Details
Presenter(s)
![Lindong Wu Headshot](https://confcats-catavault.s3.amazonaws.com/CATAVault/ieeecass/master/files/styles/cc_user_photo/s3/user-pictures/19171.jpg?h=c27cc264&itok=dbsbutBW)
Display Name
Lindong Wu
- Affiliation
-
AffiliationPeking University
- Country
Abstract
A high accuracy speech command recognition ASIC is implemented in TSMC 180nm CMOS technology, which is based on a One-Dimension Convolutional Neural Network (1-D CNN) and the MFCC algorithm. The commands are reconfigurable. Max 30 speech commands are supported. Measurements reveal that the chip achieves 95.6% accuracy on Google Speech Command Database (GSCD), while keeping a reasonable power consumption as 26.4mW, which is better than the state-of-the-art chips.