Gym Exercises Monitoring Based On Pressure Sensing Smart Gloves

Physical activity and fitness exercises have favorable effects on mental and physical health.This has resulted in the recent rise of novel technologies,such as wearable computing and sensor technology,which have been utilized to develop systems that monitor physical activity,assess performance outcomes,and help in decision-making to optimize performance efficiency.The computerized systems for monitoring strength training exercises in the gym are inferior to those for tracking cardiovascular activities like running and biking.It is noticed that fitness exercises,except for cardiovascular activities like running,involve single or multiple interactions between hand and gym equipment.By undertaking the literature review and consulting with fitness trainers,an alternative design-the pressure sensing smart glove with no wire connections at the sensing regions is proposed to track fitness activities whenever the user’s hand comes into contact with the workout environment or their body.The pressure-sensing smart glove can monitor and evaluate upper-body gym workouts by interacting with commonly available gym equipment.The prototype of the smart glove is developed by knitting a soft and stretchy pressure sensing fabric with a knitting machine.The smart glove has 93 sensing points on the wrist in a 3 × 15 matrix and on the palm in a 6 × 8 matrix,with no connecting cables in these places,making it less apparent during exercise.An experiment is conducted to collect data and analyze the performance of the pressure sensing smart glove.A total of 20 volunteers are selected to perform 10 different upper-body workouts.The instructional manual and demonstration videos are provided them with an understanding of how to perform the chosen exercises.The participants are asked to complete a questionnaire on their experience wearing the glove.Around 72.5%of participants prefer to wear gloves and generally provide positive feedback on their experience with the smart glove.The acquired raw data is preprocessed before extracting features and recognizing gym exercises.The objective of the preprocessing stage is to recover data obtained from abnormal channels and denoise the data.The retrieved data from the wrist and palm areas are then used to calculate 2,934 features,including time,frequency,and frame features.Duplicates,quasi-constant,and correlated features are removed from the retrieved features.Hence,a subset of 286 features is acquired and utilized to tune the parameters of four machine learning models using the grid searching technique.Considering both wrist and palm data,the outcomes are reviewed to determine the precise model.The findings indicate that the support vector machine model has the highest accuracy of 96%when utilizing stratified 10-fold cross validation.The development of an algorithm that uses a weight signal of an exercise to recognize repetitions of a workout is presented.There are three separate stages in the algorithm.1)Selection of true peaks by configuring height and distance criteria.2)The first dynamic time warping execution to identify an Explicit Pattern template for counting the repetitions.3)Applying the second dynamic time warping to examine and count repeats of a certain exercise.After evaluating the data signals of workouts obtained from the wrist zone,palm zone,and wrist and palm zones,the mean absolute error of 17.81(wrist and palm zones together)is calculated.The ratio index approach is offered as a means of determining the degree of imbalance in weight lifted.The positive and negative Γ±=±0.6 criteria are computed and utilized to determine the imbalance of weights lifted during a fitness activity.Based on the calculated findings,a total of 15 participants and 18 sets of various exercises are identified as having a weight imbalance.In summary,a pressure sensing smart glove system is proposed for recognizing fitness activities,counting repetitions,and quantifying imbalance.The smart glove can collect pressure-distributed data from 93 pressure sensing points located at the wrist and palm regions.The system’s performance is tested using exercise data acquired from 20 volunteers.The support vector classifier achieves an accuracy of 96%.For counting repetitions,a mean absolute error of 17.81 is computed.The imbalance of lifted weights is quantified using the ratio index and a calculated threshold of Γ±=±0.6.