Filtering And Spectrum Analysis For Audio Signals Based On Wavelet

With the development of technology, the important role of audio signal has been moreand more embodied in the actual aspects of life, such as mobile communication, radio,television and some important engineering projects. However, the actual audio signal causesus great influence because of the containing noise. At present, the most effective method forthe audio signal filtering is the use of wavelet transform. Wherein, wavelet threshold filteringalgorithm is widely used, while several methods of choosing threshold are not on the basis ofsignal decomposition level, which will cause noises cannot be completely removed or thefeatures of effective signal is filtered out by mistake. Thus, when wavelet threshold filteringalgorithm is used for audio signal filtering, the decomposition level should be considered forchoosing the threshold. Besides, the frequency spectrum analysis, power spectrum estimationand other comprehensive analysis should be considered sufficiently.The process of the audio signal filtering and spectrum analysis is divided into two steps.The first step is the application of wavelet threshold filtering algorithm for audio signalfiltering. An improved method is proposed the adaptive selection of optimal decompositionlayer is based on the comparison of high-frequency coefficients of adjacent layers. Besides,the choice of adaptive optimal threshold is based on the decomposition layer. The resultsshow the improved algorithm in two indicators is better than the three methods above. At thesame time, it effectively retains the original features of audio signal. The second step is to usethe wavelet packet analysis in frequency spectrum analysis and power spectrum estimation forfiltered audio signal. A new method is proposed based on the minimum entropy principle andwavelet binary tree. Then the equivalent AR model is established, the parameters of ARmodel are solved by the Yule-Walker algorithm, the noise variance is estimated and the powerspectrum simulated during the period of power spectrum estimation. The results show theYule-Walker algorithm is best in Spectrum characteristics, anti-aliasing properties and theestimation of noise variance of all methods. The above work can make the audio signal betterhandled during the processing of filtering and frequency spectrum analysis.