A 141 UW, 2.46 PJ/Neuron Binarized Convolutional Neural Network Based Self-Learning Speech Recognition Processor in 28NM CMOS.
Shouyi YinPeng OuyangShixuan ZhengDandan SongXiudong LiLeibo LiuShaojun WeiPublished in: VLSI Circuits (2018)
Keyphrases
- speech recognition
- neural network
- high speed
- silicon on insulator
- ibm power processor
- single chip
- cmos technology
- low power
- pattern recognition
- hidden markov models
- language model
- metal oxide semiconductor
- random access memory
- speech recognizer
- speech processing
- automatic speech recognition
- speech synthesis
- speech signal
- speech recognition technology
- speech understanding
- speech recognition systems
- speaker identification
- power consumption
- noisy environments
- image sensor
- keyword spotting
- speech recognizers
- speaker dependent
- speaker diarization
- low voltage
- document images
- speaker independent
- deep learning
- speech recognition errors
- power dissipation
- mixture model