on animal responses to acoustic stimuli, analyzing how different types of sound influence behavior and physiology. For example, acute noise exposure increases corticosterone levels and reduces weight gain in broiler chicken farming [52]. In dairy cattle, some studies have explored the impact of playing selected songs during automated milking routines [30].

Finally, Figure 4 provided a visualization of scientific production behavior over the years, highlighting that publication levels remained relatively stable until 2007, followed by a steady increase in research activity.

3.2. Animal Production Context

Figure 5 shows that 61% of the studies focus on swine. Research on dairy cattle and on broiler chickens each account for 14% of the total. The specific case of the multi-category study corresponds to the combination of swine and beef cattle in a slaughterhouse investigation conducted by [60]. Another relevant aspect is the type of facility where the studies were conducted. Figure 6 shows that the most balanced percentages are those corresponding to swine and poultry production studies.

On the other hand, in broiler and dairy cattle, only between 17% and 20% of the studies were conducted in experimental facilities, which could limit the direct application of the results found. Also, in this review, only the study by [60] was found for commercial beef cattle facilities.

Figure 7A highlights a clear research focus on the farrowing and finishing stages of swine production, each covered by ten studies. Eight of the farrowing stage studies concentrate specifically on piglets. Breeding and gestation appeared in seven studies, whereas the nursery stage received the least attention, with only five studies.

In cattle (Figure 7B), the focus is divided between calves and lactating dairy cows, with three and four studies identified, respectively. Research has explored the role of vocalizations during critical events such as the peripartum period [57], responses to environmental noise under stress [77], and the positive influence of classical music in automated milking systems [30,83].

In poultry production (Figure 7C), acoustic research has primarily focused on the characterization of vocalizations and the identification of sound sources within housing environments. Notable among these is the work of [81], who identified frequency bands associated with various elements in broiler houses, including fans, heaters, feeding systems, vocalizations, wing flapping, and dust bathing. This characterization enabled the mapping of the internal sound environment and provided a basis for assessing environmental conditions from an acoustic perspective. Additionally, ref. [55] analyzed the progression of chick vocalizations throughout growth, linking them to social and developmental behaviors.

However, the current landscape shows a predominance of studies on broilers without a defined growth phase (n = 4) and only one study focused on chicks, while two studies on laying hens did not specify the production stage.

3.3. Sound Capture Technologies

Early work in farm animal bioacoustics relied on straightforward studio microphones, for example, the Sennheiser MKH416 Tu3 and the Brüel & Kjær 4165, which researchers connected to portable tape recorders such as the Uher 4200 or the Racal Store 4DS [75,79].

Later work introduced high-sensitivity condenser microphones, including the Monacor ECM3005 and the Sennheiser ME67, which researchers plugged into laptops through sound cards such as the Realtek AC97 [28,29,57,70]. At the same time, projects that examined background noise in livestock buildings began to use hand-held sound-level meters like the Extech 407764 and the Voltcraft SL-300 [69,71,72,82].

Storage practices have evolved in step with recording hardware. In the late 1980s and 1990s, investigators preserved vocalizations on reel-to-reel or compact-cassette decks [75,77,79]. Most recent studies instead use solid-state recorders, for example, the Marantz PMD 661 MK II [56,57].

Similarly, several authors report using professional calibrators such as the Brüel & Kjær 4238 and the Norsonic AS 1251 to check levels before data collection [60,61,71,72,75,78,82]. However, many papers do not describe their calibration procedures. A detailed summary of the recording and calibration technologies reported in the reviewed studies, including representative brand/model examples and references, is presented in Table 5.

Table 5. Sound capture and calibration technologies reported in the reviewed studies.

Device Category	Representative Equipment (Brand/Model)	Typical Application	References
Early studio microphones	Sennheiser MKH416 Tu3; Brüel & Kjær 4165	Direct vocalization recording, often connected to analog tape recorders	[75,79]
Portable recorders (cassette/reel)	Uher 4200; Racal Store 4DS	Storage of vocalizations in early studies (1980s–1990s)	[75,77,79]
Condenser microphones	Monacor ECM3005; Sennheiser ME67	High-sensitivity recordings, often connected via laptop sound cards (e.g., Realtek AC97)	[28,29,57,70]
Sound level meters (SLM)	Extech 407764; Voltcraft SL 300	Background noise measurement in livestock housing	[69,71,72,82]
Solid-state recorders	Marantz PMD 661 MK II	Modern portable storage of acoustic data	[56,57]
Calibrators	Brüel & Kjær 4238; Norsonic AS 1251	Checking equipment levels before data collection	[60,61,71,72,75,78,82]
Video systems with external microphones	SOMO system (SoundTalks NV); other unspecified video + mic setups	Complement acoustic and visual data; automated detection	[56,62]
Multi-equipment setups	Microphone + SLM; microphone + video	Combine multiple data sources to enhance environmental and acoustic characterization	[57,64,65,73,81,82]

Note: The table summarizes the main sound capture and calibration technologies identified across the 36 studies included in this review. Not all articles specified equipment brands or models; therefore, the references listed correspond only to those studies that explicitly reported the use of the indicated devices. “SLM” refers to portable sound level meters used for measuring overall noise levels (e.g., dB(A), dB(C)). “Early studio microphones” refers to professional microphones (e.g., Sennheiser MKH416 Tu3, Brüel & Kjær 4165) employed in early farm animal bioacoustics studies, often connected to analog tape recorders.

Table 5. Sound capture and calibration technologies reported in the reviewed studies.

Device Category	Representative Equipment (Brand/Model)	Typical Application	References
Early studio microphones	Sennheiser MKH416 Tu3; Brüel & Kjær 4165	Direct vocalization recording, often connected to analog tape recorders	[75,79]
Portable recorders (cassette/reel)	Uher 4200; Racal Store 4DS	Storage of vocalizations in early studies (1980s–1990s)	[75,77,79]
Condenser microphones	Monacor ECM3005; Sennheiser ME67	High-sensitivity recordings, often connected via laptop sound cards (e.g., Realtek AC97)	[28,29,57,70]
Sound level meters (SLM)	Extech 407764; Voltcraft SL 300	Background noise measurement in livestock housing	[69,71,72,82]
Solid-state recorders	Marantz PMD 661 MK II	Modern portable storage of acoustic data	[56,57]
Calibrators	Brüel & Kjær 4238; Norsonic AS 1251	Checking equipment levels before data collection	[60,61,71,72,75,78,82]
Video systems with external microphones	SOMO system (SoundTalks NV); other unspecified video + mic setups	Complement acoustic and visual data; automated detection	[56,62]
Multi-equipment setups	Microphone + SLM; microphone + video	Combine multiple data sources to enhance environmental and acoustic characterization	[57,64,65,73,81,82]

Note: The table summarizes the main sound capture and calibration technologies identified across the 36 studies included in this review. Not all articles specified equipment brands or models; therefore, the references listed correspond only to those studies that explicitly reported the use of the indicated devices. “SLM” refers to portable sound level meters used for measuring overall noise levels (e.g., dB(A), dB(C)). “Early studio microphones” refers to professional microphones (e.g., Sennheiser MKH416 Tu3, Brüel & Kjær 4165) employed in early farm animal bioacoustics studies, often connected to analog tape recorders.

In terms of device placement, two main strategies were observed. The first involved placing devices at a short distance from the animal, typically between 0.3 and 1.5 m. The second consisted of positioning the device slightly above the activity area, taking the ground level as a reference, particularly in broiler and swine housing environments [59,81].

Analysis of the proportion of technologies used by species reveals relevant differences (Figure 8). In laying hens, the use of video cameras was reported [62]. In broilers, a combination of microphones, sound level meters, and, to a lesser extent, specialized equipment such as the SOMO system by SoundTalks NV has been applied [56]. In swine and cattle, recording technologies are diversified and include condenser microphones, sound level meters, and video systems with external microphones.

An additional aspect highlighted in Figure 9 is the use of the term “multi-equipment”, which refers to the simultaneous combination of more than one capture device within a single experiment. This strategy, documented in six of the reviewed articles [57,64,65,73,81,82], aims to complement acoustic information with visual data or environmental sound pressure levels.

3.4. Variables and Parameters Assessed

Variables such as sound pressure level were expressed using different scales (dB, dB(A), dB(C)), while frequency-related parameters referred to various aspects of sound, such as fundamental frequency, mean frequency, and resonance frequency, despite sharing the same unit of measurement (Hz). In addition, several studies grouped multiple acoustic characteristics into general descriptions.

The analysis of variable co-occurrence provides a clearer view of the relationships established between acoustic and non-acoustic parameters (Figure 10). The network plot shows that the variable “frequency” occupies a central position in the network, connecting with duration, overall acoustic profile, and physiological or behavioral aspects. Complementarily, the UpSet plot (Figure 11) shows that the most frequent combinations involve “Frequency” alongside “Animal behavior” or “Physiology/health”.

Finally, the risk of bias assessment showed clear inclusion criteria, adequately described exposure measurements, and appropriate statistical analyses across studies (Figure 12). Overall, 35 out of 36 studies presented a “Low” risk of bias, while 1 out of 36 was classified as “High”. Questions Q5 and Q6, related to the identification and management of confounding factors, accumulated the majority of judgments other than “Yes”.

4. Discussion

4.1. General Overview of Studies

The prominence of terms in Figure 2 such as pigs, vocalization, animal welfare, noise, environment, acoustic and stress corroborates the notion that sound is primarily explored as an indicator of environmental conditions, management practices, or animal health in production systems. These preliminary findings indicate a particular concentration of research on swine, with frequent references to acoustic monitoring, sound analysis, and precision livestock farming.

The predominance of vocalization-focused studies in Figure 3 indicates a clear trend toward using animal-generated acoustic parameters to infer physical and emotional states. Conversely, the small proportion addressing responses to external noise suggests relatively limited attention to this approach within the reviewed literature.

Consistent with this trend, animal vocal emissions can reflect fear, pain, illness, or stress, and their spectral and temporal analysis has become a key tool for inferring physiological, health, or emotional conditions. Vocalizations are used from different perspectives as non-invasive indicators of animal welfare. For instance, they have been applied to monitor heat stress [59], which strengthens their role as a complementary tool in welfare assessment.

These data lay the groundwork for estimating the animals’ actual noise exposure and for proposing operational limits. The discussion presented by [28] stands out, in which the difficulty of separating coughing sounds from background noise in production environments is detailed. Years later, ref. [53] provided a means of differentiating vocalizations from environmental interferences such as noise generated by fans. These studies demonstrate the challenge of analyzing vocalizations when mechanical and biological signals coexist, reinforcing the importance of acoustic source differentiation in welfare-focused research.

Exposure to unpleasant or unwanted sounds, which are defined as noise, may act as a relevant environmental stressor with physical and psychological effects [84]. Although not part of the 36 articles included in this review, ref. [85] highlighted that anthropogenic noise may lead to behavioral alterations, disrupted sleep, and even hearing damage in livestock animals. Additionally, ref. [86] reported that heifers show a clear preference for quiet environments over noisy milking parlors, as indicated by increased heart rate in the latter. These findings underscore the dual role of sound as both a potential stressor and a tool for welfare improvement in controlled environments.

Finally, the steady increase in research activity shown in Figure 4 may be attributed to enhanced accessibility to recording devices, as well as significant advances in signal processing and machine learning in industrial settings [87]. In particular, the rise in artificial intelligence has led to new applications in acoustic monitoring of animals, a trend clearly reflected in recent studies [88]. Institutional interest in promoting PLF has also strengthened this line of research. A notable example is the European Precision Livestock Farming (EU-PLF) initiative, launched in 2012, which brought together universities and companies, such as SoundTalks NV, with the aim of transferring PLF technologies to the industrial scale [89].

4.2. Animal Production Context

The histogram in Figure 5 suggests that slaughterhouses processing multiple animal species may face additional challenges related to traceability, since animals arriving at the facility can differ in characteristics such as breed and age. Additionally, Figure 6 reveals a low representation of acoustic research conducted in slaughterhouse facilities among the 36 selected articles. This knowledge gap may represent a significant opportunity for future studies, especially considering that the acoustic environment in slaughterhouses can have an impact on animal welfare. Evidence of this are the studies presented by [60,90] who observed handling resistance behaviors associated with noise from a noisy truck near the facility and documented sound intensity levels of up to 100 dB that could affect both animal and staff welfare.

In this sense, early detection of stress indicators through monitoring technology that combines sound sensors and artificial intelligence algorithms could represent an effective solution to mitigate animal welfare problems. This facilitates timely interventions and helps optimize inspection and quality control processes. A successful example is the STREMODO system, which employs sound analysis and neural networks to detect stress-related vocalizations in swine [91]. Furthermore, recent research shows that sensors integrated with artificial intelligence and acoustic analysis can identify anomalies by detecting vocal patterns associated with animal stress [92].

In the same way, the larger number of studies that examine the farrowing stage, many centered on piglets, could reflect the intensive handling routine procedures associated with pain, and the weaning stress characteristic of this period. A similar rationale may explain the attention given to finishing pigs, where management practices in the final production phase, including those that precede slaughter, raise specific welfare concerns. In comparison, the nursery stage has received less attention, leaving limited knowledge about the environmental and acoustic conditions that pigs encounter soon after weaning. Further research could fill this gap.

In cattle, acoustic research has mainly focused on the study of vocalizations as a tool for assessing physiological and emotional states. Complementing this line of research, ref. [93] demonstrated that the use of music significantly increased the voluntary approach of dairy cows to the pre-milking area, indicating a behavioral improvement associated with a favorable acoustic environment. Despite the advances observed in the dairy sector, the reviewed studies show limited attention to heifers and few investigations addressing acoustic aspects in beef cattle, particularly during critical stages such as finishing, where stress levels are considerable and could be assessed through specific sound patterns.

Although attempts have been made to use vocalizations as welfare indicators in beef cattle, the results have been limited and inconclusive. A meta-analysis on vocalizations during castration in beef cattle identified a high risk of bias, largely due to the lack of objective methods and the subjectivity involved in acoustic interpretation [94]. Similarly, ref. [95] found insufficient evidence to support the use of vocalizations as clear indicators of welfare in these animals.

Considering that acoustic research in cattle farming is still developing, it is necessary to apply multivariate analyses that enable an integrated evaluation of both animal welfare indicators and noise emissions in production facilities [96]. This methodological approach would support informed decision-making to optimize the acoustic environment, ensuring acceptable sound levels for animals in production and improving overall welfare conditions.

In poultry production, the ambiguous classification regarding age or physiological stage, labelled as “unknown”, limits comparability between studies. Given the short life cycle of broilers, typically around 35 to 45 days, it is possible that studies with no phase specification may have covered the entire cycle. Nonetheless, it is important to standardize the description of production stages in order to strengthen scientific rigor and enable more comparative analyses.

It is also worth noting that differences in microphone placement and calibrations in swine, short production cycles of poultry, and the underrepresentation of beef cattle studies result in heterogeneous datasets that cannot be directly correlated with one another. This variability underscores the need for harmonized methods that allow more robust cross-species comparisons.

Chloupek et al. [52] studied how sudden noise affects a range of physiological stress markers in livestock. Expanding on that work, ref. [97] developed a non-invasive system for broilers that pairs pecking sounds with video footage to identify feeding behaviors, sparing the birds the disturbance of direct human observation.

Bioacoustic tools now also support remote welfare monitoring. For example, studies with laying hens have proposed automatic vocalization detectors to gauge thermal comfort [98]. Sun et al. [99] created a model that tracks the health of white-feathered broilers through their calls, while [100] used acoustic analysis to flag avian influenza at an early stage, achieving accuracy between 84% and 90%.

These advances show that bioacoustics in poultry production has moved steadily toward automated, non-invasive methods that spot early changes in health and welfare. The growing use of real-time sound capture and analysis in birds is now beginning to take hold in other livestock sectors as well.

4.3. Sound Capture Technologies

Early microphone and recorder setups were simple but reached 40 kHz, capturing the full spectral range of animal vocalizations. With direct digital recording, data handling became easier, and post processing grew more detailed. The transition to solid state recorders, with larger memory and higher sampling rates, let researchers archive long, high quality sessions with minimal data loss. Thanks to this upgrade, complex events, like the vocalizations produced by piglets during castration, can now be captured without losing fine acoustic detail [68].

Nevertheless, the limited reporting of calibration procedures observed in numerous studies may reduce reproducibility and comparability. On the other hand, the variability in device placement found in the 36 articles could reflect the adaptation of methods to the type of signal to be captured and the specific acoustic context of each installation.

Differences in the proportion of recording technologies across animal categories, as shown in Figure 8, may be attributed to a combination of factors, including the behavioral characteristics of each species, the feasibility of using certain devices without interfering with production activities, and the objectives of each study. For instance, research aiming to link animal behavior with welfare through vocalization analysis may rely primarily on video cameras, whereas studies focused on identifying the frequency content of animal vocalizations may prioritize devices such as microphones.

The combined use of microphones with video cameras, or microphones with sound level meters, allows for a more comprehensive characterization of the acoustic environment and the behavioral or physiological responses of the animals. Although microphones and sound level meters remain the most widely used technologies (Figure 9), the incorporation of emerging tools is noteworthy. These include mobile applications for sound measurement [60] and automatic vocal pattern recognition models [62]. Such innovations reflect a growing trend towards automation, portability, and real-time analysis in acoustic monitoring systems for animal production. Nevertheless, the limited number of studies per animal category poses a constraint in this review, preventing definitive conclusions about trends in the use of recording technologies across species.

4.4. Variables and Parameters Assessed

Heterogeneous units and parameterizations of acoustic variables impede direct comparison across studies, indicating the need for standardized protocols. Figure 10 places frequency at the center of the network, showing its ability to connect diverse acoustic dimensions and to serve as a key parameter in animal-production sound analysis. By contrast, ‘Environmental conditions’ and ‘Body weight’ display limited connectivity, mainly to ‘Physiology/health’ and ‘Frequency’, suggesting room to strengthen links between acoustic metrics and environmental and physiological variables.

Taken together, the results presented in the network plot (Figure 10) and the UpSet plot (Figure 11) underscore the importance of advancing towards more integrated experimental designs that simultaneously incorporate detailed acoustic metrics alongside biological, physiological, and environmental variables. Through this multivariable approach, sound could become a reliable and robust tool to support the objective assessment of health and welfare in animal production systems.

In this regard, the classification presented in Table 2 and Table 3 can serve as a basis to guide future work by clustering non-acoustic and acoustic parameters within meaningful groups, useful when designing standardized terms and measuring procedures. While this review is not in a position to provide a detailed standardization, it highlights key aspects that should be consistently reported, such as calibration methods, sampling rates, and frequency ranges. These are key points crucial when seeking greater comparability and reproducibility between species and scenarios of production and for moving toward more unified approaches in future research.

4.5. Key Findings and Perspectives of the Articles

Evidence shows that acoustic analysis is gaining ground as a tool for assessing animal welfare, even though important methodological gaps persist. In swine, several studies link high-pitched, prolonged screams to intense pain or stress during routine procedures such as castration and ear tagging [63,68,73,76,79]. In cattle, vocal behavior varies by phenotype and can reveal how animals perceive a situation, yet these shifts rarely alter the acoustic properties of the calls [77]. Studies with Holstein-Friesian cows show that peripartum calls combine sound and posture, open mouth and tongue protrusion, to project the signal over longer distances, a feature thought to reflect the urgency of calving [57].

In poultry, the call’s characteristics change with age and body weight, making them potential indicators of health and growth. Broilers, for instance, show a fall in peak frequency as they mature [54], while studies in laying hens have isolated acoustic traits that let algorithms sort calls into distinct categories [53]. Recurrent neural networks trained on chick vocalizations can even pick up early behavioral patterns [62]. Still, many investigations do not cross-validate sound data with physiology, and few compare results across breeds, ages, or management systems [64,67].

Acoustic monitoring is proving useful for health surveillance as well. Trials with swine show that classifiers built from cough recordings can tell sick animals from healthy pen mates [29,70]. In a related study, ref. [65] trained support vector machines to recognize sneezes linked to swine influenza and reported accuracy levels approaching 100 percent. Similar work in dairy calves isolates coughs by frequency and duration [28]. Though outside the scope of this review, broiler studies report more than 80 percent accuracy in spotting diseases such as infectious bronchitis and Newcastle disease [101].

However, the adoption of these tools in field conditions faces significant limitations. Background noise from equipment such as fans or engines, along with high inter-individual variability, can compromise the robustness of algorithms, which are often trained on limited datasets under controlled conditions. This challenge is also evident in animal production systems, where peak noise levels often occur during routine procedures or stressful events (such as handling or pre-slaughter operations), occasionally exceeding 120 dB, even when daily averages remain below 85 dB(A) [60,61,75,78].

Noise levels also shift with time of day, season, and wind direction [71,72], whereas barn design seems to play a lesser role [42,43]. Passive measures can dampen sound [82], and mobile apps now permit quick, low-cost audits in slaughterhouses [60]. Precision-livestock projects are starting to add acoustic streams to real-time dashboards that track stress, growth, and behavior [54,55,62,66].

Even so, high background noise and lower call rates in older animals often reduce system accuracy. These issues argue for multimodal setups that blend audio with computer vision and for automated labelling workflows that toughen models against on-farm variability. It is also worth noting that control equipment meant to improve climate can itself become a stressor, especially for pigs [74]. By contrast, the acoustic sensors themselves (e.g., microphones, sound level meters) are non-invasive tools and have not been reported to cause direct adverse health effects on animals.

Bringing noise readings together with temperature, humidity and gas measurements will allow producers to set combined thresholds that reflect the true environmental load animals experience. For sound to become a routine metric in precision farming, researchers still need to link key acoustic patterns to physiological responses, test their methods under a wider variety of farm conditions and strengthen the algorithms so they keep working despite the normal background clatter of a barn.

It should also be noted that aspects related to cost, scalability, and long-term practicality of acoustic monitoring technologies were not explicitly addressed in most of the reviewed studies. This represents a relevant gap, as the economic feasibility of these tools will be critical for their effective adoption in commercial farming systems.

In addition, the synthesis indicated a predominantly low risk of bias. Even so, deficiencies clustered in Q5 and Q6, where higher “No”/”Unclear” rates suggest that confounders were not consistently identified or addressed. Clear prespecification and transparent handling of confounders are recommended to mitigate this source of bias.

Although recent advances have incorporated machine learning (ML) into acoustic monitoring, many of these papers were excluded from this review because they reported sound data only as input variables without providing details about the recording protocols or acoustic parameters. This exclusion was necessary to preserve methodological comparability among the selected studies, yet it also highlights an important limitation: promising ML approaches could not be systematically assessed. Future systematic reviews specifically focused on ML applications may be able to address this gap by compiling and comparing quantitative performance metrics (e.g., accuracy, sensitivity, specificity) across studies.

In light of the heterogeneity observed across studies, generating quantitative recommendations such as specified sampling rates or general equipment configurations was not feasible. Even so, the synthesis identifies clear priorities for future work. These are the need to adopt common reporting conventions (e.g., calibrating methods, ranges of frequency, and device placement locations), and frequent joining of acoustic parameters and physiological markers for verification. These would then permit more robust and comparable datasets that can inform future meta-analyses to distinguish species-specific technical procedures and operating blueprints.

5. Conclusions

This systematic review provided an updated and detailed overview of research assessing sound in animal production systems, particularly in cattle, poultry, and swine farming. The findings show that scientific interest in acoustic analysis has grown steadily in recent decades, driven by advancements in capture technologies, progress in signal processing methods, and the strengthening of the precision livestock farming approach. Four main research themes were identified: (1) vocalizations as indicators of animal welfare, (2) characterization of sound sources, (3) combined analysis of sound sources and vocalizations, and (4) animal responses to acoustic stimuli.

Even so, several methodological hurdles still block routine use on commercial farms. Studies rely on a wide mix of metrics, follow no shared rules for calibrating equipment or collecting data, and seldom combine sound with physiological or environmental measures. Most models are built on narrow datasets and lose accuracy when tested in real barns. Addressing these gaps will call for stricter experimental design, larger and more varied databases and sturdier instruments that can cope with the noise and other interference found in production settings.

In this context, it is essential to move towards the development of technologies that are more accessible, replicable, and suited to field conditions, allowing for the effective integration of acoustic analysis with the evaluation of relevant environmental parameters. Moreover, enhancing the linkage between acoustic signals and physiological indicators remains a key challenge in establishing sound as a reliable biomarker of welfare. The evolution of automated systems for sound capture and analysis offers a strategic opportunity to strengthen both acoustic and environmental monitoring, contributing to more efficient, sustainable, and welfare-focused animal production systems.

Nonetheless, this study has certain limitations related to the eligibility criteria applied. Only articles published in English were included, which may have excluded relevant research disseminated in other languages, particularly from regions with strong local scientific production. Studies that used acoustic data as input for ML models were excluded if they did not report quantitative metrics directly associated with sound or with relevant environmental and physiological variables. This decision aimed to ensure a sufficient technical characterization of measurement methods, rather than limiting the analysis to the predictive performance of the models. Future reviews could consider expanding these criteria to explore more deeply the role of artificial intelligence in acoustic monitoring.

The results of this review indicate that in-practice application of acoustic monitoring is possible in livestock farming. Sound analysis provides in-real-time, non-invasive information that is supplementary to environmental and physiological variables and aids in the early detection of stress or disease and contributes to automated responses in the system of Precision Livestock Farming. Recent advances, such as the use of neural networks to recognize vocalizations or mobile devices for on-farm sound measurement, illustrate that these tools are moving from the laboratory into everyday farm routines. Extending their use in real conditions will help transform current research into practical strategies that improve animal welfare, productivity, and the sustainability of farming systems.