Implementation of Mel Frequency Cepstrum Coefficients and Dynamic Time Warping Algorithms on Door Physical Security System using Voice Recognition Pattern

Abstract

The door physical security system with voice recognition pattern is intended to amplify the security system and make it easier for users to access the house entrance. The experiment was carried out on a proportional small-scale prototype consisting of input media in the form of a reed switch sensor and a microphone on the user's smartphone, an actuator in the type of a solenoid door lock controlled by a relay, and the output media as a display on the Android application. By using a sensor reed switch to read the condition of a closed or open door, a servo motor to move the automatic entry and a solenoid door lock to perform automatic door locking. We can match the voice of the user entered with voice data that has been recorded on the embedded system. The application of the Mel Frequency Cepstrum Coefficients method in determining the extraction of sound features and Dynamic Time Warping to determine the sound match score. If the sound matching process is successful (valid), the system will give an output in the form of the action of opening, closing, and locking the house door automatically. The success rate of the system in doing pattern voice recognition is 80-90%. The optimal distance for accessing house doors in this system is less than 10 meters. The total response time required by the system to open and close and lock the user's door based on the allowed user input is ± 6 seconds.