Batch-normalized joint training for DNN-based distant speech recognition

By Mirco Ravanelli, Philemon Brakel, Maurizio Omologo, Yoshua Bengio. 2016

35 citations recorded by the source corpus.

Improving distant speech recognition is a crucial step towards flexible human-machine interfaces. Current technology, however, still exhibits a lack of robustness, especially when adverse acoustic conditions are met. Despite the significant progress made in the last years on both speech enhancement and speech recognition, one potential limitation of state-of-the-art technology lies in composing modules that are not well matched because they are not trained jointly.

Subfields: Artificial Intelligence, Optimization & Training, Speech & Audio, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Yoshua Bengio
2. Mirco Ravanelli
3. Titouan Parcollet
4. Philemon Brakel
5. Maurizio Omologo
6. Jianyuan Zhong
7. Loren Lugosch
8. Kyunghyun Cho

In Optimization & Training

1. Mirco Ravanelli
2. Yoshua Bengio
3. Maurizio Omologo
4. Philemon Brakel
5. Titouan Parcollet
6. R. Mori
7. Chien-Feng Liao
8. Samuele Cornell

In Speech & Audio

1. Yoshua Bengio
2. Mirco Ravanelli
3. Titouan Parcollet
4. Philemon Brakel
5. Maurizio Omologo
6. Jianyuan Zhong
7. R. Mori
8. Chien-Feng Liao

In Signal Processing

1. Yoshua Bengio
2. Mirco Ravanelli
3. Titouan Parcollet
4. Philemon Brakel
5. Maurizio Omologo
6. Jianyuan Zhong
7. Loren Lugosch
8. Samuele Cornell

Citations in the corpus

• Light Gated Recurrent Units for Speech Recognition (2018, methodology): “nique is effective both to improve the system performance and to speed-up the training procedure [20], [22], [37], [49], [50].”
• Learning Speaker Representations with Mutual Information (2018, methodology): “Next, two fully-connected leaky-ReLU layers [39] composed of 2048 and 1024 neurons (normalized with batch normalization [40, 41]) were applied.”
• SpeechBrain: A General-Purpose Speech Toolkit (2021, background): “A single toolkit, for instance, encourages the exploration of transfer learning and joint training techniques across different tasks [20–23] and enables the creation of fully differentiable graphs where multiple technologies are trained jointly and learn to interact.”
• The Pytorch-kaldi Speech Recognition Toolkit (2018, unknown): “It is also possible to employ models based on an ecosystem of neural networks performing different tasks, as done in the context of joint training between speech enhancement and speech recognition [28, 29] or in the context of the recently-proposed cooperative networks of deep neural networks [30].”
• A network of deep neural networks for Distant Speech Recognition (2017, background): “The third line reports the performance achieved with the joint training approach recently proposed in [24].”