Research On Compressed Sensing Algorithm Of Speech Signal

As people improved quality of life and the level of scientific and technological advances,more and more people need a message,which requires more storage capacity and transmission loss,people need higher sampling level and efficiency.Because compressed sensing the "edge sampling edge compression" feature,so in recent years,people's interest in exploring it is also growing.Compressed Sensing main features are:under certain premise,the signal can be completely restored,will greatly increase the signal sampling rate can largely reduce the loss of data storage and transmission.In this paper,Newton's method as a reference,proposed regularized Newton method,Newton's method to good effect Reconstruction and contractility quickly together,and improved the old version of the regularization process,extending the definition of average energy packets.Let the average energy and the total energy Grouped together constitute the atoms grouped,allowing increased precision atomic selected,which is where this article to its innovation.Extensive testing showed that after improving regularized Newton method in addition to the reconstruction has higher accuracy,but also greatly enhance the speed of reconstruction,we can see positive improvements after the Newton method has a high value.Newton's method involves matrix inversion calculations have good computing power,if we can integrate improved over the quasi-Newton method and regularization idea,we can better reduce the computation time and computational reconstruction,so the idea is there practice necessary.In addition,the paper also tried to put the idea of using compressed sensing in the time domain processing speech signals.The second place of innovation of this paper is an adaptive voice compression perceived ideas.The experimental results show: the audio signal based on the difference between the opening and Voicing Adaptive Allocation observation points can greatly reduce the compression ratio,the test sound signal to noise ratio is also greatly enhanced;and the compression ratio in a certain area can also be measured greatly The data.The results of this paper can prove adaptive approach is to observe the points can be used,which greatly enhance the sound compression ratio.