REAL TIME ABNORMAL SOUND DETECTION AND CLASSIFICATION FOR HOME ENVIRONMENT

This paper is a final report for the final year project titled Real Time Abnormal Sound
Detection and Classification for Home Environment. This project utilizes signal processing
and signal analyzing techniques to classify any abnormal sounds in any house environment.
The project will be applied specifically for surveillance system in the house so that the
system could recognise and differentiate the abnormal sounds in house environment. There
are various algorithms available to identify and classify the abnormal event sound. In this
project, a system is designed to have abnormal sound detection and classification. For
abnormal sound detection, mean signal approach being used while for classification process,
there two main methods being used which are features extraction using Mel-Frequency
Cepstral Coefficient (MFCC) and classifier using Gaussian Mixture Model (GMM). Mel
scale frequencies in MFCC are distributed linearly in the low range but logarithmically in the
high range. These characteristic corresponds to the physiological of the hnman ear.
Therefore, MFCC has assurance of high accuracy in output. Gaussian Mixture Model (GMM)
technique is suitable for usage in Matlab software. Signals from sensor will be detected and
MFCC will extract the features which meet the requirement decided. The valuable features
will be feed in to the classifier GMM. For this project, a microphone will be used as audio
sensor, while all programming codes will be tested using Matlab software. As a result, the
project would be able to classify the detected any abnormal event sound. The user would be
able to employ the system as a supervision and precaution for any unexpected situation.