A unified generative model using generative adversarial network for activity recognition

  • PDF / 2,363,358 Bytes
  • 10 Pages / 595.276 x 790.866 pts Page_size
  • 10 Downloads / 217 Views

DOWNLOAD

REPORT


ORIGINAL RESEARCH

A unified generative model using generative adversarial network for activity recognition Mang Hong Chan1 · Mohd Halim Mohd Noor1  Received: 21 April 2020 / Accepted: 11 September 2020 © Springer-Verlag GmbH Germany, part of Springer Nature 2020

Abstract The recent advancement of deep learning methods has seen a significant increase in recognition accuracy in many important applications such as human activity recognition. However, deep learning methods require a vast amount of sensor data to automatically extract the most salient features for activity classification. Therefore, in this paper, a unified generative model is proposed to generate verisimilar data of different activities for activity recognition. The proposed generative model not only able to generate data that have a similar pattern, but also data with diverse characteristics. This allows for data augmentation in activity classification to improve the overall recognition accuracy. Three similarity measures are proposed to assess the quality of the synthetic data in addition to two visual evaluation methods. The proposed generative model was evaluated on a public dataset. The training data was prepared by systematically varying the combination of original and synthetic data. Results have shown that classification using the hybrid training data achieved a comparable recognition accuracy with the classification using the original training data. The performance of the classifiers maintained at the recognition accuracy of 85%. Keywords  Activity recognition · Data generation · Generative adversarial network · Data augmentation

1 Introduction The rapid growth and development of machine learning techniques and ubiquitous computing have led computer scientists to analyses and interpret sensor data with the main purpose of extracting knowledge from the pervasive sensor. Human Activity Recognition (HAR) has gain attention from a lot of researchers for it has extreme practicality in health monitoring, medical support, entertainment and personal health tracking service (Qi et al. 2018; Prati et al. 2019). Most of the research aims to improve and enhance the speed, efficiency and accuracy of the algorithm in pattern recognition by extracting knowledge from the raw data in order to provide useful information under realistic conditions. HAR is one of the many fields that use machine learning techniques to learn from sensor data to recognize the * Mohd Halim Mohd Noor [email protected] Mang Hong Chan [email protected] 1



School of Computer Sciences, Universiti Sains Malaysia, Pulau Pinang, Malaysia

latent patterns in the datasets for different types of human activities. One of the very important factors of high accuracy of deep learning is the sufficient amount of training data. A robust and reliable model needs a vast amount of data to precisely capture the underlying pattern of the data. In fact, getting a good amount of data for the training of the model is an elementary need for deep learning (Ramasamy and Roy 2018) as it ca