AAD-KWS: a sub-$\mu\mathrm{W}$ keyword spotting chip with a zero-cost, acoustic activity detector from a 170nW MFCC feature extractor in 28nm CMOS.
Lixuan ZhuWeiwei ShanJiaming XuYicheng LuPublished in: ESSDERC (2021)
Keyphrases
- feature extractor
- keyword spotting
- speech recognition
- cmos technology
- feature extraction
- nm technology
- silicon on insulator
- metal oxide semiconductor
- power consumption
- speech processing
- low power
- speaker identification
- pattern recognition
- speech signal
- hidden markov models
- recognition scheme
- image processing
- image sensor
- texture descriptors
- signal processing
- sound source
- feature vectors
- gaussian mixture model
- image classification