|Author's Email Address
||This thesis had been viewed 5559 times. Download 0 times.|
|Type of Document
||Detection and Deep-Learning Recognition of Hand Gestures Using Wi-Fi Signals|
|Date of Defense
injection-locked quadrature receiver
||This thesis presents a passive radar which is able to detect hand gesture by Doppler effect. The Wi-Fi signals are utilized as the signal sources in this radar for detecting a moving target. Since this is a passive radar, the architecture of the radar receiver is the main focus of the work. The radar receiver architecture is divided into two parts, an injection-locked oscillator (ILO) and a quadrature demodulator.|
Firstly, this thesis introduces the ILO design method. To fulfill the ILO with a high locking range and wide frequency tuning range, a ring oscillator based ILO was used and implemented using TSMC 0.18μm process. Next, the quadrature demodulator is implemented with hybrid components on a printed circuit board. Then this work combine the ILO and the quadrature demodulator to form the radar receiver, and compares the performance with the radar receiver used in the previous work. It is concluded that the presented radar receiver outperforms the previous one.
To recognize the gesture signals collected from the receiver and to raise the accuracy of recognition, this work uses the deep learning algorithm to train the gesture recognition process. The training procedure outputs different characteristic parameters of the signals, which is useful to recognize the gesture using the neural network. Moreover, different neural network structures were used to improve the accuracy of the gesture recognition.
||Chao-Hsiung Tseng - chair|
Fu-Kang Wang - co-chair
Huey-Ru Chuang - co-chair
Tzyy-Sheng Horng - advisor
Indicate in-campus at 3 year and off-campus access at 3 year.|
|Date of Submission