Introduction to Accent Classification with Neural Networks 1.1 2005/12/14 18:35:09.558 US/Central 2005/12/14 23:59:14.956 US/Central Scott Novich snovich@rice.edu Scott Novich snovich@rice.edu Accent Accent Classification Accent Detection Accent Recognition Classification Detection Neural Neural Network Recognition This module provides an overview of Team SSB's final ELEC 301 project -- accent recognition and classification with a neural network

Overview Although seemingly subtle, accents have important influences in many areas – from business, to sociology, technology, security, and intelligence. While much linguistic analysis has been done on the subject matter, very little work has done with regards to potential applications.

Goals The goal of this project is to generate a process for accurate accent detection. The algorithm developed should have the flexibility to choose how many accents to differentiate between. Currently, the algorithm is aimed at differentiating accents by languages, rather than regions, but should be able to conform to the latter as well. Finally, the application should produce an output showing the relative strength of a speaker's primary accent compared to the rest in the system.

Design Choices The agreed-upon option for achieving the desired flexibility in the project's algorithm is to use a neural network. A neural network is a matrix containing weights that correspond to how certain parameters fed to the network tie the inputs to the outputs. Parameters of known inputs with corresponding outputs are fed to the network to train it. Training the network produces the weighted matrix, to which test samples can then be fed. This provides a powerful and flexible tool that can be used to generate the desired algorithm. Utilizing this limits the project group only by the amount of overall samples collected to train the matrix with, and how they are defined. For this project, approximately 800 samples from over 70 people have been collected for the purposes of training and testing. The group of language-based accents to test with consists of American Northern English, American Texan English, Farsi, Russian, Mandarin, and Korean.

Applications Potential applications for this project are incredibly diverse. One example might be for tagging information about a subject for intelligence purposes. The program could also be used as a potential aid/error check for voice-recognition based systems such as customer service or bioinformatics in security systems. The project can even aid in showing a student's progress in learning a foreign language.