Research On Hand Shape Recognition System Based On DSP

With the improvement of technology, information technology has been rapidly developed, more and more attention was paid to the information security. It is necessary to study a unique, safety, reliability, and convenient identification technology. The biological recognition technology has came into being in this condition and it has been widely used in the field of identification. As a kind of typical biometric identification technology, the hand shape recognition is easy to collect data, well receptive, high stability, and works simplely, so it has important research significance and practical value, which has been the hotspot. The hand shape recognition system based on PC has the shortcomings of high cost and large volume, so its application is limited. The study of hand shape recognition system based on DSP, can effectively reduce the cost and improve the work efficiency. It has good practicability, and is convenient to commercialise. The research in this field has just unfolding.Supported by the project “Hand and palm vein recognition system based on DSP” of Science and Technology Department of Jilin Province(Grant No. 20140204046), this paper has a deep research on the hand shape recognition system based on DSP, the system can achieve the image acquisition, feature extraction, and identification. The main research contents are as follows:(1) This research improved the hand features extraction method and geometry contour features extraction method. Except the thumb, the other four fingers were separatrd, curve fitting algorithm was used to locat axis of fingers. The instability contour of the fingers was cut off, and a step alignment method was used to standardize the fingers. Then, the geometric features were width of the fingers, profile features were coincidence rate, they were both extracted. This method can fully utilize the features of hand information, and improve the stability of hand positioning. The extracted hand shape features were more effective and reliable.(2) This research proposed a mix algorithm based on particle swarm algorithm and differential evolution algorithm to optimize the hand parameters. To achieve better matching results, we determine the optimal coefficient value based on the effectiveness characteristics of each finger. Experimental results show that hand parameter optimization effectively improved the recognition accuracy.(3) This research proposed a multi-feature fusion hand shape recognition method based on support vector machine(SVM), with the fusion of the contour features and geometric features of the finger on decision level. This method achieved the integration of hand features with good recognition performance. The proposed method had a better identifying performance than traditional method with single characteristic and threshold determination, and the recognition accuracy rate was up to 98.65%.(4) This research designed the hardware and software of hand shape recognition system based on DSP. We determined the system selection, and determined the hardware system of hand shape recognition with the core of TMS320DM642. The design of the system structure, image output/ input modules, and memory modules were described in detail.We also designed the software system framework, and developed the video driver based on DSP/BIOS. Finally, the proposed hand shape recognition algorithm was transplanted to the DSP system. The experimental results show that the experimental identification rate was up to 96%, which validated the practicality and reliability of the proposed algorithm.Theoretical and experimental studies have shown that the proposed hand shape recognition system based on DSP effectively used the information of hand features. The parameters optimization improved the recognition accuracy. The decision-making of feature fusion based on SVM has high classification accuracy, and the recognition performance is good.