Search

Abstract Acoustic identification of vocalizing individuals opens up new and deeper insights into animal communications, such as individual-/group-specific dialects, turn-taking events, and dialogs. However, establishing an association between an individual animal and its emitted signal is usually non-trivial, especially for animals underwater. Consequently, a collection of marine species-, array-, and position-specific ground truth localization data is extremely challenging, which strongly limits possibilities to evaluate localization methods beforehand or at all. This study presents ORCA-SPY, a fully-automated sound source simulation, classification and localization framework for passive killer whale (Orcinus orca) acoustic monitoring that is embedded into PAMGuard, a widely used bioacoustic software toolkit. ORCA-SPY enables array- and position-specific multichannel audio stream generation to simulate real-world ground truth killer whale localization data and provides a hybrid sound source identification approach integrating ANIMAL-SPOT, a state-of-the-art deep learning-based orca detection network, followed by downstream Time-Difference-Of-Arrival localization. ORCA-SPY was evaluated on simulated multichannel underwater audio streams including various killer whale vocalization events within a large-scale experimental setup benefiting from previous real-world fieldwork experience. Across all 58,320 embedded vocalizing killer whale events, subject to various hydrophone array geometries, call types, distances, and noise conditions responsible for a signal-to-noise ratio varying from $$-14.2$$ - 14.2 dB to 3 dB, a detection rate of 94.0 % was achieved with an average localization error of 7.01 $$^\circ$$ ∘ . ORCA-SPY was field-tested on Lake Stechlin in Brandenburg Germany under laboratory conditions with a focus on localization. During the field test, 3889 localization events were observed with an average error of 29.19 $$^\circ$$ ∘ and a median error of 17.54 $$^\circ$$ ∘ . ORCA-SPY was deployed successfully during the DeepAL fieldwork 2022 expedition (DLFW22) in Northern British Columbia, with a mean average error of 20.01 $$^\circ$$ ∘ and a median error of 11.01 $$^\circ$$ ∘ across 503 localization events. ORCA-SPY is an open-source and publicly available software framework, which can be adapted to various recording conditions as well as animal species.

Oral diadochokinesis (oral-DDK) is one of the common maximum performance task used in clinical practice to evaluate the oral motor mechanism. Although there are reference values ​​for some languages, there are no recent publications of adult population speaking Spanish, German or Czech. Aims: 1) to describe the values ​​of oral-DDK in adult speakers of Spanish, German and Czech, 2) to determine the influence of gender on the values ​​of oral-DDK, 3) to determine the influence of age on the values ​​of oral-DDK and, 4) to compare the inter-language values ​​of the oral-DDK. Methods & Procedures: the recordings of 189 healthy adults were considered (between 26 to 86 years), 94 females and 95 males, while producing the sequential motion rate (SMR) /pa-ta-ka/. Count-by-time method was used using an acoustic analysis software. Oral-DDK averages were established for each language and the effect of gender and age was determined. The variability of the measure among languages ​​was also established. Results: reference data were obtained for each language. Age did not show any significant effect on oral-DDK despite its decline as age advanced; there was no gender effect on the oral-DDK, and the comparison of the oral-DDK among languages showed statistically significant differences. Conclusions: This study determined the reference data of the oral-DDK for Spanish, German and Czech and found significant differences among the three languages ​​and no significant effects of age or gender on the measure studied.

Abstract Bioacoustic research spans a wide range of biological questions and applications, relying on identification of target species or smaller acoustic units, such as distinct call types. However, manually identifying the signal of interest is time-intensive, error-prone, and becomes unfeasible with large data volumes. Therefore, machine-driven algorithms are increasingly applied to various bioacoustic signal identification challenges. Nevertheless, biologists still have major difficulties trying to transfer existing animal- and/or scenario-related machine learning approaches to their specific animal datasets and scientific questions. This study presents an animal-independent, open-source deep learning framework, along with a detailed user guide. Three signal identification tasks, commonly encountered in bioacoustics research, were investigated: (1) target signal vs. background noise detection, (2) species classification, and (3) call type categorization. ANIMAL-SPOT successfully segmented human-annotated target signals in data volumes representing 10 distinct animal species and 1 additional genus, resulting in a mean test accuracy of 97.9%, together with an average area under the ROC curve (AUC) of 95.9%, when predicting on unseen recordings. Moreover, an average segmentation accuracy and F1-score of 95.4% was achieved on the publicly available BirdVox-Full-Night data corpus. In addition, multi-class species and call type classification resulted in 96.6% and 92.7% accuracy on unseen test data, as well as 95.2% and 88.4% regarding previous animal-specific machine-based detection excerpts. Furthermore, an Unweighted Average Recall (UAR) of 89.3% outperformed the multi-species classification baseline system of the ComParE 2021 Primate Sub-Challenge. Besides animal independence, ANIMAL-SPOT does not rely on expert knowledge or special computing resources, thereby making deep-learning-based bioacoustic signal identification accessible to a broad audience.

Abstract Biometric identification techniques such as photo-identification require an array of unique natural markings to identify individuals. From 1975 to present, Bigg’s killer whales have been photo-identified along the west coast of North America, resulting in one of the largest and longest-running cetacean photo-identification datasets. However, data maintenance and analysis are extremely time and resource consuming. This study transfers the procedure of killer whale image identification into a fully automated, multi-stage, deep learning framework, entitled FIN-PRINT. It is composed of multiple sequentially ordered sub-components. FIN-PRINT is trained and evaluated on a dataset collected over an 8-year period (2011–2018) in the coastal waters off western North America, including 121,000 human-annotated identification images of Bigg’s killer whales. At first, object detection is performed to identify unique killer whale markings, resulting in 94.4% recall, 94.1% precision, and 93.4% mean-average-precision (mAP). Second, all previously identified natural killer whale markings are extracted. The third step introduces a data enhancement mechanism by filtering between valid and invalid markings from previous processing levels, achieving 92.8% recall, 97.5%, precision, and 95.2% accuracy. The fourth and final step involves multi-class individual recognition. When evaluated on the network test set, it achieved an accuracy of 92.5% with 97.2% top-3 unweighted accuracy (TUA) for the 100 most commonly photo-identified killer whales. Additionally, the method achieved an accuracy of 84.5% and a TUA of 92.9% when applied to the entire 2018 image collection of the 100 most common killer whales. The source code of FIN-PRINT can be adapted to other species and will be publicly available.

Parkinson’s disease (PD) is a neurodegenerative disorder that affects the coordination of muscles and limbs, including those responsible of the speech production. The lack of control of the limbs and muscles involved in the speech production process can generate intelligibility problems and this situation has a negative impact in the social interaction of the patients. It is already demonstrated that constant speech therapy can improve the communication abilities of the patients; however, the measurement of the recovery progress is done subjectively by speech therapists and neurologists. Due to this, it is required the development of flexible tools able to asses and guide the speech therapy of the patients. In this paper the design and deployment of a new device for the real time assessment of speech signals of people with PD is presented. The processes of design and deployment include the development on three platforms: first, a graphic user interface is developed on Matlab, second the first prototype is implemented on a digital signal processor (DSP) and third, the final device is developed on a mini-computer. The device is equipped with an audio codec, storage capacity and the processing unit. Besides, the system is complemented with a monitor to display the processed information on real time and with a keyboard enabling the interaction of the end-user with the device. Different acoustics and nonlinear dynamics measures which have been used in the state of the art for the assessment of speech of people with PD are implemented on the three mentioned platforms. In accordance with the state of the art, the designed platforms show an increment in the variation of the fundamental period of speech (commonly called pitch) of people with PD. Additionally, the decrease of the vocal space area is validated for the case of patients with PD. These results indicate that the designed device is useful to perform the assessment and monitoring of the speech therapy of people with PD.

Este artículo presenta un sistema para la detección automática de señales de voz hipernasales basado en la combinación de dos diferentes esquemas de caracterización aplicados en las cinco vocales del español y dos palabras seleccionadas. El primer esquema está basado en características clásicas como perturbaciones del periodo fundamental, medidas de ruido y coeficientes cepstrales en la frecuencia de Mel. El segundo enfoque está basado en medidas de dinámica no lineal. Las características más relevantes son seleccionadas usando dos técnicas: análisis de componentes principales y selección flotante hacia adelante secuencial. La decisión acerca de si un registro de voz es hipernasal o sano es tomada usando una máquina de soporte vectorial de margen suave. Los experimentos consideran grabaciones de las cinco vocales del idioma español y las palabras y se consideran, asimismo, tres conjuntos de características: (1) el enfoque clásico, (2) el análisis de dinámica no lineal y (3) la combinación de ambos esquemas. En general, los aciertos son mayores y más estables cuando las características clásicas y no lineales son combinadas, indicando que el análisis de dinámica no lineal se complementa con el esquema clásico.

La enfermedad de Parkinson (EP) es un desorden neurodegenerativo queafecta la coordinación de músculos y extremidades, incluyendo aquellosresponsables de la producción del habla, generando alteraciones en lainteligibilidad de la señal de voz. Está demostrado que el ejercicio terapéuticoconstante puede mejorar las habilidades de comunicación de los pacientes; sinembargo, el diagnóstico acerca del avance en el proceso de recuperación esrealizado de forma subjetiva por los fonoaudiólogos o neurólogos. Debido aesto se requiere el desarrollo de herramientas flexibles que valoren y guíen laterapia fonoaudiológica de los pacientes. En este artículo se presenta el diseñoe implementación de un sistema embebido para el análisis en tiempo real dela voz de pacientes con EP. Para esto se desarrollan tres plataformas; primero,se construye una interfaz gráfica en Matlab; luego, se crea un primer prototipobasado en un DSP TMS320C6713 de Texas Instruments. La aplicación finales desarrollada sobre un mini-ordenador que cuenta con un códec de audio,capacidad de almacenamiento, y una unidad de procesamiento. El sistemaademás se complementa con un monitor LCD para desplegar información entiempo real, y un teclado para la interacción con el usuario.En todas las plataformas se evalúan diferentes medidas usadas comúnmenteen la valoración de la voz de pacientes con EP, incluyendo característicasacústicas y de dinámica no lineal. En concordancia con otros trabajos delestado del arte donde se analiza la voz de personas con EP, la plataformadiseñada muestra un incremento en la variación del pitch en la voz de lospacientes, además de un decremento en el valor del área del espacio vocálico.Este resultado indica que la herramienta diseñada puede ser útil para hacer laevaluación y seguimiento de la terapia fonoaudiológica de pacientes con EP.

NeuroSpeech is a software for modeling pathological speech signals considering different speech dimensions: phonation, articulation, prosody, and intelligibility. Although it was developed to model dysarthric speech signals from Parkinson’s patients, its structure allows other computer scientists or developers to include other pathologies and/or measures. Different tasks can be performed: (1) modeling of the signals considering the aforementioned speech dimensions, (2) automatic discrimination of Parkinson’s vs. non-Parkinson’s, and (3) prediction of the neurological state according to the Unified Parkinson’s Disease Rating Scale (UPDRS) score. The prediction of the dysarthria level according to the Frenchay Dysarthria Assessment scale is also provided. Keywords: Speech processing, Dysarthria, Parkinson’s disease, Phonation, Articulation, Prosody, Intelligibility, Python, Software

This paper presents a system for the automatic detection of hypernasal speech signals based on the combination of two different characterization approaches applied to the five spanish vowels and two selected words. The first approach is based on classical features such as pitch period perturbations, noise measures, and Mel-Frequency Cepstral Coefficients (MFCC). The second approach is based on the Non-Linear Dynamics (NLD) analysis. The most relevant features are selected and sorted using two techniques: Principal Components Analysis (PCA) and Sequential Forward Floating Selection (SFFS). The decision about whether a voice record is hypernasal or healthy is taken using a Soft Margin - Support Vector Machine (SM-SVM). Experiments upon recordings of the five Spanish vowels and the words are performed considering three different set of features: (1) the classical approach, (2) the NLD analysis, and (3) the combination of the classical and NLD measures. In general, the accuracies are higher and more stable when the classical and NLD features are combined, indicating that the NLD analysis is complementary to the classical approach.

In patients suffering from head and neck cancer, speech intelligibility is often restricted. For assessment and outcome measurements, automatic speech recognition systems have previously been shown to be appropriate for objective and quick evaluation of intelligibility. In this study we investigate the applicability of the method to speech disorders caused by head and neck cancer. Intelligibility was quantified by speech recognition on recordings of a standard text read by 41 German laryngectomized patients with cancer of the larynx or hypopharynx and 49 German patients who had suffered from oral cancer. The speech recognition provides the percentage of correctly recognized words of a sequence, that is, the word recognition rate. Automatic evaluation was compared to perceptual ratings by a panel of experts and to an age-matched control group. Both patient groups showed significantly lower word recognition rates than the control group. Automatic speech recognition yielded word recognition rates which complied with experts' evaluation of intelligibility on a significant level. Automatic speech recognition serves as a good means with low effort to objectify and quantify the most important aspect of pathologic speech—the intelligibility. The system was successfully applied to voice and speech disorders.