Speech recognition system (SR) defines a system that converts a speech sample to text with the goal to be as accurate as a human listener. It has been used in different areas like robotics, health-care, and automotive.
As speech recognition is becoming more accurate in understanding what users say, more developers integrate it in their application. The goal is to make the SP system as accurate as possible for the user. If the trained network continually misinterprets what was said, it can drive the customer away. The developers have a choice between offline and cloud-based implementations of speech recognition systems. When the privacy matters, or we work on a very specific domain, then the offline speech recognition systems are favored. However, the accuracy of this SR system is not adequate in compare with cloud-based solutions.
In this thesis, first the training and recognizing in a SR system is observed. Then an open source SP system is chosen and reviewed in detail. The accuracy of the system is observed and steps for improvement of the accuracy are implemented. Finally, there is a comparison between the accuracy of both open source and cloud-based speech recognition systems.
«
Speech recognition system (SR) defines a system that converts a speech sample to text with the goal to be as accurate as a human listener. It has been used in different areas like robotics, health-care, and automotive.
As speech recognition is becoming more accurate in understanding what users say, more developers integrate it in their application. The goal is to make the SP system as accurate as possible for the user. If the trained network continually misinterprets what was said, it can drive...
»