基于小波包和FCM聚类的电能表内异物检测

Foreign object detection in electric energy meter based on wavelet packet and FCM clustering

针对采集的电能表内异物晃动产生的声音信号,通过对声学识别流程传统端点检测方法的研究,提出了一种新的电能表内异物声音信号端点检测方法。结合小波包能量特征、短时TEO对数能量和模糊C-均值聚类(FCM),使用提取的小波包能量特征同支持向量机(SVM)完成电能表内异物声音的训练与识别。相较传统的基于阈值的端点检测算法,该端点检测算法处理后的异物检测准确率明显较高,能够更好地检测电能表内异物。

For the sound signal generated by the shaking of foreign objects in the electric energy meter,a new endpoint detection method for foreign matter sound signal in electric meter is proposed by studying on traditional endpoint detection method of acoustic identification process.The method combines wavelet packet energy features,short-time TEO logarithmic energy and fuzzy C-means clustering(FCM),and finally uses the extracted wavelet packet energy features and support vector machine(SVM)to complete the training and recognition of foreign matter in the electric energy meter.Compared with the traditional threshold-based endpoint detection algorithm,the endpoint detection algorithm has higher accuracy of detecting on foreign objects,and can better detect foreign objects in the energy meter.