Search Results (11)

Electron ionization (EI) mass spectrometry (MS) is a widely used technique for the compound identification and production of spectra. However, incomplete coverage of reference spectral libraries limits reliable analysis of newly characterized molecules. This study presents a hybrid deep learning model for predicting EI-MS spectra directly from molecular structure. The approach combines a graph neural network encoder with a residual neural network decoder, followed by refinement using cross-attention, bidirectional prediction, and probabilistic, chemistry-informed masks. Trained on the NIST14 EI-MS database (≤500 Da), the model achieves strong library matching performance (Recall@10 ≈ 80.8%) and high spectral similarity. The proposed hybrid GNN (Graph Neural Network)-ResNet (Residual Neural Network) model can generate high-quality synthetic EI-MS spectra to supplement existing libraries, potentially reducing the cost and effort of experimental spectrum acquisition. The obtained results demonstrate the potential of data-driven models to augment EI-MS libraries, while highlighting remaining challenges in generalization and spectral uniqueness. Full article

Data augmentation techniques are widely applied in speech emotion recognition to increase the diversity of data and enhance the performance of models. However, existing research has not deeply explored the impact of these data augmentation techniques on emotional data. Inappropriate augmentation algorithms may distort emotional labels, thereby reducing the performance of models. To address this issue, in this paper we systematically evaluate the influence of common data augmentation algorithms on emotion recognition from three dimensions: (1) we design subjective auditory experiments to intuitively demonstrate the impact of augmentation algorithms on the emotional expression of speech; (2) we jointly extract multi-dimensional features from spectrograms based on the Librosa library and analyze the impact of data augmentation algorithms on the spectral features of speech signals through heatmap visualization; and (3) we objectively evaluate the recognition performance of the model by means of indicators such as cross-entropy loss and introduce statistical significance analysis to verify the effectiveness of the augmentation algorithms. The experimental results show that “time stretching” may distort speech features, affect the attribution of emotional labels, and significantly reduce the model’s accuracy. In contrast, “reverberation” (RIR) and “resampling” within a limited range have the least impact on emotional data, enhancing the diversity of samples. Moreover, their combination can increase accuracy by up to 7.1%, providing a basis for optimizing data augmentation strategies. Full article

Precision agriculture relies on highly detailed soil maps to optimize resource use. Proximal sensing methods, such as EMI, require a certain number of soil samples and laboratory analysis to interpolate the characteristics of the soil. NIR diffuse reflectance spectroscopy offers a rapid, low-cost alternative that increases datapoints and map accuracy. This study tests and optimizes a methodology for high-detail soil mapping in a 2.5 ha hazelnut grove in Grosseto, Southern Tuscany, Italy, using both EMI sensors (GF Mini Explorer, Brno, Czech Republic) and a handheld NIR spectrometer (Neospectra Scanner, Si-Ware Systems, Menlo Park, CA, USA). In addition to two profiles selected by clustering, another 35 topsoil augerings (0–30 cm) were added. Laboratory analyses were performed on only five samples (two profiles + three samples from the augerings). Partial least square regression (PLSR) with a national spectral library, augmented by the five local samples, predicted clay, sand, organic carbon (SOC), total nitrogen (TN), and cation exchange capacity (CEC). The 37 predicted datapoints were used for spatial interpolation, using the ECa map, elevation, and DEM derivatives as covariates. Kriging with external drift (KED) was used to spatialize the results. The errors of the predictive maps were calculated using five additional validation points analyzed by conventional methods. The validation showed good accuracy of the predictive maps, particularly for SOC and TN. Full article

The composition of clay minerals in soils, and more particularly the presence of montmorillonite (as part of the smectite family), is a key factor in soil swell–shrinking as well as off–road vehicle mobility. Detecting these topsoil clay minerals and quantifying the montmorillonite abundance are a challenge since they are usually intimately mixed with other minerals, soil organic carbon and soil moisture content. Imaging spectroscopy coupled with unmixing methods can address these issues, but the quality of the estimation degrades the coarser the spatial resolution is due to pixel heterogeneity. With the advent of UAV-borne and proximal hyperspectral acquisitions, it is now possible to acquire images at a centimeter scale. Thus, the objective of this paper is to evaluate the accuracy and limitations of unmixing methods to retrieve montmorillonite abundance from very-high-resolution hyperspectral images (1.5 cm) acquired from a camera installed on top of a bucket truck over three different agricultural fields, in Loiret department, France. Two automatic endmember detection methods based on the assumption that materials are linearly mixed, namely the Simplex Identification via Split Augmented Lagrangian (SISAL) and the Minimum Volume Constrained Non-negative Matrix Factorization (MVC-NMF), were tested prior to unmixing. Then, two linear unmixing methods, the fully constrained least square method (FCLS) and the multiple endmember spectral mixture analysis (MESMA), and two nonlinear unmixing ones, the generalized bilinear method (GBM) and the multi-linear model (MLM), were performed on the images. In addition, several spectral preprocessings coupled with these unmixing methods were applied in order to improve the performances. Results showed that our selected automatic endmember detection methods were not suitable in this context. However, unmixing methods with endmembers taken from available spectral libraries performed successfully. The nonlinear method, MLM, without prior spectral preprocessing or with the application of the first Savitzky–Golay derivative, gave the best accuracies for montmorillonite abundance estimation using the USGS library (RMSE between 2.2–13.3% and 1.4–19.7%). Furthermore, a significant impact on the abundance estimations at this scale was in majority due to (i) the high variability of the soil composition, (ii) the soil roughness inducing large variations of the illumination conditions and multiple surface scatterings and (iii) multiple volume scatterings coming from the intimate mixture. Finally, these results offer a new opportunity for mapping expansive soils from imaging spectroscopy at very high spatial resolution. Full article

In-situ infrared soil spectroscopy is prone to the effects of ambient factors, such as moisture, shadows, or roughness, resulting in measurements of compromised quality, which is amplified when multiple sensors are used for data collection. Aiming to provide accurate estimations of common physicochemical soil properties, such as soil organic carbon (SOC), texture, pH, and calcium carbonates based on in-situ reflectance captured by a set of low-cost spectrometers operating at the shortwave infrared region, we developed an AI-based spectral transfer function that maps fields to laboratory spectra. Three test sites in Cyprus, Lithuania, and Greece were used to evaluate the proposed methodology, while the dataset was harmonized and augmented by GEO-Cradle regional soil spectral library (SSL). The developed dataset was used to calibrate and validate machine learning models, with the attained predictive performance shown to be promising for directly estimating soil properties in-situ, even with sensors with reduced spectral range. Aiming to set a baseline scenario, we completed the exact same modeling experiment under laboratory conditions and performed a one-to-one comparison between field and laboratory modelling accuracy metrics. SOC and pH presented an R² of 0.43 and 0.32 when modeling the in-situ data compared to 0.63 and 0.41 of the laboratory case, respectively, while clay demonstrated the highest accuracy with an R² value of 0.87 in-situ and 0.90 in the laboratory. Calcium carbonates were also attempted to be modeled at the studied spectral region, with the expected accuracy loss from the laboratory to the in-situ to be observable (R² = 0.89 for the laboratory and 0.67 for the in-situ) but the reduced dataset variability combined with the calcium carbonate characteristics that are spectrally active in the region outside the spectral range of the used in-situ sensor, induced low RPIQ values (less than 0.50), signifying the importance of the suitable sensor selection. Full article

Technological innovations and advanced multidisciplinary research increase the demand for multisensor data fusion in Earth observations. Such fusion has great potential, especially in the remote sensing field. One sensor is often insufficient in analyzing urban environments to obtain comprehensive results. Inspired by the capabilities of hyperspectral and Light Detection and Ranging (LiDAR) data in multisensor data fusion at the feature level, we present a novel approach to the multitemporal analysis of urban land cover in a case study in Høvik, Norway. Our generic workflow is based on bitemporal datasets; however, it is designed to include datasets from other years. Our framework extracts representative endmembers in an unsupervised way, retrieves abundance maps fed into segmentation algorithms, and detects the main urban land cover classes by implementing 2D ResU-Net for segmentation without parameter regularizations and with effective optimization. Such segmentation optimization is based on updating initial features and providing them for a second iteration of segmentation. We compared segmentation optimization models with and without data augmentation, achieving up to 11% better accuracy after segmentation optimization. In addition, a stable spectral library is automatically generated for each land cover class, allowing local database extension. The main product of the multitemporal analysis is a map update, effectively detecting detailed changes in land cover classes. Full article

Resolving small molecule mixtures by nuclear magnetic resonance (NMR) spectroscopy has been of great interest for a long time for its precision, reproducibility, and efficiency. However, spectral analyses for such mixtures are often highly challenging due to overlapping resonance lines and limited chemical shift windows. The existing experimental and theoretical methods to produce shift NMR spectra in dealing with the problem have limited applicability owing to sensitivity issues, inconsistency, and/or the requirement of prior knowledge. Recently, we resolved the problem by decoupling multiplet structures in NMR spectra by the wavelet packet transform (WPT) technique. In this work, we developed a scheme for deploying the method in generating highly resolved WPT NMR spectra and predicting the composition of the corresponding molecular mixtures from their ${}^1$H NMR spectra in an automated fashion. The four-step spectral analysis scheme consists of calculating the WPT spectrum, peak matching with a WPT shift NMR library, followed by two optimization steps in producing the predicted molecular composition of a mixture. The robustness of the method was tested on an augmented dataset of 1000 molecular mixtures, each containing 3 to 7 molecules. The method successfully predicted the constituent molecules with a median true positive rate of 1.0 against the varying compositions, while a median false positive rate of 0.04 was obtained. The approach can be scaled easily for much larger datasets. Full article

There is a need to minimize the usage of traditional laboratory reference methods in favor of spectroscopy for routine soil carbon monitoring, with potential cost savings existing especially for labile pools. Mid-infrared spectroscopy has been associated with accurate soil carbon predictions, but the method has not been researched extensively in connection to C lability. More studies are also needed on reducing the numbers of samples and on how to account for the compositional nature of C pools. This study compares performance of two classes of partial least squares regression models to predict soil carbon in a global (models trained to data from a spectral library), local (models trained to data from a target area), and calibration-spiking (spectral library augmented with target-area spectra) scheme. Topsoil samples were+ scanned with a Fourier-transform infrared spectrometer, total and hot-water extractable carbon determined, and isometric log-ratio coordinates derived from the latter measurements. The best RMSEP was estimated as 0.38 and 0.23 percentage points $\mathrm{TC}$ for the district and field scale, respectively—values sufficiently low to make only qualitative predictions according to the RPD and RPIQ criteria. Models estimating soil carbon lability performed unsatisfactorily, presumably due to low labile pool concentration. Traditional weighing of spiking samples by including multiple copies thereof in training data yielded better results than canonical partial least squares regression modeling with embedded weighing. Although local modeling was associated with the most accurate predictions, calibration spiking addressed better the trade-off between data acquisition costs and model quality. Calibration spiking with compositional data analysis is, therefore, recommended for routine monitoring. Full article

The ‘inverse problem’ of mass spectrometric molecular identification (‘given a mass spectrum, calculate/predict the 2D structure of the molecule whence it came’) is largely unsolved, and is especially acute in metabolomics where many small molecules remain unidentified. This is largely because the number of experimentally available electrospray mass spectra of small molecules is quite limited. However, the forward problem (‘calculate a small molecule’s likely fragmentation and hence at least some of its mass spectrum from its structure alone’) is much more tractable, because the strengths of different chemical bonds are roughly known. This kind of molecular identification problem may be cast as a language translation problem in which the source language is a list of high-resolution mass spectral peaks and the ‘translation’ a representation (for instance in SMILES) of the molecule. It is thus suitable for attack using the deep neural networks known as transformers. We here present MassGenie, a method that uses a transformer-based deep neural network, trained on ~6 million chemical structures with augmented SMILES encoding and their paired molecular fragments as generated in silico, explicitly including the protonated molecular ion. This architecture (containing some 400 million elements) is used to predict the structure of a molecule from the various fragments that may be expected to be observed when some of its bonds are broken. Despite being given essentially no detailed nor explicit rules about molecular fragmentation methods, isotope patterns, rearrangements, neutral losses, and the like, MassGenie learns the effective properties of the mass spectral fragment and valency space, and can generate candidate molecular structures that are very close or identical to those of the ‘true’ molecules. We also use VAE-Sim, a previously published variational autoencoder, to generate candidate molecules that are ‘similar’ to the top hit. In addition to using the ‘top hits’ directly, we can produce a rank order of these by ‘round-tripping’ candidate molecules and comparing them with the true molecules, where known. As a proof of principle, we confine ourselves to positive electrospray mass spectra from molecules with a molecular mass of 500Da or lower, including those in the last CASMI challenge (for which the results are known), getting 49/93 (53%) precisely correct. The transformer method, applied here for the first time to mass spectral interpretation, works extremely effectively both for mass spectra generated in silico and on experimentally obtained mass spectra from pure compounds. It seems to act as a Las Vegas algorithm, in that it either gives the correct answer or simply states that it cannot find one. The ability to create and to ‘learn’ millions of fragmentation patterns in silico, and therefrom generate candidate structures (that do not have to be in existing libraries) directly, thus opens up entirely the field of de novo small molecule structure prediction from experimental mass spectra. Full article

Sparse unmixing has been successfully applied in hyperspectral remote sensing imagery analysis based on a standard spectral library known in advance. This approach involves reformulating the traditional linear spectral unmixing problem by finding the optimal subset of signatures in this spectral library using the sparse regression technique, and has greatly improved the estimation of fractional abundances in ubiquitous mixed pixels. Since the potentially large standard spectral library can be given a priori, the most challenging task is to compute the regression coefficients, i.e., the fractional abundances, for the linear regression problem. There are many mathematical techniques that can be used to deal with the spectral unmixing problem; e.g., ordinary least squares (OLS), constrained least squares (CLS), orthogonal matching pursuit (OMP), and basis pursuit (BP). However, due to poor prediction accuracy and non-interpretability, the traditional methods often cannot obtain satisfactory estimations or achieve a reasonable interpretation. In this paper, to improve the regression accuracy of sparse unmixing, least angle regression-based constrained sparse unmixing (LARCSU) is introduced to further enhance the precision of sparse unmixing. Differing from the classical greedy algorithms and some of the cautious sparse regression-based approaches, the LARCSU algorithm has two main advantages. Firstly, it introduces an equiangular vector to seek the optimal regression steps based on the simple underlying geometry. Secondly, unlike the alternating direction method of multipliers (ADMM)-based algorithms that introduce one or more multipliers or augmented terms during their optimization procedures, no parameters are required in the computational process of the LARCSU approach. The experimental results obtained with both simulated datasets and real hyperspectral images confirm the effectiveness of LARCSU compared with the current state-of-the-art spectral unmixing algorithms. LARCSU can obtain a better fractional abundance map, as well as a higher unmixing accuracy, with the same order of magnitude of computational effort as the CLS-based methods. Full article

High resolution imaging spectroscopy data have been recognised as a valuable data resource for augmenting detailed material inventories that serve as input for various urban applications. Image-specific urban spectral libraries are successfully used in urban imaging spectroscopy studies. However, the regional- and sensor-specific transferability of such libraries is limited due to the wide range of different surface materials. With the developed methodology, incomplete urban spectral libraries can be utilised by assuming that unknown surface material spectra are dissimilar to the known spectra in a basic spectral library (BSL). The similarity measure SID-SCA (Spectral Information Divergence-Spectral Correlation Angle) is applied to detect image-specific unknown urban surfaces while avoiding spectral mixtures. These detected unknown materials are categorised into distinct and identifiable material classes based on their spectral and spatial metrics. Experimental results demonstrate a successful redetection of material classes that had been previously erased in order to simulate an incomplete BSL. Additionally, completely new materials e.g., solar panels were identified in the data. It is further shown that the level of incompleteness of the BSL and the defined dissimilarity threshold are decisive for the detection of unknown material classes and the degree of spectral intra-class variability. A detailed accuracy assessment of the pre-classification results, aiming to separate natural and artificial materials, demonstrates spectral confusions between spectrally similar materials utilizing SID-SCA. However, most spectral confusions occur between natural or artificial materials which are not affecting the overall aim. The dissimilarity analysis overcomes the limitations of working with incomplete urban spectral libraries and enables the generation of image-specific training databases. Full article