Speaker identification using wavelet transform and probabilistic neural network

Abstract EN

Speaker identification is the process of automatically identify who is speaking on the basis of individual information included in speech waves.

This technique makes it possible to use the speakers voices to verify their identity and control access to services such as voice dialing, banking by telephone, telephone shopping, database access services, information services, voice mail, security control for confidential information areas, and remote access to computers.

The goal of this project is to build an automatic speaker identification system for a closed set Text-Dependent & Text-Independent.

This generally will include two main phases : The Training Phase, which is used to built the speakers database and the Identification (Testing) Phase, which is used to compare the unknown model with the models stored in the speakers Data base.

The wavelet transform is diffused into most digital signal processing applications.

It is plays very important role in speech signal processing and analysis, and mainly in speaker identification because of its superior performance when used particularly in multi-resolution analysis.

The proposed system constructs from three stages, the first stage is the Preprocessing stage, in which the speech signal is separated into many frames, and each frame is multiplied by (Hamming Window).

In feature extraction stage, which is the second stage, the discriminative features of each spoken words are extracted by using the DWT technique, the resultant of this stage is the feature vector for each speaker.

In the third stage, which is the classification stage, the feature vector of each speaker is used as an input to the neural network.

The results show that phonetic features are powerful for speaker identification and the proposed algorithm is efficient concerning the minimizing of the calculation operations and reducing the execution time.