International Conference on Advances in Science and Technology |
Foundation of Computer Science USA |
ICAST2014 - Number 4 |
February 2015 |
Authors: Sneha B. Lonkar, Nadir N. Charniya |
242cf39e-5f67-46a2-88bb-c368fb607381 |
Sneha B. Lonkar, Nadir N. Charniya . Design of Optimal MLP NN for Speaker Dependent Spoken Words Recognition Application. International Conference on Advances in Science and Technology. ICAST2014, 4 (February 2015), 14-18.
Spoken words recognition provides applications like spoken commands recognitions in robotics command, speech based number dialing for phones and mobiles, etc. It also provides applications in railway and banking areas. This work aims at designing of optimal Multilayer Perceptron Neural Network (MLP NN) based classifiers for speaker dependent spoken digits recognition. The classifier attempted as optimal leading to less number of computations and few components requirement for its future implementation in hardware leading to a low cost speech recognition system. Isolated spoken digits were used as an input data to the neural networks based classifiers. Each spoken word was analyzed for the feature like Mel Frequency Cepstral Coefficients (MFCC). The MLP NN based classifier was designed meticulously with the condition of minimum components and attempting maximum classification accuracy.