Single Channel Speech and Background Segregation Through Harmonic-Temporal Clustering
Jonathan Le Roux, Hirokazu Kameoka, Nobutaka Ono, Alain de Cheveigné, Shigeki Sagayama · 2007
The design of effective algorithms for single-channel analysis of complex and varied acoustical scenes is a very important and challenging problem. We present here the application of the recently introduced Harmonic-Temporal Clustering (HTC) framework to single channel speech enhancement, background retrieval and speaker separation. HTC processing relies on a precise parametric description of the voiced parts of speech derived from the power spectrum. We explain the positioning of the algorithm inside the Computational Acoustic Scene Analysis (CASA) area, describe the theoretical background of the method, show through preliminary experiments its basic feasibility, and discuss potential improvements.