Design And Implementation Of Sound Recognition System In Factory Noise Environment

With the rapid development of our country’s manufactory industry, thedemands of automatic sound detection of product sound on production line aregreatly increasing. Most of the current sound detection methods basically rely onlistening of human ears. This may increase both the product cost and the errors.Moreover, the hearing will be badly damaged if the ears are under high pitch soundfor long time. Therefore, with more and more demands for the automatic production,more and more high quality automatic sound detecting systems are in need. Thispaper studied the sound recognition system in noisy factory environment. It wasdesigned based on the actual demand of toy manufacturer and had highpracticability.According to the requirements of automatic sound detection in toys production,this paper analyzed and studied the key technologies involved in the soundrecognition system, such as basic principles of sound recognition system, in-depthstudy on voice signal processing, voice recognition feature parameter extraction,pattern matching algorithms and so on. Firstly, in terms of the analysis of the toyssound characteristics in noisy factory environment, we used a method whichcombines short-time energy and short-time zero crossing rate for the endpointdetection. Secondly, in terms of the analysis and comparison of LPCC and MFCC, weproposed new characteristic parameters, which synthesized Mel-Scale cepstrum(MFCC), inverse Mel-Scale cepstrum (IMFCC) and their differential parameters. Thirdly, in terms of the issues that pattern match methods in traditional speechrecognition system are no longer suitable because of the special characteristics of thetoy’s sound, we proposed a new No Warping sliding pattern match algorithm forfinding the shortest distance.The following are the major contributions:(i) This dissertation did detail demandanalysis for the system according to the toy manufacturer’s demand.(ii) Simulated allthe algorithms on MATLAB.(iii) Designed both hardware and softwareimplementations for the sound recognition system under the actual production lineenvironment.(iv) Realized and tested the system using MYSQL and Visual C++on theWindows platform.