• JSTS:Journal of Semiconductor Technology and Science
• /
• 제17권3호
• /
• pp.465-472
• /
• 2017

Security has become one of the most important requirements for various devices for multi-sensor based embedded systems. The AES (Advanced Encryption Standard) algorithm is widely used for security, however, it requires high computing power. In order to reduce the CPU power for the data encryption of images, we propose a new image encryption module using hardware memoization, which can reuse previously generated data. However, as image pixel data are slightly different each other, the reuse rate of the simple memoization system is low. Therefore, we further apply an approximate concept to the memoization system to have a higher reuse rate by sacrificing quality. With the novel technique, the throughput can be highly improved by 23.98% with 14.88% energy savings with image quality loss minimization.

• Journal of Information Processing Systems
• /
• 제18권1호
• /
• pp.97-114
• /
• 2022

Interest is increasing in electrocardiogram (ECG) signal analysis for embedded devices, creating the need to develop an algorithm suitable for a low-power, low-memory embedded device. Linear approximation of the ECG signal facilitates the detection of fiducial points by expressing the signal as a small number of vertices. However, dynamic programming, a global optimization method used for linear approximation, has the disadvantage of high complexity using memoization. In this paper, the calculation area and memory usage are improved using a linear approximated template. The proposed algorithm reduces the calculation area required for dynamic programming through local optimization around the vertices of the template. In addition, it minimizes the storage space required by expressing the time information using the error from the vertices of the template, which is more compact than the time difference between vertices. When the length of the signal is L, the number of vertices is N, and the margin tolerance is M, the spatial complexity improves from O(NL) to O(NM). In our experiment, the linear approximation processing time was 12.45 times faster, from 18.18 ms to 1.46 ms on average, for each beat. The quality distribution of the percentage root mean square difference confirms that the proposed algorithm is a stable approximation.