End-to-End Domain-Adversarial Voice Activity Detection

Lavechin, Marvin; Gill, Marie-Philippe; Bousbib, Ruben; Bredin, Hervé; García-Perera, Leibny Paola

doi:10.21437/interspeech.2020-2285

Cited by 30 publications

(16 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Generally, noise-robust VAD systems are developed using audio from clean speech datasets augmented with different types of noise [21,11,8]. Hebbar et al [11] compiled the Subtitle-Aligned Movie (SAM) corpus, a dataset based on 117 hours of movie audio.…”

Section: Training Datamentioning

confidence: 99%

“…Then a smoothing filter is applied to decide the label for a frame spanned by multiple segments. We examined two common smoothing filters: majority vote (median) [11] and average (mean) [8]. We also experimented with different amounts of overlap ranging from 12.5% to 87.5% to understand the effect of this parameter for frame-level performance.…”

Section: Modelmentioning

confidence: 99%

“…Median and mean smoothing filters are generally used when generating frame level predictions [30,8]. In Table 4, we show the results of these two filters using three different amounts of overlap: 12.5%, 50.0%, and 87.5%.…”

Section: Overlapping Predictionsmentioning

confidence: 99%

“…Apart from feature-based [1,2,3] and statistical modeling approaches [4,5], recent research effort has been devoted to finding efficient deep-learning-based VAD model architectures. Notable examples include Recurrent Neural Networks (RNN) [6,7,8], Convolutional Neural Networks (CNN) [9,10,11,12], and Convolutional Long Short-Term Memory (LSTM) Deep Neural Networks (CLDNN) [13], which conduct frequency modeling with CNN and temporal modeling with LSTM. LSTM is a popular choice for sequential modeling of VAD tasks [13,6].…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

MarbleNet: Deep 1D Time-Channel Separable Convolutional Neural Network for Voice Activity Detection

Jia

Majumdar

Ginsburg

2021

ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

We present MarbleNet, an end-to-end neural network for Voice Activity Detection (VAD). MarbleNet is a deep residual network composed from blocks of 1D time-channel separable convolution, batchnormalization, ReLU and dropout layers. When compared to a stateof-the-art VAD model, MarbleNet is able to achieve similar performance with roughly 1/10-th the parameter cost. We further conduct extensive ablation studies on different training methods and choices of parameters in order to study the robustness of MarbleNet in realworld VAD tasks.

show abstract

Section: Training Datamentioning

confidence: 99%

Section: Modelmentioning

confidence: 99%

Section: Overlapping Predictionsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

MarbleNet: Deep 1D Time-Channel Separable Convolutional Neural Network for Voice Activity Detection

Jia

Majumdar

Ginsburg

2021

ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

show abstract

“…In order to alleviate the effect of loudness difference shown in Table 1, we scale the audio signal for each speaker turn separately, and make sure each turn is in the range of -1 and 1. We apply a domain-adversarial neural network based voice activity detection model [18] for intraspeaker-turn segmentation. We also exclude speaker turns that are shorter than 2 seconds to extract robust acoustic features.…”

Section: Data Preprocessingmentioning

confidence: 99%

Modelling Paralinguistic Properties in Conversational Speech to Detect Bipolar Disorder and Borderline Personality Disorder

Wang

Vaci

et al. 2021

ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

Bipolar disorder (BD) and borderline personality disorder (BPD) are two chronic mental health conditions that clinicians find challenging to distinguish based on clinical interviews, due to their overlapping symptoms. In this work, we investigate the automatic detection of these two conditions by modelling both verbal and non-verbal cues in a set of interviews. We propose a new approach of modelling short-term features with visibility-signature transform, and compare it with widely used high-level statistical functions. We demonstrate the superior performance of our proposed signature-based model. Furthermore, we show the role of different sets of features in characterising BD and BPD.

show abstract

Application of Adversarial Domain Adaptation to Voice Activity Detection

Kim

2021

Lecture Notes in Networks and Systems

View full text Add to dashboard Cite

End-to-End Domain-Adversarial Voice Activity Detection

Cited by 30 publications

References 0 publications

MarbleNet: Deep 1D Time-Channel Separable Convolutional Neural Network for Voice Activity Detection

MarbleNet: Deep 1D Time-Channel Separable Convolutional Neural Network for Voice Activity Detection

Modelling Paralinguistic Properties in Conversational Speech to Detect Bipolar Disorder and Borderline Personality Disorder

Application of Adversarial Domain Adaptation to Voice Activity Detection

Contact Info

Product

Resources

About