Details
Presenter(s)
![Yi Sheng Chong Headshot](https://confcats-catavault.s3.amazonaws.com/CATAVault/ieeecass/master/files/styles/cc_user_photo/s3/user-pictures/10301_0.jpg?h=7bc4e76b&itok=N6WYGOBj)
Display Name
Yi Sheng Chong
- Affiliation
-
AffiliationNanyang Technological University
- Country
-
CountrySingapore
Abstract
Mel frequency cepstral coefficient (MFCC) features are widely used in applications such as keyword spotting, bearing fault detection and heart sound classification. This work proposes a low power MFCC engine that enables the use of MFCC on the battery-powered edge applications. Various optimizations are adopted to allow energy efficient implementation for all MFCC processing steps. Our proposed engine maintains good accuracy when deployed to several applications as compared to pure software implementation while consuming only 128 nW at 0.3 V supply. It occupies only 0.08mm^2 in 40nm CMOS, which is 5x and 2.75x power and area reduction respectively when compared to prior arts.