Search Results (13)

Voice conversion (VC) is an advanced technology that enables the transformation of raw speech into high-quality audio resembling the target speaker’s voice while preserving the original linguistic content and prosodic patterns. In this study, we propose a voice conversion algorithm, Multi-Dimensional Perception Flow Matching (MPFM-VC). Unlike traditional approaches that directly generate waveform outputs, MPFM-VC models the evolutionary trajectory of mel spectrograms with a flow-matching framework and incorporates a multi-dimensional feature perception network to enhance the stability and quality of speech synthesis. Additionally, we introduce a content perturbation method during training to improve the model’s generalization ability and reduce inference-time artifacts. To further increase speaker similarity, an adversarial training mechanism on speaker embeddings is employed to achieve effective disentanglement between content and speaker identity representations, thereby enhancing the timbre consistency of the converted speech. Experimental results for both speech and singing voice conversion tasks show that MPFM-VC achieves competitive performance compared to existing state-of-the-art VC models in both subjective and objective evaluation metrics. The synthesized speech shows improved naturalness, clarity, and timbre fidelity in both objective and subjective evaluations, suggesting the potential effectiveness of the proposed approach. Full article

Advancements in audio synthesis and manipulation technologies have reshaped applications such as personalised virtual assistants, voice cloning for creative content, and language learning tools. However, the misuse of these technologies to create audio deepfakes has raised serious concerns about security, privacy, and trust. Studies reveal that human judgement of deepfake audio is not always reliable, highlighting the urgent need for robust detection technologies to mitigate these risks. This paper provides a comprehensive survey of recent advancements in audio deepfake detection, with a focus on cutting-edge developments in the past few years. It begins by exploring the foundational methods of audio deepfake generation, including text-to-speech (TTS) and voice conversion (VC), followed by a review of datasets driving progress in the field. The survey then delves into detection approaches, covering frontend feature extraction, backend classification models, and end-to-end systems. Additionally, emerging topics such as privacy-preserving detection, explainability, and fairness are discussed. Finally, this paper identifies key challenges and outlines future directions for developing robust and scalable audio deepfake detection systems. Full article

Introduction: Dysarthria, a motor speech disorder caused by neurological damage, significantly hampers speech intelligibility, creating communication barriers for affected individuals. Voice conversion (VC) systems have been developed to address this, yet accurately predicting phonemes in dysarthric speech remains a challenge due to its variability. This study proposes a novel approach that integrates Fuzzy Expectation Maximization (FEM) with diffusion models for enhanced phoneme prediction, aiming to improve the quality of dysarthric voice conversion. Methods: The proposed method combines FEM clustering with Diffusion Probabilistic Models (DPM). Diffusion models simulate noise addition and removal to enhance the robustness of speech signals, while FEM iteratively optimizes phoneme boundaries, reducing uncertainty. The system was trained using the Saarland University Voice Disorder dataset, consisting of dysarthric and normal speech samples, with the conversion process represented in the Mel-spectrogram domain. The framework employs both subjective (Mean Opinion Score, MOS) and objective (Word Error Rate, WER) metrics for evaluation, complemented by ablation studies. Results: Experimental results showed that the proposed method significantly improved phoneme prediction accuracy and overall voice conversion quality. It achieved higher MOSs for naturalness, intelligibility, and speaker similarity compared to existing models like StarGAN-VC and CycleGAN-VC. Additionally, the proposed method demonstrated a lower WER for both mild and severe dysarthria cases, indicating better performance in producing intelligible speech. Discussion: The integration of FEM with diffusion models offers substantial improvements in handling the irregularities of dysarthric speech. The method’s robustness, as evidenced by the ablation studies, shows that it can maintain speech naturalness and intelligibility even without a speaker-encoder. These findings suggest that the proposed approach can contribute to the development of more reliable assistive communication technologies for individuals with dysarthria, providing a promising foundation for future advancements in personalized speech therapy. Full article

Voice conversion (VC) refers to the technique of modifying one speaker’s voice to mimic another’s while retaining the original linguistic content. This technology finds its applications in fields such as speech synthesis, accent modification, medicine, security, privacy, and entertainment. Among the various deep generative models used for voice conversion, including variational autoencoders (VAEs) and generative adversarial networks (GANs), diffusion models (DMs) have recently gained attention as promising methods due to their training stability and strong performance in data generation. Nevertheless, traditional DMs focus mainly on learning reconstruction paths like VAEs, rather than conversion paths as GANs do, thereby restricting the quality of the converted speech. To overcome this limitation and enhance voice conversion performance, we propose a cycle-consistent diffusion (CycleDiffusion) model, which comprises two DMs: one for converting the source speaker’s voice to the target speaker’s voice and the other for converting it back to the source speaker’s voice. By employing two DMs and enforcing a cycle consistency loss, the CycleDiffusion model effectively learns both reconstruction and conversion paths, producing high-quality converted speech. The effectiveness of the proposed model in voice conversion is validated through experiments using the VCTK (Voice Cloning Toolkit) dataset. Full article

Speech is a crucial source of personal information, and the risk of attackers using such information increases day by day. Speaker privacy protection is crucial, and various approaches have been proposed to hide the speaker’s identity. One approach is voice anonymization, which aims to safeguard speaker identity while maintaining speech content through techniques such as voice conversion or spectral feature alteration. The significance of voice anonymization has grown due to the necessity to protect personal information in applications such as voice assistants, authentication, and customer support. Building upon the S3PRL-VC toolkit and on pre-trained speech and speaker representation models, this paper introduces a feature disentanglement approach to improve the de-identification performance of the state-of-the-art anonymization approaches based on voice conversion. The proposed approach achieves state-of-the-art speaker de-identification and causes minimal impact on the intelligibility of the signal after conversion. Full article

We present an any-to-one voice conversion (VC) system, using an autoregressive model and LPCNet vocoder, aimed at enhancing the converted speech in terms of naturalness, intelligibility, and speaker similarity. As the name implies, non-parallel any-to-one voice conversion does not require paired source and target speeches and can be employed for arbitrary speech conversion tasks. Recent advancements in neural-based vocoders, such as WaveNet, have improved the efficiency of speech synthesis. However, in practice, we find that the trajectory of some generated waveforms is not consistently smooth, leading to occasional voice errors. To address this issue, we propose to use an autoregressive (AR) conversion model along with the high-fidelity LPCNet vocoder. This combination not only solves the problems of waveform fluidity but also produces more natural and clear speech, with the added capability of real-time speech generation. To precisely represent the linguistic content of a given utterance, we use speaker-independent PPG features (SI-PPG) computed from an automatic speech recognition (ASR) model trained on a multi-speaker corpus. Next, a conversion model maps the SI-PPG to the acoustic representations used as input features for the LPCNet. The proposed autoregressive structure enables our system to produce the following prediction step outputs from the acoustic features predicted in the previous step. We evaluate the effectiveness of our system by performing any-to-one conversion pairs between native English speakers. Experimental results show that the proposed method outperforms state-of-the-art systems, producing higher speech quality and greater speaker similarity. Full article

Voice spoofing attempts to break into a specific automatic speaker verification (ASV) system by forging the user’s voice and can be used through methods such as text-to-speech (TTS), voice conversion (VC), and replay attacks. Recently, deep learning-based voice spoofing countermeasures have been developed. However, the problem with replay is that it is difficult to construct a large number of datasets because it requires a physical recording process. To overcome these problems, this study proposes a pre-training framework based on multi-order acoustic simulation for replay voice spoofing detection. Multi-order acoustic simulation utilizes existing clean signal and room impulse response (RIR) datasets to generate audios, which simulate the various acoustic configurations of the original and replayed audios. The acoustic configuration refers to factors such as the microphone type, reverberation, time delay, and noise that may occur between a speaker and microphone during the recording process. We assume that a deep learning model trained on an audio that simulates the various acoustic configurations of the original and replayed audios can classify the acoustic configurations of the original and replay audios well. To validate this, we performed pre-training to classify the audio generated by the multi-order acoustic simulation into three classes: clean signal, audio simulating the acoustic configuration of the original audio, and audio simulating the acoustic configuration of the replay audio. We also set the weights of the pre-training model to the initial weights of the replay voice spoofing detection model using the existing replay voice spoofing dataset and then performed fine-tuning. To validate the effectiveness of the proposed method, we evaluated the performance of the conventional method without pre-training and proposed method using an objective metric, i.e., the accuracy and F1-score. As a result, the conventional method achieved an accuracy of 92.94%, F1-score of 86.92% and the proposed method achieved an accuracy of 98.16%, F1-score of 95.08%. Full article

Voice conversion (VC) consists of digitally altering the voice of an individual to manipulate part of its content, primarily its identity, while maintaining the rest unchanged. Research in neural VC has accomplished considerable breakthroughs with the capacity to falsify a voice identity using a small amount of data with a highly realistic rendering. This paper goes beyond voice identity manipulation and presents an original neural architecture that allows the manipulation of voice attributes (e.g., gender and age). The proposed architecture is inspired by the fader network, transferring the same ideas to voice manipulation. The information conveyed by the speech signal is disentangled into interpretative voice attributes by means of minimizing adversarial loss to make the encoded information mutually independent while preserving the capacity to generate a speech signal from the disentangled codes. During inference for voice conversion, the disentangled voice attributes can be manipulated and the speech signal can be generated accordingly. For experimental evaluation, the proposed method is applied to the task of voice gender conversion using the freely available VCTK dataset. Quantitative measurements of mutual information between the variables of speaker identity and speaker gender show that the proposed architecture can learn gender-independent representation of speakers. Additional measurements of speaker recognition indicate that speaker identity can be recognized accurately from the gender-independent representation. Finally, a subjective experiment conducted on the task of voice gender manipulation shows that the proposed architecture can convert voice gender with very high efficiency and good naturalness. Full article

Electro-laryngeal (EL) speech has poor intelligibility and naturalness, which hampers the popular use of the electro-larynx. Voice conversion (VC) can enhance EL speech. However, if the EL speech to be enhanced is with complicated tone variation rules in Mandarin, the enhancement will be less effective. This is because the source speech (Mandarin EL speech) and the target speech (normal speech) are not strictly parallel. We propose using cycle-consistent generative adversarial networks (CycleGAN, a parallel-free VC framework) to enhance continuous Mandarin EL speech, which can solve the above problem. In the proposed framework, the generator is designed based on the neural networks of a 2D-Conformer-1D-Transformer-2D-Conformer. Then, we used Mel-Spectrogram instead of traditional acoustic features (fundamental frequency, Mel-Cepstrum parameters and aperiodicity parameters). At last, we converted the enhanced Mel-Spectrogram into waveform signals using WaveNet. We undertook both subjective and objective tests to evaluate the proposed approach. Compared with traditional approaches to enhance continuous Mandarin EL speech with variable tone (the average tone accuracy being 71.59% and average word error rate being 10.85%), our framework increases the average tone accuracy by 12.12% and reduces the average errors of word perception by 9.15%. Compared with the approaches towards continuous Mandarin EL speech with fixed tone (the average tone accuracy being 29.89% and the average word error rate being 10.74%), our framework increases the average tone accuracy by 42.38% and reduces the average errors of word perception by 8.59%. Our proposed framework can effectively address the problem that the source and target speech are not strictly parallel. The intelligibility and naturalness of Mandarin EL speech have been further improved. Full article

The goal of emotional voice conversion (EVC) is to convert the emotion of a speaker’s voice from one state to another while maintaining the original speaker’s identity and the linguistic substance of the message. Research on EVC in the Arabic language is well behind that conducted on languages with a wider distribution, such as English. The primary objective of this study is to determine whether Arabic emotions may be converted using a model trained for another language. In this work, we used an unsupervised many-to-many non-parallel generative adversarial network (GAN) voice conversion (VC) model called StarGANv2-VC to perform an Arabic EVC (A-EVC). The latter is realized by using pre-trained phoneme-level automatic speech recognition (ASR) and fundamental frequency (F0) models in the English language. The generated voice is evaluated by prosody and spectrum conversion in addition to automatic emotion recognition and speaker identification using a convolutional recurrent neural network (CRNN). The results of the evaluation indicated that male voices were scored higher than female voices and that the evaluation score for the conversion from neutral to other emotions was higher than the evaluation scores for the conversion of other emotions. Full article

Laryngectomees are individuals whose larynx has been surgically removed, usually due to laryngeal cancer. The immediate consequence of this operation is that these individuals (laryngectomees) are unable to speak. Esophageal speech (ES) remains the preferred alternative speaking method for laryngectomees. However, compared to the laryngeal voice, ES is characterized by low intelligibility and poor quality due to chaotic fundamental frequency $F0$, specific noises, and low intensity. Our proposal to solve these problems is to take advantage of voice conversion as an effective way to improve speech quality and intelligibility. To this end, we propose in this work a novel esophageal–laryngeal voice conversion (VC) system based on a sequence-to-sequence (Seq2Seq) model combined with an auditory attention mechanism. The originality of the proposed framework is that it adopts an auditory attention technique in our model, which leads to more efficient and adaptive feature mapping. In addition, our VC system does not require the classical DTW alignment process during the learning phase, which avoids erroneous mappings and significantly reduces the computational time. Moreover, to preserve the identity of the target speaker, the excitation and phase coefficients are estimated by querying a binary search tree. In experiments, objective and subjective tests confirmed that the proposed approach performs better even in some difficult cases in terms of speech quality and intelligibility. Full article

Voice conversion (VC) transforms the speaking style of a source speaker to the speaking style of a target speaker by keeping linguistic information unchanged. Traditional VC techniques rely on parallel recordings of multiple speakers uttering the same sentences. Earlier approaches mainly find a mapping between the given source–target speakers, which contain pairs of similar utterances spoken by different speakers. However, parallel data are computationally expensive and difficult to collect. Non-parallel VC remains an interesting but challenging speech processing task. To address this limitation, we propose a method that allows a non-parallel many-to-many voice conversion by using a generative adversarial network. To the best of the authors’ knowledge, our study is the first one that employs a sinusoidal model with continuous parameters to generate converted speech signals. Our method involves only several minutes of training examples without parallel utterances or time alignment procedures, where the source–target speakers are entirely unseen by the training dataset. Moreover, empirical study is carried out on the publicly available CSTR VCTK corpus. Our conclusions indicate that the proposed method reached the state-of-the-art results in speaker similarity to the utterance produced by the target speaker, while suggesting important structural ones to be further analyzed by experts. Full article

In voice conversion (VC), it is highly desirable to obtain transformed speech signals that are perceptually close to a target speaker’s voice. To this end, a perceptually meaningful criterion where the human auditory system was taken into consideration in measuring the distances between the converted and the target voices was adopted in the proposed VC scheme. The conversion rules for the features associated with the spectral envelope and the pitch modification factor were jointly constructed so that perceptual distance measurement was minimized. This minimization problem was solved using a deep neural network (DNN) framework where input features and target features were derived from source speech signals and time-aligned version of target speech signals, respectively. The validation tests were carried out for the CMU ARCTIC database to evaluate the effectiveness of the proposed method, especially in terms of perceptual quality. The experimental results showed that the proposed method yielded perceptually preferred results compared with independent conversion using conventional mean-square error (MSE) criterion. The maximum improvement in perceptual evaluation of speech quality (PESQ) was 0.312, compared with the conventional VC method. Full article