一种扬声器异常音的时域特征检测方法

扬声器的异常音故障常见于生产过程中，通常可由有经验的听音员凭人耳听音检测出来。与人耳听测异常音的机理类似，一种在时域检测异常音的方法被提出。通过对声响应信号进行经验模态分解得到本征模态函数，再根据本征模态函数与激励信号的瞬时频率差对各模态去混淆处理，可得到包含异常振动信息的模态函数，据此可判断是否存在异常音。算法验证分为仿真和实验两部分。仿真验证中分析了模拟的合格扬声器与故障扬声器的声响应，结果表明算法对于摩擦和碰触导致的异常音是灵敏的。实验验证中检测了20只扬声器，测量得到的故障频率与用高阶谐波失真法得到的结果一致。且对于不同异常音故障，可在经验模态函数的局部特征中观察到不同的时域特征。

A method for detecting rub and buzz defects of loudspeaker in time domain

A method which can be used for detecting Rub and Buzz defects in time domain is presented. Empirical Mode Decomposition of sound response signal is performed and aliasing parts in Intrinsic Mode Functions (IMFs) are removed by referencing the difference between stimulus and IMFs. Vibration modes, which are caused by Rub and Buzz defects, are extracted and be used for fault detection. The validation was composed of simulation and measurement of defective samples. In the simulation, constructed sound responses representing quality and fault products were analyzed. Simulated symptoms from rubbing, buzzing and loose particles problems were detected. In the measurement validation, IMFs and high-order harmonic distortions (HOD) of sound responses from 20 samples were measured and analyzed. The results show that the frequencies and amplitude characteristics of defects are concordant with the results by HOD method. Furthermore, according to the type of defects, different time-domain features can be found in local features of IMFs.