VOICE DETECTION WITH NOISE REDUCTION USING DYNAMIC TIME WARPING AND MEL-FREQUENCY CEPSTRAL COEFFICIENTS ALGORITHM APPLIED TO HOME AUTOMATION

摘要

Voice produces multiple forms of sounds heard in varied pitches. This study aims to apply a noise reduction system using the Dynamic Time Warping Algorithm and Mel-Frequency Cepstral Coefficient built in a mobile application and using a microcontroller to communicate with electronic devices to increase the voice detection accuracy. The experiment used 320 voice datasets with different background noises from music, vehicles, air-condition units and people's noise. The result showed that the prototype model outperformed existing techniques and methods with a score of: i) 89.86% for both female and male voices, ii) 91.9% for pitch and background noise from air-conditioning units, iii) 86.98% for vehicle noise, iv) 91.68% for music noise, v) and 88.95% for people's noise, specifically women's voices.