Dataset Open Access
Shibata Hayato; Kato Taku; Shinozaki Takahiro; Watanabe Shinji
Deep neural networks (DNNs) were trained for posterior and bottleneck features using Japanese and other language speech data. We explore various DNN types, their combinations, and dimension reduction by principal component analysis (PCA).
This version (version 1) extracts DNN bottleneck features obtained from GMM based SAT features. The DNN and GMM were trained by speech data from the corpus of spontaneous Japanese (CSJ).