NORMA eResearch @NCI Library

Human Activity Recognition using Deep Learning Approach

Rehman, Laiba (2022) Human Activity Recognition using Deep Learning Approach. Masters thesis, Dublin, National College of Ireland.

[thumbnail of Master of Science]
PDF (Master of Science)
Download (1MB) | Preview
[thumbnail of Configuration manual]
PDF (Configuration manual)
Download (1MB) | Preview


This research is performed to detect human activities from the triaxial body acceleration and angular velocity recorded by accelerometer and gyroscope sensors from 30 volunteers. Throughout the research, we have performed excessive data exploration, data preprocessing, visualization and build deep learning models where we performed single and multi-layer LSTM model, CNN model, and divide-and-conquer based CNN model. t-distributed Stochastic Neighbour Embedding (t-SNE) is performed to separate the data belonging to 6 activities (walking, walking upstairs, walking downstairs, sitting, laying and standing) using different perplexity values. On the test data, we have applied t-SNE of 20, 50 and 90 perplexity values and observed that the laying activity is distinct from all other activities whereas standing and sitting activities overlap each other. Throughout the research, we also performed hyperparameter tuning where we picked the best parameters from all the models to give the best accuracy. The combined divide and conquer-based model where data sharpening is applied to combine the static and dynamic activities and fed into a final pipeline with the different number of epochs give a training accuracy of 98.3% and test accuracy of 96.8%. These indicate that the combined approach can be used in a better way compared to other deep learning methods. Also, the higher accuracy indicates that this can be also improved in future research where we can see that Static and dynamic activities are highly predicted with the help of divide and conquer-based approaches.

Item Type: Thesis (Masters)
Uncontrolled Keywords: LSTM; CNN; Divide and conquer; hyper parameter
Subjects: Q Science > QA Mathematics > Electronic computers. Computer science
T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science
Q Science > QP Physiology
Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions: School of Computing > Master of Science in Data Analytics
Depositing User: Tamara Malone
Date Deposited: 08 Mar 2023 17:51
Last Modified: 08 Mar 2023 17:51

Actions (login required)

View Item View Item