MetricAug: A Distortion Metric-Lead Augmentation Strategy for Training Noise-Robust Speech Emotion Recognizer.
Ya-Tse WuChi-Chun LeePublished in: INTERSPEECH (2023)
Keyphrases
- noisy environments
- speech recognition
- geometric distortions
- image noise
- digit recognition
- speech signal
- isolated word
- additive noise
- automatic speech recognition
- speech recognizer
- word recognition
- emotion recognition
- estimation error
- speech corpus
- hidden markov models
- speaker identification
- quality metrics
- training set
- signal to noise ratio
- noise reduction
- noisy data
- distance measure
- speech enhancement
- facial expressions
- text to speech synthesis
- hearing impaired
- emotional state
- robust estimation
- quality assessment
- metric space
- spoken language
- missing data
- emotional speech
- video sequences