Study Of The Auditory Signal Denoising Algorithms For A Mobile Robot

The significance of study of speech recognition on robot is not limited by time and place on the interactive. However,when robots get speech commands, they are affected inevitably by noise from the surrounding environment, room reverberation, as well as speech interference of other speakers. They reduce the quality of speech communication. In order to solve the practical applications of speech control robot, speech enhancement recognition become bottlenecks for robots capacity. In this paper, in a noisy environment, study the speech denoising about the mobile robot, the main work is as follows:(1) This paper in the introduction summarizes firstly speech denoising course of development as well as enhance the basic theory of the algorithm. The second chapter describes the speech and noise models, and performance evaluation methods. Discuss the single-channel speech enhancement algorithm, and focus on spectral subtraction, Wiener filtering method and Kalman filtering method denoising performance comparison.(2) The third chapter introduces the microphone array theory and explains the delay estimation algorithm, and finally compares the characteristics of various beam forming algorithm. Introduce some key technologies in speech enhancement process, focus on compare about the generalized cross delay estimation method and time delay estimation method based on adaptive filtering. Simulation results show that:the time delay estimation method based on adaptive filtering anti-noise performance is better than the generalized cross correlation time delay estimation method.(3) In the fourth chapter carried out a detailed analysis of spectral subtraction algorithm, the improved spectral subtraction algorithm is applied to the microphone array, improvements of speech enhancement algorithm come true. Improved spectral subtraction noise spectral gain selection. Simulation results show that:the improved spectral subtraction and microphone array combined algorithm is better than separate single an algorithm, and also is better than no improved algorithm.(4) Chapter V compares the performance of the robot moves around the microphone array speech enhancement, Finally, verify the effect of speech enhancement speech recognition algorithm. Experimental results show that the joint algorithm to improve speech recognition rate significantly. Microphone array speech enhancement algorithm based the spectral subtraction is the focus of this work. The proposed modified algorithm compares with traditional algorithm to eliminate noise and reduce speech distortion, which has been a marked improvement to achieve the intended purpose.