Research And Design Of DM6437-based Artificial Vision System

Eye is the core organ for human to obtain visual information,but there are still some disabilities or even loss of vision caused by diseases.Medical studies showed that excepting the damaged photoreceptor cells,the other parts of visual pathways of some blind people are still intact.Therefor,a new visual remediation technology known as artificial vision is generated.In an artificial vision system,visual information is captured by using an extern camera and converted into electric signal.Then electrical signals wirelessly transmit to implant electrodes to activate electrical stimulation on the visual pathway,which causes phosphenes on visual cortex and restores the visual perception of the blind.By now,the clinical applications of artificial vision are reported aboard.In China,it's still in the laboratory stage.Therefore,the research and design on artificial vision system has a strong theoretical and practical significance.Based on the functional requirements of ophthalmic hospitals on artificial vision,this thesis designs a dm6437-based,low-cost,high-performance and integrated artificial vison system which named Vision Glass.Main work and achievements are presented as follows.(1)The scheme of the artifical vision system is designed.Then main software modules are implemented,such as video capture and display,debug and train after operation,system parameter maintenance,system software upgrade and so on.(2)To solve the conflict between the huge visual information and a limited array of implant electrode,a reducing pixelated image processing method based on block segmentation is presented.A 4×6 simulated phosphene map is generated from original 240×360 visual image by image processing,including perspective conversion,image resizing,gray-scale transformation,reducing pixilated etc.To facilitate the eye hospital to carry out visual training tasks for the blind,such as reading words,a single character recognition algorithm based on BP neural network is proposed and implemented.The experimental results show that the precision of character recognition has 96% probably.(3)After designing and analysising an OOK-based RF modulation circuit,an electrode stimulation intensity control strategy based on brightness grading is executed.To detect the work status of implant electrodes,a measurement method for impedance is put forward.Experimental results indicate that the measurement deviation is within 10%.(4)Considering features of C64 x DSP core,the system performance is optimized from various perspectives.To ensure the electrode stimulation accuracy and uniqueness before products manufactured,a specialized space response standard for artificial vision is constructed in this thesis.Moreover,the system power consumption and visual image processing performance is compared and analyzed in different working frequency.Test results show that the system could run continuously for two days after charged at the frequency of 432 MHz.Meanwhile,the speed of image processing after optimization is up to 30 fps.