A Real-Time Voice-Control Model to Access Protected Resources

The paper presents a model developed for real-time speaker recognition system that can be used to access restricted information or resources using human voice. Speaker recognition involves identification and verification of the speaker. At each stage, the voiceprint is compared with model voices of all speakers in the database. The comparison is a measure of the similarity (score) from which rejection or acceptance of the verified speaker is chosen. The speaker recognition noise cleaning techniques used were the Linear Predictive Coding (LPC) and Cepstral analysis for feature extraction. The system was trained and tested using a population of ten users, with additional ten impostors. The average call success of a true speaker was 96.5%. The impostor’s success rate was found to be 0%. The impostor’s success rate was statistically computed to be 0.0036%, which shows that the speaker recognition system exhibited good performance in real-time operation.