11-05-2013, 04:53 PM
Speaker Identification using Neural Networks
Speaker Identification.pptx (Size: 288.97 KB / Downloads: 19)
Introduction
Biometric security systems are convenient for users because it is not necessary to remember passwords or carry identification cards.
The particular characteristics of the sound and the way of talking of each person are unique.
It is easy to distinguish the sound of people using computers, even though the sounds we hear cannot be distinguished by our ears.
Thus, we can precisely recognize each person by analyzing their voice using computers.
Algorithm:Backpropagation
In the design of the speaker identification system, a recognition method of the features extracted is chosen. For this Backpropagation algorithm is used. It has two phases.
1. Feed forward phase : In this phase, the external input information at the input nodes is propagated forward to compute the output information signal at the output unit.
2. Backward phase : In this phase, modification to the connection strengths is made based on the differences between the computed and observed information signals at the outputs units.
Conclusion
A module of Neural Network based Speaker Identification System (Neural Network training) is built.
Another module of Speaker Identification System (Feature Extraction) is to be made in the coming days of project work.