We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
Intelligent Embedded Advanced Machine Learning Approach for Smartphone Movement Identification using Stacked Auto-Encoders.
- Authors
Kumar, A. Durgapraveen; Terlapu, Panduranga Vital; G., JagadeeswaraRao; Jayaram, D.; Chand, K. Poorna; Rambabu, P.
- Abstract
Nowadays, the embedded system technology is emerging and being used in every area, especially for monitoring applications using the cloud and IoT. The subfield of machine learning (ML) that deals with embedded systems are denoted as embedded ML (EML) or Tiny ML. Using and deploying ML on devices with embedded systems has significant advantages. In this research, a smartphone was connected to a computer device through embedded technology for gathering information about the movements of the cell, such as clockwise and counterclockwise rotations, up and down, left and right, idle movements, and wave or snake motions. The data of the motion of the cell was in a three-dimensional array (X, Y, and Z axes) over time. Every data sample was loaded within 10 seconds, and the data size was 3X625. We split the data into 80:20 for training and testing and applied stacked auto-encoders for feature extraction. In the first step, we got the output features1 from auto-encoder 1, which was the input of the second auto-encoder. The softmax classifier classified the features2 generated by autoencoder2 and got more extreme results than other applied models like MLP and SVM. The model stacked auto-encoders performed highly with 0.998 accuracies and a value of 1 for AUC for the testing dataset, without any over fitting problem. This analysis is very useful for tracking smartphone movements easily. As per the motion of the cell, we can analyze the carrying smartphone human motions using apps and tools.
- Subjects
MACHINE learning; SMARTPHONES; FEATURE extraction; CELL motility; CLOUD computing; MOTION capture (Human mechanics)
- Publication
Grenze International Journal of Engineering & Technology (GIJET), 2023, Vol 9, Issue 2, p9
- ISSN
2395-5287
- Publication type
Article