Improving Performance of the Fitness Exercises Repetitions Counter via Computational Complexity Reduction

Abstract

: The COVID-19 precautions had forced us to look for different techniques that enable the continuity of our ordinary life activities, especially the sports ones. In addition, the need for accurate and fast auto judgment techniques to measure physical fitness performance is constantly emerging. The artificial intelligence (AI) with multi-resolution counter had introduced method relays on Artificial Intelligence to realize this purpose, but this method has a high processing time. Modifying the algorithm structure and the inputs features leads to low computational cost. This paper presents a modified algorithm that reduces the computational costs for the optical flow equation This reduction is executed via two techniques; the first one is to execute Gunner Franeback algorithm for number of pixels less than had been used in the previous model via selecting the more weighted pixels that closer to the central pixel, the second one is to employ Model Quantization technique by using Tensor flow Lite as a proposed model. Experimental results indicate that the proposed method has low computational cost, reliable and robust, and can be applied as practical applications. The performance of the experiments was verified by comparing its time complexity with the AI with multi-resolution counter depending on ground truth data. complexity.