Introduction on Sensor-Based Human Activity Analysis: Background

The constant growth of sensor-based systems and technologies for the detection of human activities has made notable progress in the field of human-computer interaction. The continuation of Internet connectivity into daily objects and physical devices has made it possible for the researchers to use IoT sensors for healthcare, elderly people monitoring, fitness tracking, working activity monitoring, and so on. The prominent application fields of sensor-based activity monitoring systems are many, but not limited to, pattern recognition, machine learning, context awareness, and human-centric sensing. If a salient investigation is performed on this topic by fellow researchers, this can create a vital turn in the way of interaction among people and mobile devices. In this book, we have bestowed a comprehensive survey showing the various aspects of human activity recognition based on wearable, environmental, and smartphone sensors. After discussing the background, numerous factors have been analyzed for the data pre-processing part regarding noise filtering and segmentation methods. The list of sensing devices, sensors, and application tools listed in this book can be used for the activity data collection efficiently. Moreover, a detailed analysis of more than 150 benchmark datasets and dataset repositories in this book includes information about sensors, attributes, activity classes, etc. These datasets sum up several types of sensor-based daily activities, medical activities, fitness activities, transportation activities, device usage, fall detection, and hand gesture data. In addition to these, we have shown the feature extraction and classical machine learning methods in detail. Moreover, the overview of different types of classification problems has been given along with the discussion on several performance evaluation techniques showing their advantages and limitations. Furthermore, we have also discussed the importance of deep learning methods to solve the problem of shallow learning using hand-crafted features in conventional pattern recognition approaches. Finally, we have presented a summary of activity recognition methods focused on recent works in several benchmark datasets, and mentioned some future challenges regarding data collection, design issues, and other prospects.