An improved phase-space voicing-state classification for co-channel speech based on pitch detection

This paper presents an improved phase-space voicing state classification method based on pitch detection to simultaneously determine the voicing state of two speakers present in a segment of co-channel speech. Three possible voicing states are considered: Unvoiced/Unvoiced (U/U), Voice/Unvoiced (V/U), Voiced/Voiced (V/V). Firstly, the method employs a phase-space voicing-state classification algorithm to classify co-channel speech into three parts: U/U frames, V/U frames and V/V frames. Secondly, in order to decrease misjudgment between V/U and V/V frames, we introduce mulitpitch detection based on enhanced summary autocorrelation function (ESACF) to modify the voicing states of V/V frames and single pitch detection based on autocorrelation function (ACF) to modify the voicing states of V/U frames. Experiments show the proposed method effectively reduces the classification error rate and outperforms the voicing-state classification algorithm only based on phase-space reconstruction.