Noise robust feature for automatic speech recognition based on mel-spectrogram gradient histogram.
Taejin ParkSeungkwon BeackTaejin LeePublished in: SLAM@INTERSPEECH (2014)
Keyphrases
- automatic speech recognition
- noisy environments
- speech recognition
- speech signal
- word error rate
- hidden markov models
- conversational speech
- speaker verification
- broadcast news
- mid frequency
- noise reduction
- speech retrieval
- feature set
- signal to noise ratio
- noise model
- word recognition
- language model
- speaker identification
- spontaneous speech
- edge detection
- spoken words
- information retrieval systems
- speech corpus
- gray level
- speech synthesis
- wavelet transform