thehanimo / speaker-identification Goto Github PK

Speaker Identification is rapidly gaining popularity with the rise of Voice Assistants and the like. Personalization of such products often require Speaker Identification. There are numerous issues that need to be addressed before Speaker Identification can be implemented on a large scale. At the highest level, all speaker recognition systems contain two main modules:\textit{feature extraction and feature matching.} One popular feature extraction technique, MFCC, is often susceptible to noise. MFCC is used in many speech identification applications and has a range of other applications. VQ, a classical feature mapping technique, works by dividing a large set of vectors into groups having approximately the same number of points closest to them. In this project, a user authentication system has been realised with a second-layer of security by speaker identification, implemented by training a single recording of a passphrase upon registration and testing a real-time recording against the trained model to authenticate user access using Matlab.