The Study Of Threshold And Type Judgement In Loudspeaker’s Abnormal Sound Detection

In the production process of loudspeaker, all manufacturers hope that their production on the assembly line have characteristic of both high quality and excellent sound.So the detection of abnormal sound about loudspeaker is particularly important, although both at home and abroad, there are some testing equipment have been produced, with the disadvantage of expensive and their detection insufficient are very slowly. They not suitable for online monitoring, resulting in most businesses detection is dependent on the manual detection using Sweep judgment through pure tone seized listen. This subj ective detection method requires a separate listening station. Not only in the production line noisy environment cause a wrong judgment, but also caused by human hearing serious injury if long-term tolerance to high sound pressure levels generated by high-power speaker driver signal. With the increasing shortage of labor, workers’ physical and mental health more attention, as well as automation technology progress, study a reliable speaker abnormal sound detection technology is extremely important.According to these defects, the scholars also made a lot of speakers abnormal sound detection methods, all can detect effectively, however, they are only distinguish the loudspeaker whether the abnormal, and not detect their types. So, to further detect the loudspeakers’ type and extent are difficult and a new research direction. Detect the loudspeakers’abnormal extent can choose the trade-offs. Those that serious fault can be directly returned to the manufacturer production, and the minor fault can make some simple processing and correction. Detect the loudspeakers’abnormal type can targeted product troubleshooting, and improve the quality of products. Aim at the loudspeakers abnormal sound detection’s shortcomings, this paper studies the abnormal sound signal classification and identification of the loudspeaker deeper on the basis of previous research.The basic knowledge and the type of abnormal acoustic signal of the loudspeaker are introduced at first in this paper. Then the overall flow chart of the loudspeakers abnormal sound detection and the flow charts of each module are shown. After that, the abnormal sound threshold and abnormal sound type were analyzed deeply. The abnormal sound threshold includes two parts:one is abnormal threshold detection. It uses the pattern recognition method to do synthetic judgment through the result of electric response signal and sound response signal. If and only if the two signals both meet the requirements, the speaker is no abnormalities.The other part is abnormal level threshold module. Through doing short-time Fourier transform(STFT) to the response signal of loudspeaker for time-frequency diagram, then to each column of the time frequency chart, achieve the high frequency energy mean characteristic curve.The standard characteristic curve was obtained through multiple normal sample drawing characteristic curve repeatedly,the abnormal degree is determined though the distance that the measured speaker’s characteristic curve deviates from the standard curve.Finally,the MATLAB software is used to simulate all kinds of contrast diagram between abnormal degree and standard curve threshold;To abnormal signal type module, its mainly task is simulating all abnormal signals.Though doing wavelet packet decomposition to audible abnormal sound signal extracted from psychoacoustic model,the energy is achieved and normalize them as the eigenvalues, then input the characteristic value to the BP neural network to learning and training,the pattern energy of the measured speaker is input to the BP network,to achieve the recognition results. It is verified by experiments that this method can detect the type of the abnormal sound and abnormal threshold of the loudspeaker effectively.