Composition of Deep and Spiking Neural Networks for Very Low Bit Rate Speech Coding

Cerňak, Miloš; Lazaridis, Alexandros; Asaei, Afsaneh; Garner, Philip N.

doi:10.1109/taslp.2016.2604566

Cited by 25 publications

(13 citation statements)

References 57 publications

(59 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As far as we know, this is only the second published work to learn an audio compression pipeline end-to-end -the previous being an obscure early attempt by Morishima et al in 1990 [7] -and the first to compete with a contemporary standard. Cernak et al [8] proposed a nearly end-to-end design for a very-low-bitrate low-quality speech coder in 2016; however, their pipeline still required extraction of acoustic features and pitch (and was also quite complex, composing several different deep and spiking neural networks together). All other related designs we know of employ ANNs as a mere component of a larger hand-designed system.…”

Section: Introductionmentioning

confidence: 99%

End-To-End Optimized Speech Coding with Deep Neural Networks

Kankanahalli

2018

2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

Modern compression algorithms are often the result of laborious domain-specific research; industry standards such as MP3, JPEG, and AMR-WB took years to develop and were largely hand-designed. We present a deep neural network model which optimizes all the steps of a wideband speech coding pipeline (compression, quantization, entropy coding, and decompression) end-to-end directly from raw speech data -no manual feature engineering necessary, and it trains in hours. In testing, our DNN-based coder performs on par with the AMR-WB standard at a variety of bitrates (∼9kbps up to ∼24kbps). It also runs in realtime on a 3.8GhZ Intel CPU.

show abstract

Section: Introductionmentioning

confidence: 99%

End-To-End Optimized Speech Coding with Deep Neural Networks

Kankanahalli

2018

2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

show abstract

“…An example use case was shown on DNN-HMM speech recognition. Beyond ASR, other applications that rely on estimation of DNN posteriors can also benefit from the proposed approach, such as spoken query detection [26], parametric speech coding [27] and linguistic parsing [28]. Thorough experiments on large speech corpora for a broad range of applications is planned for future research.…”

Section: Discussionmentioning

confidence: 99%

Low-Rank Representation of Nearest Neighbor Posterior Probabilities to Enhance DNN Based Acoustic Modeling

Luyet¹,

Dighe²,

Asaei³

et al. 2016

Interspeech 2016

Self Cite

View full text Add to dashboard Cite

We hypothesize that optimal deep neural networks (DNN) class-conditional posterior probabilities live in a union of lowdimensional subspaces. In real test conditions, DNN posteriors encode uncertainties which can be regarded as a superposition of unstructured sparse noise over the optimal posteriors. We aim to investigate different ways to structure the DNN outputs by exploiting low-rank representation (LRR) techniques. Using a large number of training posterior vectors, the underlying low-dimensional subspace of a test posterior is identified through nearest neighbor analysis, and low-rank decomposition enables separation of the "optimal" posteriors from the spurious uncertainties at the DNN output. Experiments demonstrate that by processing subsets of posteriors which possess strong subspace similarity, low-rank representation enables enhancement of posterior probabilities, and leads to higher speech recognition accuracy based on the hybrid DNN-hidden Markov model (HMM) system.

show abstract

“…In addition to the phonetic posteriors, our previous studies on phonological posteriors show that they conform to a small number of unique binary structures which are a tiny fraction of the number of permissible codes [12]. Exploiting this property enables construction of a small-size codebook for very low-bit rate speech coding [12,13]. More recently, we also found structured sparsity of phonological posteriors highly effective for classification of supra-segmental linguistic events [14].…”

Section: Introductionmentioning

confidence: 94%

Phonetic and Phonological Posterior Search Space Hashing Exploiting Class-Specific Sparsity Structures

Asaei¹,

Luyet²,

Cerňak³

et al. 2016

Interspeech 2016

Self Cite

View full text Add to dashboard Cite

This paper shows that exemplar-based speech processing using class-conditional posterior probabilities admits a highly effective search strategy relying on posteriors' intrinsic sparsity structures. The posterior probabilities are estimated for phonetic and phonological classes using deep neural network (DNN) computational framework. Exploiting the class-specific sparsity leads to a simple quantized posterior hashing procedure to reduce the search space of posterior exemplars. To that end, small number of quantized posteriors are regarded as representatives of the posterior space and used as hash keys to index subsets of neighboring exemplars. The k nearest neighbor (kNN) method is applied for posterior based classification problems. The phonetic posterior probabilities are used as exemplars for phonetic classification whereas the phonological posteriors are used as exemplars for automatic prosodic event detection. Experimental results demonstrate that posterior hashing improves the efficiency of kNN classification drastically. This work encourages the use of posteriors as discriminative exemplars appropriate for large scale speech classification tasks.

show abstract

Composition of Deep and Spiking Neural Networks for Very Low Bit Rate Speech Coding

Cited by 25 publications

References 57 publications

End-To-End Optimized Speech Coding with Deep Neural Networks

End-To-End Optimized Speech Coding with Deep Neural Networks

Low-Rank Representation of Nearest Neighbor Posterior Probabilities to Enhance DNN Based Acoustic Modeling

Phonetic and Phonological Posterior Search Space Hashing Exploiting Class-Specific Sparsity Structures

Contact Info

Product

Resources

About