Configuring artificial neural network using optimisation techniques for speaker voice recognition
Namburi Dhana Laksmi, M. Satya Sai Ram · International Journal of Bioinformatics Research and Applications · 2022
Speaker recognition is proposed in this work using artificial neural network (ANN) and optimisation technique which finds wide variety of applications. Mel-frequency cepstral coefficient (MFCC) and linear prediction-filter coefficients (LPC) coefficients are utilised to extract features from voice signal as preliminary process. In this work, these features are applied to ANN to recognise speaker. This research focuses on configuring conventional ANN structure with zero hidden layers to multiple hidden layers. It is possible to improve the accuracy by utilising the large training dataset or by increasing the number of hidden layers. Over-fitting and under-fitting problems can be addressed by optimising the number of hidden layers. Genetic algorithm is applied to optimise hidden layers and the number of neurons for performance enhancement. The result reveals that the performance of GA in configuring ANN accomplishes 98% accuracy which is superior to conventional ANN. This research utilises two types of databases.