The aim of the project was to develop a Limited Vocabulary Isolated Word Recognition System adopting the Hidden Markov Model to statistically model the words in the dictionary. For the project an Irish database containing 10 numerals from zero-nine, spoken by 10 different speakers was used. So, this speech recognition system is speaker dependent.
The project involved the following modules
Feature Extraction Vector Quantization HMM Training or Building the HMMs Recognition
Overall, there were 100 samples
of speech ( 10 words x 10 speakers ). Out of this, 90 were used for training
and 10 were used for testing.
On the training set, 100% recognition was achieved. On the test set, the recognition made 2 mistakes. So, on the whole the performance of the recognizer was good. However, the performance of the system can be improved if we have more training samples.
All the MATLAB Code, Codebook vectors, HMMs can be downloaded here
Download ZIP File
Back to my Home