Description
Convergence of Machine Learning, Internet of Things and computationally powerful single-board computers has boosted research and implementation of smart spaces. Smart spaces make predictions based on historical data and have become an integral part of our daily routines to improve quality of life for many different groups of people. The use of embedded systems to build these smart spaces, in combination with data analytic, can provide real-time information about the environment and how it interacts with the people in it.In this thesis, we demonstrate how one embedded system that acquires data based on a 2-dimensional positional-grid, movement, temperature and vibration is used to build a smart and pervasive space. Data collected from these sensors is used for real time localization in conjunction with machine learning mechanisms to analyze human activities. Average localization accuracy achieved is 30cm. Low-cost, low-energy, Low-resolution (4x16) and non-intrusive thermal sensor is used to detect static and dynamic human activities . We train six machine learning algorithms, namely Logistic Regression, Naive Bayes, Support Vector Machine, Decision Tree, Random Forest and Artificial Neural Network (Vanilla Feed Forward) on the dataset collected in our lab. Our experiment revealed very high static activity detection rate with all algorithms with Feed Forward Neural Network giving best accuracy of 99.96% . We also show how data collection methods and sensor placement plays an important role deciding machine learning algorithms accuracy. We used method of connected components analysis to detect forward-backward movements with 87.59% accuracy and detection of left-right moment is carried out using cross-correlation method with 100% accuracy. Our implementation does not use any camera or microphone deployment, hence addressing potential privacy issues. This smart pace implementation find it’s application in wide domain from conference, study rooms to elderly care due to it’s real-time nature.
