Energy-efficient adaptive optical character recognition for wearable devices

초록

As the computing power of mobile/wearable devices increases, computation-intensive real-time optical character recognition (OCR) becomes feasible for high-resolution images. Developing mobile/wearable OCR applications is challenging because they should perform highly accurate OCR within users' tolerable waiting time and achieve low energy consumption. An adaptive power management scheme is presented that predicts the execution time for OCR and minimises its energy consumption via dynamic voltage and frequency scaling while meeting its time constraint. Tesseract, a popular open source OCR engine, is used to verify the proposed scheme. The experimental results show up to 48.25% reduction in energy consumption (with an average reduction of 34.53%).