Towards Human Activity Recognition: A Hierarchical Feature Selection Framework

Abstract

The inherent complexity of human physical activities makes it difficult to accurately recognize activities with wearable sensors. To this end, this paper proposes a hierarchical activity recognition framework and two different feature selection methods to improve the recognition performance. Specifically, according to the characteristics of human activities, predefined activities of interest are organized into a hierarchical tree structure, where each internal node represents different groups of activities and each leaf node represents a specific activity label. Then, the proposed feature selection methods are appropriately integrated to optimize the feature space of each node. Finally, we train corresponding classifiers to distinguish different activity groups and to classify a new unseen sample into one of the leaf-nodes in a top-down fashion to predict its activity label. To evaluate the performance of the proposed framework and feature selection methods, we conduct extensive comparative experiments on publicly available datasets and analyze the model complexity. Experimental results show that the proposed method reduces the dimensionality of original feature space and contributes to enhancement of the overall recognition accuracy. In addition, for feature selection, returning multiple activity-specific feature subsets generally outperforms the case of returning a common subset of features for all activities.

Step by Step Towards Effective Human Activity Recognition: A Balance between Energy Consumption and Latency in Health and Wellbeing Applications

“…In feature selection a subset of features is chosen from the original feature set [83], and a new feature vector with fewer features is used for activity description [33,77,83,84]. This approach is used in wearable sensor systems with limited hardware resources for real time activity recognition [33].…”

“…Feature selection approaches improve the initial baseline efficiency of HAR [21]. In general, feature selection approaches are divided into filter-based, wrapper-based, and embedded approach-based methods [19,83,84]. Nevertheless, some authors also use terms such as basic features (statistics applied to raw sensor data) and graphical features (generated from graph representations) [50].…”

“…However, this type of classification is rarely used. When filter-based methods are applied, the feature selection process is dispart from the classification verification process, which makes them fast, but their drawback is that they require a threshold to stop the feature selection process [84]. Table 3 lists the authors who have reported and implemented different filter-based methods for feature selection.…”

Comparison of Different Sets of Features for Human Activity Recognition by Wearable Sensors

“…Accelerometric signals are common to all HAR applications. Some studies used this information alone [25,26,27], but more often accelerometers were combined with gyroscopes [12,28,29] and magnetometers [30,31]. In few cases other signals were taken into account such as quaternions [32], temperature [1], gravity [33] or data acquired from ambient sensors [1].…”

A Weighted Late Fusion Framework for Recognizing Human Activity from Wearable Sensors

“…However when used alone, they lack the ability to recognize similar activities [15], therefore combining them with other wearable or ambient sensors, improves the performance of a system. Gyroscopes measure the angular velocity of rotation [15] hence they detect the object's orientation [24]. They are also found in most activity recognition studies, however they are not so often used individually.…”