Research on Novel Normal Fuzzy Kalman Filter for Speech Enhancement in Noisy Environment
Mao-Lin Chen, Kai‐Jung Chen · 2021 IEEE International Conference on Power Electronics, Computer Applications (ICPECA) · 2021
The filter design plays a vital role in digital signal processing. A novel filter for reducing the noisy signal's effect was developed for analyzing a noisy speech signal in the time-domain and frequency-domain. The novel filter, namely normalized fuzzy Kalman (NFLK) filter, was based on the Normalized least mean squares (NLMS) filter, fuzzy logic filter, and Kalman filter. Through mathematical derivation, the numerical model of the NFKL filter was developed. The NFKL filter was validated using a speech signal with noise and compared with three standard filters as KALMAN filter, NLMS filter, and recursive least squares (RLS) filter. The spectrum signal comparison of before-after filter used the signal to noise ratio to quantify each filter's efficacy. The comparison results proved that the NFLK filter could effectively improve the noisy speech signal with a slight SNR decrease. Thus, the novel NFLK digital filter achieved the desired specified requirements for the noise reduction of the speech signal, and its speech enhancement performance better than traditional filters.