Fig. 1From: Single microphone speech separation by diffusion-based HMM estimationTraining in HFHMM scheme. The training phase of the hybrid log-spectral-diffusion approach, for speaker 1 (top) and speaker 2 (bottom)Back to article page