Search

Multipliers are demanded in a variety of applications. They consume higher power than other blocks. On the other hand, squarer units are less complex than a general-purpose multiplier, with wide use in cryptography, image processing, and video coding. This paper presents a new radix-2m squarer unit. It uses the essential sum trees to avoid redundant operations. Omitting some duplicate multiplications simplifies the final VLSI architecture. This work uses the radix-2m squarer units with m equal to 2 (radix-4), 3 (radix-8) and 4 (radix-16). We evaluate the proposed squarer units using high-efficiency video coding (HEVC) and cryptography algorithms for a range of bits. For performance results, the architectures use random input vectors. The results show that radix-4 and radix-8 units are more area and power efficient than radix-16 units. Moreover, the high energy savings of our proposed radix-4 and radix-8 units are preserved in the case studies addressed, compared to squarer units from the literature. Particularly, we show that radix-4 and radix-8 provide more energy savings in computing sum of squared differences (SSD). Furthermore, radix-4 and radix-8 squarer units are also efficient in the Montgomery-ladder architecture, reducing energy by 58.76% (radix-8) and 63.39% (radix-4) compared to the multiplier automatically selected by the logic synthesis tool. [ABSTRACT FROM AUTHOR]

Continuous and non-invasive cardiovascular monitoring has gained attention due to the miniaturization of wearable devices. Particularly, wrist-worn photoplethysmography (PPG) sensors present an alternative to electrocardiogram recording for heart rate (HR) monitoring as it is cheaper and non-intrusive for daily activities. Yet, the accuracy of PPG measurements is heavily affected by motion artifacts which are inherent to ambulatory environments. In this paper, we propose a low-complexity LSTM-only neural network for HR estimation from a single PPG channel during intense physical activity. This work explored the trade-off between model complexity and accuracy by exploring different model dataflows, number of layers, and number of training epochs to capture the intrinsic time-dependency between PPG samples. The best model achieves a mean absolute error of 4.47 ± 3.68 bpm when evaluated on 12 IEEE SPC subjects.Clinical relevance- This work aims to improve the quality of HR inference from PPG signals using neural network, enabling continuous vital signal monitoring with little interference in daily activities from embedded monitoring devices.