A visualized acoustic saliency feature extraction method for environment sound signal processing

Environment perception is an important research issue for both unmanned ground vehicles and robots. To improve the capacity of perception, a visualized acoustic saliency feature extraction (VASFE) method based on both the short-time Fourier transform (STFT) and the Mel-Frequency Cepstrum Coefficient (MFCC) for environment sound signal processing is proposed in this paper. Sound signal is visualized by using the STFT algorithm as local image feature and the Mel-Frequency Cepstrum Coefficient (MFCC) is used to represent the local acoustic feature of the signal. The proposed VASFE method is tested by the natural sound data which collected from real world of both indoor and outdoor environment. The results show that this method is able to extract the saliency features of both long-term and short-term sound signal successfully and clearly, and conducts to very distinguishable features for future processing of environment sound information.