Multi-stream spectro-temporal and cepstral features based on data-driven hierarchical phoneme clusters.

Shang-wen Li, Liang-Che Sun, Lin-Shan Lee

Browse the full ICASSP paper archive.