为了有效地检测扬声器异常音，提出一种基于希尔伯特-黄变换的异常音识别及去噪方法。分析了小波变换的不足，再进行HHT分析。建立常见的几类扬声器异常音数学模型，并经EMD分解得到包含异常振动信息的IMF分量，利用IMF分量特点进行去噪处理。再求出各模拟异常音信号的Hilbert谱，采用图像二值化技术处理Hilbert谱，和小波变换时频谱比较后证明了HHT方法具有更多的优势。而且，根据不同类型异常音在时频域的特征，即可判断出扬声器异常音类型。matlab仿真实验结果表明，当归一化阈值取为0.065时，会得到清晰的图像，能更精确地检测出异常音类型。%In order to detect the loudspeaker defects effectively, a new method of recognition and denoising based on HHT algorithm for loudspeaker defects is presented. The shortcomings of the wavelet transform are analyzed, the HHT analysis again. Several loudspeaker defect models are established to be processed by the use of the EMD method, in which the characteristics of the IMF component including abnormal vibration information are obtained for denoising. The Hilbert spectrum of loudspeaker defect signal is obtained to be processed by the use of the image binarization technology. It proves that the HHT method has more advantages in comparison with wavelet transform. What’s more, according to the characteristics of different loudspeaker defects in the time-frequency domain, the types of the loudspeaker defects can be well detected. Matlab simulation results show that, a clear image will appear when the normalized threshold is 0.065, with which the types of loudspeaker defects can be detected more accurately.
展开▼
