Iss2Image: A Novel Signal-Encoding Technique for CNN-Based Human Activity Recognition

• DOI: 10.3390/s18113910
• Publication Date: 2018-11-14T15:58:22Z
• Authors: Taeho Hur; Jaehun Bang; Thien Huynh-The; Jongwon Lee; Jee-In Kim; Sungyoung Lee
• Keywords: Activity Recognition; SIGNAL (programming language); Distortion (music)

The most significant barrier to success in human activity recognition is extracting and selecting the right features. In traditional methods, features are chosen by humans, which requires user have expert knowledge or do a large amount of empirical study. Newly developed deep learning technology can automatically extract select Among various convolutional neural networks (CNNs) advantages local dependency scale invariance suitable for temporal data such as accelerometer (ACC) signals. this paper, we propose an efficient method, namely Iss2Image (Inertial sensor signal Image), novel encoding technique transforming inertial into image with minimum distortion CNN model image-based classification. converts real number values from X, Y, Z axes three color channels precisely infer correlations among successive different dimensions. We experimentally evaluated our method using several well-known datasets own dataset collected smartphone smartwatch. proposed shows higher accuracy than other state-of-the-art approaches on tested datasets.