A master-slave architecture for parallel speaker recognition
Availability Unrestricted Thesis/Dissertation
Abstract
Speaker recognition is one of the popular research interests in speech processing. A speaker recognition system receives the speech signal (data) and determines who the speaker is from a known set of speakers. This process involves the task of matching the input speech signal to the models for all the speakers enrolled in the system. Important factors that determine the success of these systems are response time and accuracy.
The objective of my thesis is to optimize response time by dividing the task of recognition into a number of sub tasks and to execute these individual tasks on load- balanced multiple processors. There has been limited research in improving the response time with the use of parallelism. This idea has been implemented by using a master-slave model in which the master divides the recognition tasks and initiates their parallel processing on multiple slaves. Tests performed showed that the response time achieved is better than those obtained from the conventional system, which does not involve any parallel processing. This thesis justifies that a parallel processing approach can be used to optimize the response time of a speaker recognition system.
Rights Statement
In Copyright. URI: http://rightsstatements.org/vocab/InC/1.0/
This Item is protected by copyright and/or related rights. You are free to use this Item in any way that is permitted by the copyright and related rights legislation that applies to your use. For other uses you need to obtain permission from the rights-holder(s).