# ImportantAug: a data augmentation agent for speech

Source: [ImportantAug: a data augmentation agent for speech](<https://www.sota2.com/research/paper/68fa6bd41e209131cea3061d>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-01-01T14:09:46.915Z

Authors: Viet Anh Trinh, Hassan Salami Kavaki, Michael I Mandel (1 and 2) \_\_INSTITUTION\_3\_\_ CUNY Graduate Center, (2) Brooklyn College)

Published: 2021-12-14T04:37:04.000Z

## Abstract

We introduce ImportantAug, a technique to augment training data for speech classification and recognition models by adding noise to unimportant regions of the speech and not to important regions. Importance is predicted for each utterance by a data augmentation agent that is trained to maximize the amount of noise it adds while minimizing its impact on recognition performance. The effectiveness of our method is illustrated on version two of the Google Speech Commands (GSC) dataset. On the standard GSC test set, it achieves a 23.3% relative error rate reduction compared to conventional noise augmentation which applies noise to speech without regard to where it might be most effective. It also provides a 25.4% error rate reduction compared to a baseline without data augmentation. Additionally, the proposed ImportantAug outperforms the conventional noise augmentation and the baseline on two test sets with additional noise added.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Speech Command Recognition on GSC-QUT (test)](<https://www.sota2.com/research/sota/speech-command-recognition-on-gsc-qut-test.md>): Speech Command Recognition; GSC-QUT (test); Error Rate: 4.8.

- [Speech Command Recognition on GSC-MUSAN (test)](<https://www.sota2.com/research/sota/speech-command-recognition-on-gsc-musan-test.md>): Speech Command Recognition; GSC-MUSAN (test); Error Rate: 5.1.

- [Speech Command Recognition on Google Speech Command (GSC) v2 (test)](<https://www.sota2.com/research/sota/speech-command-recognition-on-google-speech-command-gsc-v2-test.md>): Speech Command Recognition; Google Speech Command (GSC) v2 (test); Error Rate: 0.05.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2112.07156>)

[PDF](<https://arxiv.org/pdf/2112.07156.pdf>)

[Implementation](<https://github.com/tvanh512/importantAug>)