Speech and Noise Dual-Stream Spectrogram Refine Network With Speech Distortion Loss For Robust Speech Recognition.
Haoyu LuNan LiTongtong SongLongbiao WangJianwu DangXiaobao WangShiliang ZhangPublished in: ICASSP (2023)
Keyphrases
- speech recognition
- speech signal
- noisy environments
- automatic speech recognition
- spectral subtraction
- noisy speech
- speech synthesis
- speech recognizer
- background noise
- speaker identification
- hidden markov models
- speech enhancement
- pattern recognition
- language model
- speaker verification
- speech recognition systems
- speech processing
- linear prediction
- speaker recognition
- speech recognition technology
- vocal tract
- word error rate
- noise model
- recognition engine
- keyword spotting
- voice activity detection
- non stationary
- noise reduction
- digit recognition
- speech retrieval
- word recognition
- handwriting recognition
- broadcast news
- speech recognizers
- machine learning
- dialogue system
- acoustic models
- speaker dependent