A Building Security System using Speaker Identification with an Artificial Neural Network Model as a Classifier

摘要

This work presents an implementation of a security system for elevators using speaker identification to validate the identity of each inhabitant The identity of each inhabitant is extracted from his or her voice by calculating features, such as pitch, formant frequencies, cepstral coefficients and mel cepstral coefficients. In order to classify the features, the system uses an Artificial Neural Network model called Multi-Layer Perception (MLP) as a classifier. The MLP is used because of its generalization property in patterns classification and low computational requirements. In addition, this work presents the comparison among the features used in the recognition process.