Paper Info
Title
Lightweight Machine Learning-Based Approach for Supervision of Fitness Workout
Abstract
It is widely known that physical activity helps preventing several diseases. However, unsupervised training often results in low exercise quality, ineffective training, and, in worst cases, injuries. Automatic tracking and quantification of exercises by means of wearable devices could be an effective mean for the monitoring of exercise correctness. As a consequence, such devices could help motivating people, thus improving the quantity of performed physical exercise, with positive effects on users’ health conditions. However, despite the availability of several commercial devices, the performance and effectiveness are not well documented. This work proposes a new solution for fitness workout supervision exploiting machine learning techniques, in particular Linear Discriminant Analysis for analyzing data coming from wearable Inertial Measurement Units. Efforts have been done in order to reduce the computational requirements, thus assuring compatibility in perspective of embedded implementation. The experimental tests carried out to assess the proposed approach performance showed an accuracy in exercise detection over 93% and error in exercise counting less than 6%.
Year
DOI
Venue
2019
10.1109/SAS.2019.8706106
2019 IEEE Sensors Applications Symposium (SAS)
Keywords
Field
DocType
Performance evaluation,Training,Accelerometers,Band-pass filters,Principal component analysis,Feature extraction,Classification algorithms
Units of measurement,Accelerometer,Computer science,Wearable computer,Correctness,Feature extraction,Artificial intelligence,Linear discriminant analysis,Statistical classification,Wearable technology,Machine learning
Conference
ISBN
Citations 
PageRank 
978-1-5386-7713-1
0
0.34
References 
Authors
0
5
Name
Order
Citations
PageRank
Alessandro Depari112127.34
Paolo Ferrari239259.01
Alessandra Flammini349287.79
Stefano Rinaldi419031.39
Emiliano Sisinni545756.63