A hybrid coder for hidden Markov models using a recurrent neural networks

Bengio, Yoshua; Cardin, R.; Mori, Renato De; Normandin, Yves

doi:10.1109/icassp.1990.115768

Cited by 12 publications

(6 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…For the DNN architecture, we used 4 hidden layers, each with 2048 units and a Rectified Linear Unit (ReLU) [92] activation function. For classification, we used a softmax layer [93], which assigns a probability to each class, i.e., a probability to each HMM state. The system's architecture is depicted in Figure 6.…”

Section: Deep Neural Network For Acoustic Modellingmentioning

confidence: 99%

An HMM-DNN-Based System for the Detection and Classification of Low-Frequency Acoustic Signals from Baleen Whales, Earthquakes, and Air Guns off Chile

Buchan

Duran

Rojas³

et al. 2023

Remote Sensing

View full text Add to dashboard Cite

Marine passive acoustic monitoring can be used to study biological, geophysical, and anthropogenic phenomena in the ocean. The wide range of characteristics from geophysical, biological, and anthropogenic sounds sources makes the simultaneous automatic detection and classification of these sounds a significant challenge. Here, we propose a single Hidden Markov Model-based system with a Deep Neural Network (HMM-DNN) for the detection and classification of low-frequency biological (baleen whales), geophysical (earthquakes), and anthropogenic (air guns) sounds. Acoustic data were obtained from the Preparatory Commission for the Comprehensive Nuclear-Test-Ban Treaty Organization station off Juan Fernandez, Chile (station HA03) and annotated by an analyst (498 h of audio data containing 30,873 events from 19 different classes), and then divided into training (60%), testing (20%), and tuning (20%) subsets. Each audio frame was represented as an observation vector obtained through a filterbank-based spectral feature extraction procedure. The HMM-DNN training procedure was carried out discriminatively by setting HMM states as targets. A model with Gaussian Mixtures Models and HMM (HMM-GMM) was trained to obtain an initial set of HMM target states. Feature transformation based on Linear Discriminant Analysis and Maximum Likelihood Linear Transform was also incorporated. The HMM-DNN system displayed good capacity for correctly detecting and classifying events, with high event-level accuracy (84.46%), high weighted average sensitivity (84.46%), and high weighted average precision (89.54%). Event-level accuracy increased with higher event signal-to-noise ratios. Event-level metrics per class also showed that our HMM-DNN system generalized well for most classes but performances were best for classes that either had a high number of training exemplars (e.g., generally above 50) and/or were for classes of signals that had low variability in spectral features, duration, and energy levels. Fin whale and Antarctic blue whale song and air guns performed particularly well.

show abstract

Section: Deep Neural Network For Acoustic Modellingmentioning

confidence: 99%

An HMM-DNN-Based System for the Detection and Classification of Low-Frequency Acoustic Signals from Baleen Whales, Earthquakes, and Air Guns off Chile

Buchan

Duran

Rojas³

et al. 2023

Remote Sensing

View full text Add to dashboard Cite

show abstract

“…A variety of solutions have been proposed to tackle this problem. One solution uses the ANN to compute an additional set of symbols as transformed observations for the HMM [90]. A further improvement of this method is achieved through a global optimization of both the ANN and HMM [91].…”

Section: Proposed Solutionsmentioning

confidence: 99%

Speech Processing, Recognition and Artificial Neural Networks

Chollet¹,

Benedetto²,

Esposito³

et al. 1999

View full text Add to dashboard Cite

“…Otra posible solución es utilizar redes neuronales para computar conjuntos de símbolos adicionales, que puedan entregarse como observaciones transformadas para los HMM [BEN90]. Más específicamente, la red genera grados de certeza para los rasgos básicos de los sonidos como sonoridad, fricación y oclusión/silencio.…”

Section: Rasgounclassified

Aportación a la extracción paramétrica en reconocimiento de voz robusto basada en la aplicación de conocimiento de fonética acústica

Álvarez-Marquina¹

View full text Add to dashboard Cite

Deseo agradecer a Pedro Gómez Vilda, director de la presente tesis, los consejos, ayuda y orientación que a lo largo de estos últimos cinco años me ha venido proporcionando de manera regular.Así mismo, quisiera reconocer la importancia que los proyectos e investigaciones desarrollados con los profesores Víctor Nieto Lluís y Rafael Martínez Olalla han tenido en este trabajo y en mi carrera investigadora.De igual manera, no quisiera olvidar a los componentes del laboratorio de comunicación oral "Robert Wayne Newcomb" y en particular a su maestro de laboratorio José María García.Por último, quisiera dar las gracias a todas aquellas personas que han compartido conmigo todo lo bueno y lo malo que supone realizar unos estudios tan largos y muy especialmente a mi familia, a quien dedico este libro.With usura the line grows thick Ezra Pound, "XLV", The Cantos

show abstract

A hybrid coder for hidden Markov models using a recurrent neural networks

Cited by 12 publications

References 8 publications

An HMM-DNN-Based System for the Detection and Classification of Low-Frequency Acoustic Signals from Baleen Whales, Earthquakes, and Air Guns off Chile

An HMM-DNN-Based System for the Detection and Classification of Low-Frequency Acoustic Signals from Baleen Whales, Earthquakes, and Air Guns off Chile

Speech Processing, Recognition and Artificial Neural Networks

Aportación a la extracción paramétrica en reconocimiento de voz robusto basada en la aplicación de conocimiento de fonética acústica

Contact Info

Product

Resources

About