Study Of Adaptive Noise Cancellation Based On Un-uniform Sub-band

With the development of society and industrialization, noise pollution increasingly becomes a major source of environment pollution.Various of noise snot only affect the practical engineering and people’s production and life, but also threaten people’s healthy. Adaptive noise cancellation can effectively eliminate the disturbance of noise and extract useful signal from the environment when the noise characteristic is unknown and the noise transmission is changing. This paper focuses on the technology of adaptive noise cancellation. The main works of this article are as follow:(1)A new algorithm for nun-uniform sub-band decomposition and reconstruction based on Power Spectral Density(PSD) is researched. The power spectrum is estimated by using the method of cycle diagram. By dividing the amplitude of PSD, implement the group of spectrum. Then shift the spectrum and the non-uniform subband decomposition of the signal based on power spectrum is finished. The signal reconstruction process can be finished by the reverse stemps. The eigen-value spread of sub-band signal and signal reconstruction performance are analyzed by using Matlab. The simulation results show that, compared with the existing methods with equal number of sub-bands, the proposed one can effectively control the eigen-value spread of sub-band while maintaining well signal reconstruction performance.(2)The proposed algorithm is used in adaptive noise cancellation. The desired signal and reference signal use the same decomposition method based on power spectrum. In different sub-bands, spectrum shifting, decimation and noise cancellation are finished independently. Then through spectrum restoration and sub-band signal combining,the full band signal is restored. Compared with the full band method, the proposed one can effectively realize the process of adaptive noise cancellation and its convergence speed is faster than the full band method.(3)The problem of model mismatch is researched.For the problem of adaptive filter tap-length mismatch, the variable tap-length algorithm is introduced. In this method, the cost function is defined as steady-state mean square fragment error. The filter tap-length is iterative calculated in the form of fraction. The simulation results show that the algorithm can converge to appropriate tap-length with the initial order when the best tap-length message is unknown.(4)Adaptive noise cancellation experiment is research. According to the principle of adaptive noise cancellation, the experimental system is designed and constructed. The cancellation performance and model mismatch are studied. When the data is collected in experiment environment, a piece of music is used as useful signal. The noise including sine noise, white noise and air fan noise. After the data collecting is finished, noise coherence and cancellation are analyzed. The results show that the noise coherence is related to the sound of the noise, the distance of the microphone and so on. Under the interference of air fan noise, compared with uniform sub-band algorithm, the SNR of proposed sub-band algorithm is 1.0521 d B higher than the uniform sub-band algorithm when the tap-length is 50. For the problem of model mismatch, the FTLMS algorithm can realize noise cancellation process. But because of the complex of the environment, the tap-length is floating.