Text Independent Speaker Recognition and Classification using KNN Algorithm

The method of automatically identifying the speaker using the speaker-specific data included in voice waves is known as speaker recognition. For speaker recognition, a variety of uses have been investigated. Monitoring, speech-activated secure access control, voice-activated customization of services or information for certain users, instances include using recorded voice samples in forensic and criminal investigations. The application that is now mentioned most often is access control, which also includes voice dialing, banking, telephone shopping, and database access services. Thus, it is projected that speaker recognition technology would provide new services in smart environments and enhance the comfort of daily life. Research has been done on the phenomenon known as “speaker idolization,” which occurs when speakers are automatically added to an input audio channel. It makes speech recognition easier, makes it easier to search and index audio archives, and gives machine transcriptions more depth and intelligibility. An important additional application for voice recognition technology is as a forensics tool. The speaker’s short-time spectral coefficients are described using vector quantization using a codebook. The success of these techniques is assessed from the perspective of robustness against utterance variation, such as variances in content, temporal variation, and changes in utterance pace. The voice of each individual is recorded three times. The experiment’s double distance measurement result is 96.97%, whereas the KNN technique’s single data center result is 84.85% The outcome shows that the twofold distance method increases the precision of voice recognition.