InnoHAR: A Deep Neural Network for Complex Human Activity Recognition

Abstract

Human activity recognition (HAR) based on sensor networks is an important research direction in the fields of pervasive computing and body area network. Existing researches often use statistical machine learning methods to manually extract and construct features of different motions. However, in the face of extremely fast-growing waveform data with no obvious laws, the traditional feature engineering methods are becoming more and more incapable. With the development of deep learning technology, we do not need to manually extract features and can improve the performance in complex human activity recognition problems. By migrating deep neural network experience in image recognition, we propose a deep learning model (InnoHAR) based on the combination of inception neural network and recurrent neural network. The model inputs the waveform data of multi-channel sensors end-to-end. Multi-dimensional features are extracted by inception-like modules by using various kernel-based convolution layers. Combined with GRU, modeling for time series features is realized, making full use of data characteristics to complete classification tasks. Through experimental verification on three most widely used public HAR datasets, our proposed method shows consistent superior performance and has good generalization performance, when compared with the state-of-the-art.INDEX TERMS Complex human activity, inception neural network, wearable sensor, computational efficiency.

Improving Cross-Subject Activity Recognition via Adversarial Learning

“…Taking a CNN-LSTM baseline as the classifier and PAMAP2 as the dataset, our solution provides a gain of nearly 10% in cross-subject performance (in terms of mean F1-score) compared to the sole use of the CNN-LSTM baseline. Applied to the state-of-the-art Inno-HAR [36] classifier, the leap in performance reaches almost 5%, also for the PAMAP2 dataset. These improvements correspond to a decreased need for variability of subject behavior in the training set, which can be translated into fewer subjects with which to collect and label data.…”

MCC-CKF: A Distance Constrained Kalman Filter Method for Indoor TOA Localization Applications

“…The experiment scene was shown in Figure 4, which is located in the hall (10 m × 17 m × 4 m) inside the building of School of Computer and Communication Engineering, University of Science and Technology Beijing (USTB). Four reference nodes are respectively located as (1, 1), (9, 1), (1,16) and (9,16). The experimenter is asked to periodically walk along a rectangular trajectory within the region.…”

“…The TOA ranging accuracy can be improved by smoothing TOA ranging results to mitigate the impact of the NLOS propagation [14]. However, in practice, TOA ranging is more affected by non-Gaussian noises rather than Gaussian ones, whose errors could achieve as high as 10 m [15,16]. If the system is disturbed by non-Gaussian noise, such as this mismatch, the performance of Kalman filter will be degraded, which has no ability to limit the large ranging error.…”

“…Besides, it is generally considered that KF [17] take advantage of minimum mean square error (MMSE) as the optimal criterion in linear systems with white Gaussian noise. The standard is so sensitive to large outliers and non-Gaussian noise that Kalman filters may not show good performance when the systems are mixed with colored noise [16,18] appearing in many actual application scenarios. In order to effectively work out the estimation problem when the system has colored Gaussian noise, several optimization criteria have been proposed to replace the MMSE.…”

“…To enhance the performance of the Kalman filter for TOA ranging and localization applications that have colored Gaussian noise [16,18], maximum correntropy criterion Kalman filtering (MCC-KF) has used the MCC instead of the MMSE as the optimization criterion for LOS/NLOS TOA ranging. The MCC-KF can get much better performance than traditional KF in many cases.…”