Development of a Microcontroller Based Voice Recognition System for Accessing Bank Vault

Abstract

Security is a very essential aspect in the banking industry, and bank vaults access is limited to a certain number of individuals for proper management and accountability. Many security systems have been used to ensure the safety of customer’s assets ranging from password-based system, finger print based system, and combinational locks in the bank vaults. These forms of protection
can be hacked, and duplicated. The use of combinational lock doesn’t favour the physically challenged ones (handicap) and people with cognitive disability. The developed work aims towards overcoming the challenges in security of the current vault systems in the banking industry, by using voice recognition system approach with a relative spectra frequency coefficient- perceptual linear predictive (RASTA-PLP) as the applied feature extraction technique in order to aid access to the bank vault to authorized persons with a convenient and more secure means without any need of remembering a password, lock combination, and the use of magnetic card of which can be easily stolen. The functionality of this work was implemented and tested using a model bank vault system built and controlled by Arduino ATMEGA328p microcontroller and a voice recognition system designed on MATLAB R2015a which controls the access into the bank vault system and thereby protecting the vault from robbery. Vector Quantization (VQ) approach using LBG algorithm (VQLBG) was applied for feature matching. VQ distortion between the resultant codebook and RASTA-PLP coefficients of an unknown speaker was taken as the basis for determining the speaker’s authenticity. Euclidean distance was calculated for each signal to be recognized and matched to the closest speech signal in the speech database.