Institutional Repository of Vilnius University
Žmogaus balso atpažinimas naudojantis neuroniniais tinklais /
Abstract
dc:descriptionThis paper is about human voice command recognition using neural networks. Different models will be compared to find the best solution. This is an important field to research for its practical use in medicine, communication, business. Two models have been implemented and compared to each other. A convolutional neural network M5 and a recurrent neural network LSTM. Firstly both models were trained on SPEECH COMMANDS data set and achieved 87\% and 82\% accuracy. With the convolutional model performing better. Then both networks were trained again using SYNTHETIC SPEECH COMMANDS data set and performed better than before increasing there accuracy to 90\% and 86\%. Thus pointing out the usefulness of computer generated data for training models. Also a test after every training session was implemented using data out of the training set, the accuracy of both models of course was lower. Overall the convolutional network performer better just after training and also testing with the new data out of the initial training set. The recurrent network did improve better having more data to train with.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2021
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Zapolskas, Lukas,
- Contributors dc:contributor
-
- Kybartas, Rimantas
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD107116705&prefLang=en_US