Reducing Computational Complexity of Multichannel Nonnegative Matrix Factorization Using Initial Value Setting for Speech Recognition.
Taiki IzumiRyo AiharaToshiyuki HanazawaYohei OkatoTakanobu UramotoShingo UenoharaKen'ichi FuruyaPublished in: CISIS (2018)
Keyphrases
- speech recognition
- nonnegative matrix factorization
- data representation
- hidden markov models
- negative matrix factorization
- matrix factorization
- language model
- least squares
- automatic speech recognition
- speech signal
- pattern recognition
- speech recognizer
- speech synthesis
- principal component analysis
- spectral clustering
- speech recognition systems
- computer vision
- original data
- speaker identification
- speaker independent
- bayesian networks
- linear prediction
- support vector machine