Q-ASR: Integer-only Zero-shot Quantization for Efficient Speech Recognition.
Sehoon KimAmir GholamiZhewei YaoAniruddha NrusimhaBohan ZhaiTianren GaoMichael W. MahoneyKurt KeutzerPublished in: CoRR (2021)
Keyphrases
- speech recognition
- automatic speech recognition
- hidden markov models
- language model
- speech signal
- speech recognizer
- speech recognition technology
- speech retrieval
- pattern recognition
- speech synthesis
- noisy environments
- speech understanding
- speech recognizers
- word error rate
- speech processing
- handwriting recognition
- speech recognition systems
- speaker independent
- cepstral coefficients
- speaker dependent
- machine learning
- non stationary
- feature space
- keyword spotting
- conversational speech
- speaker adaptation
- information retrieval
- neural network