TY - GEN
T1 - Accelerated SVM Algorithm for Sensors Fusion-based Activity Classification in Lightweighted Edge Devices
AU - Chang, Juneseo
AU - Kang, Myeongjin
AU - Park, Daejin
N1 - Publisher Copyright:
© 2022 IEEE.
PY - 2022
Y1 - 2022
N2 - Smart homes assist users by providing convenient services from human activity classification with the help of machine learning (ML) technology. However, most of the conventional high-performance ML algorithms require high computing power and memory usage. Therefore, they are inapplicable for resource-limited embedded systems such as smart homes. In this study, we propose a memory-efficient, high-speed ML algorithm for smart home activity data classification. We propose a method for comprehending activity data as image data, thereby using the MNIST dataset as a substitute for real-world activity data. The proposed ML algorithm consists of three parts: data preprocessing, training, and classification. In data preprocessing, training data of the same label are grouped into further detailed clusters. The training process generates hyperplanes by accumulating and thresholding each cluster of preprocessed data. Finally, the classification process is done by calculating the similarity between the input data and each hyperplane using the bitwise-operation-based error function. We verified our algorithm on 'Raspberry Pi 3' by loading trained hyperplanes and performing classification on 1, 000 training data. Compared to a linear support vector machine implemented from Tensorflow Lite, the proposed algorithm improved performance to 45%, memory usage to 15.41 %, and execution time per accuracy to 41.3%.
AB - Smart homes assist users by providing convenient services from human activity classification with the help of machine learning (ML) technology. However, most of the conventional high-performance ML algorithms require high computing power and memory usage. Therefore, they are inapplicable for resource-limited embedded systems such as smart homes. In this study, we propose a memory-efficient, high-speed ML algorithm for smart home activity data classification. We propose a method for comprehending activity data as image data, thereby using the MNIST dataset as a substitute for real-world activity data. The proposed ML algorithm consists of three parts: data preprocessing, training, and classification. In data preprocessing, training data of the same label are grouped into further detailed clusters. The training process generates hyperplanes by accumulating and thresholding each cluster of preprocessed data. Finally, the classification process is done by calculating the similarity between the input data and each hyperplane using the bitwise-operation-based error function. We verified our algorithm on 'Raspberry Pi 3' by loading trained hyperplanes and performing classification on 1, 000 training data. Compared to a linear support vector machine implemented from Tensorflow Lite, the proposed algorithm improved performance to 45%, memory usage to 15.41 %, and execution time per accuracy to 41.3%.
KW - Activity monitoring
KW - energy-accuracy trade-off
KW - machine learning
UR - http://www.scopus.com/inward/record.url?scp=85127032581&partnerID=8YFLogxK
U2 - 10.1109/ICCE53296.2022.9730557
DO - 10.1109/ICCE53296.2022.9730557
M3 - Conference contribution
AN - SCOPUS:85127032581
T3 - Digest of Technical Papers - IEEE International Conference on Consumer Electronics
BT - 2022 IEEE International Conference on Consumer Electronics, ICCE 2022
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2022 IEEE International Conference on Consumer Electronics, ICCE 2022
Y2 - 7 January 2022 through 9 January 2022
ER -