SNR and sub-band SNR estimation based on Gaussian mixture modeling in the log power domain with application for speech enhancements (第6回音声言語シンポジウム)