Projekte pro Jahr
Abstract
This paper introduces neural architecture search (NAS) for the automatic discovery of end-to-end keyword spotting (KWS) models for limited resource environments. We employ a differentiable NAS approach to optimize the structure of convolutional neural networks (CNNs) operating on raw audio waveforms. After a suitable KWS model is found with NAS, we conduct quantization of weights and activations to reduce the memory footprint. We conduct extensive experiments on the Google speech commands dataset. In particular, we compare our end-to-end models to mel-frequency cepstral coefficient (MFCC) based CNNs. For quantization, we compare fixed bit-width quantization and trained bit-width quantization. Using NAS only, we were able to obtain a highly efficient model with an accuracy of 95.55% using 75.7k parameters and 13.6M operations. Using trained bit-width quantization, the same model achieves a test accuracy of 93.76% while using on average only 2.91 bits per activation and 2.51 bits per weight
Originalsprache | englisch |
---|---|
Titel | 2022 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2022 - Proceedings |
Herausgeber (Verlag) | Institute of Electrical and Electronics Engineers |
Seiten | 3423-3427 |
Seitenumfang | 5 |
ISBN (elektronisch) | 9781665405409 |
DOIs | |
Publikationsstatus | Veröffentlicht - 2022 |
Veranstaltung | 47th IEEE International Conference on Acoustics, Speech and Signal Processing: ICASSP 2022 - Virtual, Online, Singapur Dauer: 22 Mai 2022 → 27 Mai 2022 |
Publikationsreihe
Name | ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings |
---|---|
Band | 2022-May |
ISSN (Print) | 1520-6149 |
Konferenz
Konferenz | 47th IEEE International Conference on Acoustics, Speech and Signal Processing |
---|---|
Kurztitel | ICASSP 2022 |
Land/Gebiet | Singapur |
Ort | Virtual, Online |
Zeitraum | 22/05/22 → 27/05/22 |
ASJC Scopus subject areas
- Software
- Signalverarbeitung
- Elektrotechnik und Elektronik
Fields of Expertise
- Information, Communication & Computing
Fingerprint
Untersuchen Sie die Forschungsthemen von „End-to-end Keyword Spotting using Neural Architecture Search and Quantization“. Zusammen bilden sie einen einzigartigen Fingerprint.Projekte
- 1 Laufend