SoC for IoT Devices with 18ms Noise-Robust Speech-to-Text Latency via Bayesian Speech Denoising and Attention-Based Sequence-to-Sequence DNN Speech Recognition in 16nm FinFET.
Thierry TambeEn-Yu YangGlenn G. KoYuji ChaiColeman HooperMarco DonatoPaul N. WhatmoughAlexander M. RushDavid BrooksGu-Yeon WeiPublished in: ISSCC (2021)
Keyphrases
- speech recognition
- noisy environments
- denoising
- speech synthesis
- speech signal
- hidden markov models
- pattern recognition
- automatic speech recognition
- speech processing
- language model
- noisy speech
- speaker identification
- spectral subtraction
- speech recognition systems
- speech recognizer
- keyword spotting
- speaker verification
- recognition engine
- speaker diarization
- noise reduction
- digit recognition
- word error rate
- noise removal
- speech enhancement
- edge detection
- speaker independent
- acoustic models
- cepstral coefficients
- speech retrieval
- mobile devices