An Subarea Localization Algorithm based on Combination Features using Representative Audio Fingerprint

Abstract

To solve the problem of poor subarea localization performance of a single audio feature of ambient sound in a room, an indoor localization algorithm based on audio combination features is proposed. This paper utilizes three audio features of ambient sound to combine a new feature named CSIE, which are chromagram, sonogram and improved energy spectral density, and these features can express the tonal energy, frequency characteristics of audio and signal energy in the unit band, respectively. As the measured data shown, the subarea localization algorithm proposed has good performance. The accuracy of the feature CSIE for indoor subarea localization has reached 81% and 85% in two different rooms. Compared with single features, the accuracy has increased by at least 7%, up to 60%.