Noisy Speaker Signal Adaptive Feedback Reduction by Using Time Domain Kalman Filter and Prediction Error Method Based-On Partition Block Frequency Domain Kalman Filter

Acoustic feedback is often formed when the signal source is placed in front of the speaker. This creates an acoustic loop. Acoustic feedback reduces sound quality and produces artificial sounds. Acoustic feedback elimination methods can be used to improve Maximum Stable Gain (MSG) and sound quality. One of the approaches to deal with acoustic feedback is to use the error prediction method based on the Kalman frequency domain filter using feedback partition block. Simultaneous use of the Kalman filter in the frequency domain and the segmentation of the feedback path has improved the criterion of maximum stability compared to methods such as time domain adaptive filters or without partition block of the feedback path. But this approach does not work well in the face of noisy input. In acoustic environments, the presence of ambient noise prevents the signal from being received without noise. In this study, after preprocessing (noisy signal with Gaussian noise with standard deviation of 0.0 to 0.5 dB), two steps: 1. Using Kalman filter of time domain and 2. Using frequency domain Kalman filter method using partition block of the feedback path is provided to improve the MSG criterion of about 3 to 4 dB.