Audio sounds classification using scattering features and support vectors machines for medical surveillance

This paper proposes a new approach to recognize environmental sounds for audio surveillance and security applications. The sounds are extremely versatile, including sounds generated in domestic, business, and outdoor environments. Since this variability is hard to model, investigations concentrate mostly on specific classes of sounds. Among those, the system that is able to recognize indoor environmental sounds may be of great importance for surveillance and security applications. These functionalities can also be used in portable teleassistive devices to inform disabled and elderly persons affected in their hearing capabilities about specific environmental sounds (door bells, alarm signals, etc.). We propose to apply an environmental sounds classification method, based on scattering transform and the principal component analysis (PCA). Our method integrates ability of PCA to de-correlate the coefficients by extracting a linear relationship with what of scatter transform analysis to derive feature vectors used for environmental sounds classification. The performance evaluation shows the superiority of this novel sound recognition method. The support vector machines method based on Gaussian kernel is used to classify the datasets due to its capability to deal with high-dimensional data. Our SVM−based multiclass classification approach seems well suited for real-world recognition tasks. Experimental results have revealed the good performance of the proposed system and the classification accuracy is up to 92.22%.