Cancellation Quality Assessment with Spectrum Segmentation for Speech Communication

The spectrum segmentation method is proposed for the interference cancellation quality assessment, which not only integrates the spectrum distribution characteristic of speech and interference, but also introduces the statistical optimization model for calibrating the assessment coefficient. Hence, there will be a remarkable improvement of SINR estimation accuracy and real-time performance of speech signals under various common interference. Implementation in hardware and numerical experiments demonstrate state-of-the-art performance of proposed method under the AM pattern. It is ideal for the closed-loop interference cancellation of speech communication.