ABSTRACT

The Speech has been potential mode of the interaction with computers now a days. The recognition of the speech is very much difficult for the computers due to a complexity of the human language. Therefore, in this research study we have developed a platform and speaker independent Automatic Speech Recognition (ASR) system for English and Sindhi language. The system gets input from the user in the form of voice via mi r phone and according to the voice command, action will be performed. The manipulator can open, close and switch between the applications. The open command i u ed for opening an application; close command is used for closing an application and S\ itch command is used for switching from one application to another application. The user spoke a switch command directly by microphone. A determination is then made whether the application to be switched to being run. If the application is running, the focus of the operating system is switched to the second application. If a second application is not running, then the application is launched. We ha e achieved 75% accuracy for Sindhi language and 85% accuracy for English language. The proposed system has been developed using Java, which provides platform independent features and an open-source speech recognition framework called CMU (Carnegie Mellon University) Sphinx4; it is a flexible, open source, the modular and pluggable framework which uses the statistical based approach (Hidden Markov Model) for speech recognition. The proposed system uses the model-based approach. Therefore, in the future, different language models can also be embedded in the system.