One solution to the crime and illegalimmigration problem is the use of biometrics techniques and technology. Biometrics aremethods for recognizing a user based on uniquephysiological and/or behavioural characteristics ofthe user. This paper presents the results of an ongoingwork in using neural networks for voice recognition.
Crime and illegal immigration has reached unprecedented levels. Although the gap betweenthe rich and the poor is the main cause of this crime, it isthe person in the middle who ends up paying for it. Longterm and short term solutions are needed urgently.Conventional keys, access codes, access cards areproving ineffective as they are easily lost, stolen, copied,observed or left at home. The goal of this work is to comeup with innovative, but inexpensive, solutions to thecrime problem using emerging technologies such asbiometrics, artificial intelligence, computer networks,signal processing, etc. The long-term goal of the work isto design a black box that will identify a known userbased on characteristic features such as speech, image,etc.Biometrics are methods for recognizing a user basedon unique physiological and/or behaviouralcharacteristics of the user. These characteristics includefinger prints, speech, face, retina, iris, hand-writtensignature, hand geometry, wrist veins, etc. Biometricssystem are being commercially developed for a numberof financial and security applications. The task performedby this system can be classified into identification andverification. Identification involves identifying a userfrom a database of user characteristics whereasverification involves authenticating a user's identity usinga pattern in its database.Of all the above mentioned human traits used inBiometrics, the one that humans learn to recognize first isthe voice characteristic. Infants can identify the voice oftheir mothers and telephone users can identify a caller ona noisy telephone line. Furthermore, the bandwidthassociated with the speech is also much smaller than theother image based human traits. This implies quickerprocessing and smaller storage space
read more http://cslu.cse.ogi.edu/tutordemos/nnet_...recog.html