中文摘要：

传统的扬声器出厂质量检测方法有两种：试听员听音检测和扬声器频域特性检测。听音检测方法没有量化标准,不同试听员测试结果有所差异,而且对试听员的听力有一定损害。频域特性检测方法只能检测该扬声器的物理特性,而对音色的检测作用有限,常常遇到频域曲线一致,而音色相差甚远的情况。充分考虑了音色对扬声器品质的影响,提取了能够反映声音音色属性的特征,利用自动分类方法对扬声器质量进行判别,同时比较了不同特征之间的运算量。所提方法充分考虑了人耳听觉特性,并有效地提高了识别率。

英文摘要：

In the quality inspection process of speakers,the traditional methods involve listening inspection and frequency domain detection.However,listening inspection does not have quantitative criteria,which leads to different results by different inspectors.In addition,it is harmful to the workers’ hearing.Detecting the characteristics in frequency domain can only detect some physical characteristics and rarely consider the role of timbre.Sometimes two loudspeakers have the same frequency curve with different timbre.In this paper we fully account for the effect of timber on the loudspeakers’ quality,extract the characteristics that reflect the representation of timbre,and use automatic classification method to determine the quality of loudspeaker,meanwhile comparing the amount of computation for these characteristics.This method fully takes into account the characteristics of human ear that suit more to people’s listening habits,and meantime increases the recognition rate.