The Application Of Psychoacoustic Model In The Speaker Abnormal Sounds Detection

China is the first country of producing speakers. the output of speakers can reach severalhundred million one year. The unpredictable situation in speaker’s design and production leadto speakers fault. So, after production of speakers, the first process is “pure sound detection”.It means that speakers were put under nominal power sweep signal so that we can draw theconclusion whether the speaker is qualified which the users satisfaction with. Now, thedetection of speaker’s “pure sound” is based on man’s ears. This method is not only based onman’s sense of hearing and experience but also related to working environment and thesituation of a man. So this method is not conductive to automation and quality of production.Also, this way can damage man’s ears. Due to the above reasons, many speaker detectorsarise.In the abnormal sounds of speakers, it is audible sounds which affect man’s hearing, sothis part of sounds plays an important role in judging whether the speaker is qualified. But, incommon used speakers detectors, most of all are judging all the abnormal faults. This methodis useful to correct the faults in production line for some manufacturers. But in somemanufacturing enterprises which request high output, special in some small and cheapmanufacturing enterprises, are trend to whether the speaker has the audible sounds. Itstimulates the study of detecting audible sounds.Psychoacoustic model can pick up the appreciable sounds from speaker’s response. Thisarticle uses the psychoacoustic model into speaker detectors and provides standard data forspeaker manufacturer to correct manufacturing defects.This subject is derived from the national natural science funds. Combined with the needsof the electro-acoustic industry, this article studies how to apply psychoacoustic model tospeaker detection. Signal source was used to produce excitation signal. Use microphone tocollect the audio response of speakers and transfer it to PC. In PC, convert an analog signalinto a digital signal, then use HHT to pick up high order weak signal. Finally, usepsychoacoustic model to analyze high order weak signal and get audible abnormal signal, toprepare data for next module: pattern recognition. The research of this article mainly includes:(1)Use HHT to analyze audio response of speakers and get abnormal sounds.(2)For modal aliasing phenomenon of HHT, we use empirical mode decompositioncorrect.(3)Use psychoacoustic model I to analyze abnormal sounds which was picked up byHHT, and get the audible abnormal sounds. (4) Use wavelet packet decomposition to improve the situation of the low time-frequencyresolution that psychoacoustic model I having.(5)Transplant the HHT and psychoacoustic model in Matlab into C++, to link up with theother modules of speaker detectors.This article studies psychoacoustic model in theory and practice. By playing the audibleabnormal sounds which psychoacoustic model getting, and compare with the sounds whichremove audible abnormal sounds, we can conclude that applying psychoacoustic model intospeaker detector can pick up the audible abnormal sounds effectively, and prove the concise,accurate and efficient way.