This paper proposes a novel data compression method for artificial vision systems and its low-energy implementation in order to reduce energy consumption in a wireless communication subsystem. The artificial vision systems are one of the methods for realizing visual prosthesis by controlling stimulus to visual nerves, and they consist of an inner stimulating unit and an outer image processing unit. The outer unit transmits information regarding stimulation to the inner unit via wireless communication, which occupies a large portion of the whole energy consumption. Reducing traffic in wireless communication is important to prevent damage caused by extra heat dissipation of the inner unit, which leads to excess energy consumption. The proposed compression method marks a higher compression ratio than the conventional compression methods by taking advantage of the analyses of stimuli position data, which is dominant in traffic. The proposed method is implemented as an application-domain specific instruction-set processor to achieve both configurability of stimulation control and compression efficiency. The evaluation results show that the proposed implementation reduces energy consumption by about 87% and 62% in the compression and decompression process, respectively. These results indicate that the proposed method can expect to reduce energy consumption in a wireless communication receiver dramatically.
2017 by the Information Processing Society of Japan