Realistic Multi-Microphone Data Simulation for Distant Speech Recognition

By Mirco Ravanelli, Piergiorgio Svaizer, Maurizio Omologo. 2016

37 citations recorded by the source corpus.

The availability of realistic simulated corpora is of key importance for the future progress of distant speech recognition technology. The reliability, flexibility and low computational cost of a data simulation process may ultimately allow researchers to train, tune and test different techniques in a variety of acoustic scenarios, avoiding the laborious effort of directly recording real data from the targeted environment. In the last decade, several simulated corpora have been released to the research community, including the data-sets distributed in the context of projects and international challenges, such as CHiME and REVERB. These efforts were extremely useful to derive baselines and common evaluation frameworks for comparison purposes. At the same time, in many cases they highlighted the need of a better coherence between real and simulated conditions. In this paper, we examine this issue and we describe our approach to the generation of realistic corpora in a domestic context. Experimental validation, conducted in a multi-microphone scenario, shows that a comparable performance trend can be observed with both real and simulated data across different recognition frameworks, acoustic models, as well as multi-microphone processing techniques.

Subfields: Speech & Audio

How researchers see this paper

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

In Speech & Audio

Citations in the corpus

Multi-Task Self-Supervised Learning for Robust Speech Recognition (2020, positive): “Along with the original clean version, we generated a contaminated version of TIMIT using noise sequences and real impulse responses [33] (different from those used for self-supervised training).”
Light Gated Recurrent Units for Speech Recognition (2018, methodology): “n the image method [55] and correspond to four different reverberation times T 60 (ranging from 250 to 1000 ms). More details on the realistic impulse responses adopted in this corpus can be found in [56]–[58]. Additional experiments were conducted with the CHiME 4 dataset [39], that is based on bot”
Batch-normalized joint training for DNN-based distant speech recognition (2016, methodology): “ation time (T 60) of the considered room is about 0.7 seconds. Development and test data have been simulated with the same approach. More details about the data contamination approach can be found in [45, 46, 47]. The WSJ experiments are based on the popular wsj5k task (aligned with the CHiME 3 [6] ”
VoiceHome-2, an extended corpus for multichannel speech processing in real homes (2019, unknown): “Indeed, the room impulse responses and the noise-only data in voiceHome can be used to generate simulated data for training (Ravanelli et al., 2016) while the new, real voiceHome-2 data can be used for testing.”
Twin Regularization for online speech recognition (2018, methodology): “Training was based on the original WSJ-5k corpus (consisting of 7138 sentences uttered by 83 speakers) that was contaminated with a set of impulse responses measured in a real apartment [33].”
A network of deep neural networks for Distant Speech Recognition (2017, background): “More details on this corpus and on the impulse responses adopted in this work can be found in [33, 34].”
Interpretable Convolutional Filters with SincNet (2018, methodology): “With the purpose of validating SincNet in both clean and noisy conditions, speech recognition experiments are conducted on both the TIMIT and DIRHA dataset [41, 42].”
Improving Speech Recognition by Revising Gated Recurrent Units (2017, background): “More details on this dataset and on the impulse responses can be found in [36, 37].”
Robust Laughter Detection in Noisy Environments (2021, background): “In addition to increasing the size of the training data and functioning as a regularizer, these kinds of augmentations have shown to be important for speech recognition in noisy environments [30].”
The Pytorch-kaldi Speech Recognition Toolkit (2018, methodology): “To validate our model in a more challenging scenario, experiments were also conducted in distant-talking conditions with the DIRHA-English dataset(4) [36,37].”