Fuzzy logic system for abnormal audio event detection using mel frequency cepstral coefficients

• Main Authors: Dadula, Cristina P., Dadios, Elmer P.
• Format: text
• Published: Animo Repository 2017
• Subjects: Buses—Transmission devices, Automatic; Buses—Safety measures; Voice frequency; Audio data mining; Sound—Analysis; Manufacturing
• Online Access: https://animorepository.dlsu.edu.ph/faculty_research/2033
• Institution: De La Salle University

This paper presents a fuzzy logic system for audio event detection using mel frequency cepstral coefficients (MFCC). Twelve MFCC of audio samples were analyzed. The range of values of MFCC were obtained including its histogram. These values were normalized so that its minimum and maximum values lie between 0 and 1. Rules were formulated based on the histogram to classify audio samples as normal, gunshot, or crowd panic. Five MFCC were chosen as input to the fuzzy logic system. The membership functions and rules of the fuzzy logic system are defined based on the normalized histograms of MFCC. The system was tested with a total of 150 minutes of normal sounds from different buses and 72 seconds audio clips abnormal sounds. The designed fuzzy logic system was able to classify audio events with an average accuracy of 99.4%.