Background-noise independent sound recognition using imagerial spectrogram patterns

Abstract

Audio recognition is defined as the task of recognizing a particular piece of audio (could be music, ring-tone, speech and singing as well, from a given sample set of audio tracks. The field of audio recognition tries to emulate this behavior by using concepts from Biological modeling, signal processing theory and pattern recognition theory. Several techniques have been proposed to solve the problem of audio recognition. Most of the proposed methods are divided into two processing steps: feature extraction and classification. This research proposes a Background Noise Independence Sound Recognition algorithm that is able to automatically recognize a piece of audio with background by using the concept of spectrogram pattern matching. Each signal is analyzed and generated to its spectrogram that is used to train data for the classifier. Several classification functions are used, such as feed-forward neural network and k-Nearest Neighbor. This research applies a concept of matching of spectrogram pattern with various audio problem singing voice recognition and the environment sound recognition.