To search, Click below search items. |
| All Published Papers Search Service |
|
|
|
Title |
Implementation of Deep Neural Nets on Microcontrollers for Speech Recognition |
|
Author |
Fazal Noor and Hatem M. El-Boghdadi |
| Citation |
Vol. 25 No. 4 pp. 65-74 |
|
Abstract |
Nature inspired optimization methods and Deep Learning Neural Networks have been proliferated in the literature. The optimization algorithms have been used in variety of problems such as controls, search, estimation, and many other areas whereas the Neural Networks have mainly been used in prediction and classification’s problems. In this work, Neural networks with heuristics optimization algorithms are employed to compute the optimum weights. In particular, a Parallel Distributed Bat Algorithm (PDBA) is used to obtain optimum weights of Deep Learning Neural Networks. The algorithms are implemented on an Arduino, an open source microcontroller and an application in control systems is studied using speech recognition. As the neural network process is compute intensive, multiple microcontrollers are proposed in a master-slave configuration. Speech is the fastest way for communications between humans, and extensive research has been done in speech recognition between humans and machines. In this research, Linear Predictive Coding (LPC) is used for extracting speech features and Deep Learning Neural Networks are trained with speech feature samples in a voice-controlled application. |
|
Keywords |
Robotics, Speech recognition, Bat Optimization Algorithm, Neural Networks,. |
|
URL |
