Practical Inter-Floor Noise Sensing System with Localization and Classification

Abstract

Inter-floor noise is a severe social problem which causes illegal arson, violence, and even murder. In this paper, an inter-floor noise sensing system is proposed to detect and record information related to inter-floor noise in an apartment building. The proposed system measured the noise level and estimated the direction of the noise source along with the type of noise. The noise level measurement is based on the sound pressure level (SPL) measurement, which is a logarithmic measure of the effective pressure of a sound relative to a reference sound pressure. Noise source localization was performed using the estimated time difference of arrival (TDOA) from the microphone array. For the classification of noise types, the Mel frequency cepstral coefficients (MFCC) and zero-crossing rate (ZCR) were extracted from a noise signal, and the k-nearest neighbor algorithm was used to classify the type of noise. In addition, we developed a noise monitoring hardware to evaluate our methods in the actual environment. The experimental results demonstrated that the proposed system had a reliable accuracy for each functional unit. The results showed that the error of the noise level was approximately +/- 1.5 dB(A), the error of the direction estimation was approximately +/- 10 degrees, and the accuracy of the classification for the noise type was more than 75%. These output data from the proposed system are expected to be used as important reference data for any dispute cases due to inter-floor noise.