摘要
为解决钱塘江涌潮检测难问题,提出了一种新的涌潮检测方法。首先,通过梅尔倒谱系数MFCC(Mel-Frequency Cepstral Coefficients)提取涌潮样本多维声学特征与非涌潮样本声学特征;然后,使用支持向量机(SVM)构建涌潮检测模型;最后,通过模型对输入的样本特征进行分类判断。与以采用线性预测倒谱系数(LPCC)提取声学特征方法或是采用BP神经网络构建检测模型相比,MFCC与支持向量机结合在涌潮检测的精度上有一定的提高。
In order to solve the problem of detecting Qiantang River tidal bore,this paper proposes a new detectionmethod for the tidal bore. At first,getting the sound of tidal bore and non-tidal bore Mel-frequency Cepstral Coeffi-cients（MFCCs）as the acoustic features. Then,using support vector machine（SVM）to construct a detection modelfor tidal bore. Finally,classifying input sample acoustic features by the model. Compared to using linear predictioncepstral coefficients（LPCC）as the acoustic features or constructing model by BP neural network,the new approachto detect tidal bore via the support vector machine（SVM）with the Mel-frequency Cepstral Coefficients（MFCCs）asthe acoustic features reach a higher recognition accuracy.
出处
《传感技术学报》
CAS
CSCD
北大核心
2016年第11期1773-1778,共6页
Chinese Journal of Sensors and Actuators
基金
国家自然科学基金项目(61374005)
浙江自然科学基金项目(LY14F030022)
