Predicting Speech Intelligibility of Enhanced Speech Using Phone Accuracy of DNN-Based ASR System.
Kenichi AraiShoko ArakiAtsunori OgawaKeisuke KinoshitaTomohiro NakataniKatsuhiko YamamotoToshio IrinoPublished in: INTERSPEECH (2019)
Keyphrases
- speech recognition
- automatic speech recognition
- speech signal
- speech recognizers
- speech synthesis
- word error rate
- noisy environments
- acoustic models
- speech recognizer
- spontaneous speech
- hidden markov models
- emotion recognition
- computational cost
- speech corpus
- emotional speech
- pattern recognition
- audio visual
- high accuracy
- recognition errors
- language acquisition
- broadcast news
- recognition engine
- speaker independent
- conversational speech
- dialogue system
- speaker recognition
- highly accurate
- computational efficiency
- prediction accuracy
- non stationary
- endpoint detection
- spoken words
- hearing impaired
- audio stream
- handwriting recognition
- speech recognition systems
- vocal tract
- text to speech
- neural network
- mobile phone
- language model
- classification accuracy
- feature selection