OKRELM: online kernelized and regularized extreme learning machine for wearable-based activity recognition

Human activity recognition (AR) techniques promote the development of large amounts of meaningful applications such as context awareness [1, 2], energy expenditure [3], disease detection [4] and personal healthcare [5]. Moreover, with the development of wearable techniques in recent years, diverse of sensors (accelerometer, gyroscope, et al.) are embedded into the mini-wearable devices (e.g. smartwatch [6], wristband [7], armband [8], head-belt [9, 10]). Consequently, wearable AR techniques are widely employed to improve the users’ health conditions by collecting and analyzing their data of activities of daily living (ADLs), and then giving them feedback. Wearable AR technologies grow extremely fast and a great deal of work has been proposed [11, 12]. Owing to the limited computation and storage resources of the wearable devices, a wearable AR model ought to be lightweight with reduced computation complexity [13]. In some real applications, real-time feedback is greatly important and necessary. For instance, a jogger might see how many steps he has made when he is running, then decides whether to continue running or not. The AR model inside a wearable device should fulfill the recognition task in real time. To do this, a great number of machine learning algorithms (Decision Tree [14], Support Vector Machine [15, 16], Extreme Learning Machine [17], Dynamic Bayesian Network [18], Hidden Markov Models [19], Boosting [20], etc.) and a Abstract Miscellaneous mini-wearable devices (Jawbone Up, Apple Watch, Google Glass, et al.) have emerged in recent years to recognize the user’s activities of daily living (ADLs) such as walking, running, climbing and bicycling. To better suits a target user, a generic activity recognition (AR) model inside the wearable devices requires to adapt itself according to the user’s personality in terms of wearing styles and so on. In this paper, an online kernelized and regularized extreme learning machine (OKRELM) is proposed for wearable-based activity recognition. A smallscale but important subset of every incoming data chunk is chosen to go through the update stage during the online sequential learning. Therefore, OKRELM is a lightweight incremental learning model with less time consumption during the update and prediction phase, a robust and effective classifier compared with the batch learning scheme. The performance of OKRELM is evaluated and compared with several related approaches on a UCI online available