Adopting deep learning methods for human activity recognition has been effective in extracting discriminative features from raw input sequences acquired from body-worn sensors. Although human movements are encoded in a sequence of successive samples in time, typical machine learning methods perform recognition tasks without exploiting the temporal correlations between input data samples. Convolutional neural networks (CNNs) address this issue by using convolutions across a one-dimensional temporal sequence to capture dependencies among input data. However, the size of convolutional kernels restricts the captured range of dependencies between data samples. As a result, typical models are unadaptable to a wide range of activity-recognition configurations and require fixed-length input windows. In this paper, we propose the use of deep recurrent neural networks (DRNNs) for building recognition models that are capable of capturing long-range dependencies in variable-length input sequences. We present unidirectional, bidirectional, and cascaded architectures based on long short-term memory (LSTM) DRNNs and evaluate their effectiveness on miscellaneous benchmark datasets. Experimental results show that our proposed models outperform methods employing conventional machine learning, such as support vector machine (SVM) and k-nearest neighbors (KNN). Additionally, the proposed models yield better performance than other deep learning techniques, such as deep believe networks (DBNs) and CNNs.

It is known that an ultrawideband (UWB)-based indoor positioning system (IPS) has superior positioning performance and can meet the requirements of location-based services (LBSs) as the Internet of Things (IoT) applications. However, there is a limitation of UWB positioning when it is conducted at the nonline-of-sight (NLOS) channels degrading the UWB ranging accuracy at indoor environments. In this article, we propose an artificial intelligence (AI) applied UWB positioning system that can enhance the positioning performance by classifying channel conditions with channel impulse response (CIR) of the received UWB signal. The proposed system mitigates the positioning degradation caused by the NLOS situations by performing extended Kalman filter (EKF) localization and long short-term memory (LSTM) training of the observed channel status. The main feature of the proposed UWB positioning method is that it can be used even at unknown locations not trained with the LSTM model learning the channel status, because of our training strategy of not the position coordinates, but the UWB ranging error between UWB devices corresponding to CIR of the received UWB signal. This article provides the experimental setup and performance evaluation results of the proposed system. The evaluation results showed that the proposed AI-applied UWB positioning method significantly improved its accuracy performance compared with the existing positioning methods.

In recent times, social and commercial interests in location-based services (LBS) are significantly increasing due to the rise in smart devices and technologies. The global navigation satellite systems (GNSS) have long been employed for LBS to navigate and determine accurate and reliable location information in outdoor environments. However, the GNSS signals are too weak to penetrate buildings and unable to provide reliable indoor LBS. Hence, GNSS’s incompetence in the indoor environment invites extensive research and development of an indoor positioning system (IPS). Various technologies and techniques have been studied for IPS development. This paper provides an overview of the available smartphone-based indoor localization solutions that rely on radio frequency technologies. As fingerprinting localization is mostly accepted for IPS development owing to its good localization accuracy, we discuss fingerprinting localization in detail. In particular, our analysis is more focused on practical IPS that are realized using a smartphone and Wi-Fi/Bluetooth Low Energy (BLE) as a signal source. Furthermore, we elaborate on the challenges of practical IPS, the available solutions and comprehensive performance comparison, and present some future trends in IPS development.

Securing personal authentication is an important study in the field of security. Particularly, fingerprinting and face recognition have been used for personal authentication. However, these systems suffer from certain issues, such as fingerprinting forgery, or environmental obstacles. To address forgery or spoofing identification problems, various approaches have been considered, including electrocardiogram (ECG). For ECG identification, linear discriminant analysis (LDA), support vector machine (SVM), principal component analysis (PCA), deep recurrent neural network (DRNN), and recurrent neural network (RNN) have been conventionally used. Certain studies have shown that the RNN model yields the best performance in ECG identification as compared with the other models. However, these methods require a lengthy input signal for high accuracy. Thus, these methods may not be applied to a real-time system. In this study, we propose using bidirectional long short-term memory (LSTM)-based deep recurrent neural networks (DRNN) through late-fusion to develop a real-time system for ECG-based biometrics identification and classification. We suggest a preprocessing procedure for the quick identification and noise reduction, such as a derivative filter, moving average filter, and normalization. We experimentally evaluated the proposed method using two public datasets: MIT-BIH Normal Sinus Rhythm (NSRDB) and MIT-BIH Arrhythmia (MITDB). The proposed LSTM-based DRNN model shows that in NSRDB, the overall precision was 100%, recall was 100%, accuracy was 100%, and F1-score was 1. For MITDB, the overall precision was 99.8%, recall was 99.8%, accuracy was 99.8%, and F1-score was 0.99. Our experiments demonstrate that the proposed model achieves an overall higher classification accuracy and efficiency compared with the conventional LSTM approach.

In indoor environments, the Global Positioning System (GPS) and long-range tracking radar systems are not optimal, because of signal propagation limitations in the indoor environment. In recent years, the use of ultra-wide band (UWB) technology has become a possible solution for object detection, localization and tracking in indoor environments, because of its high range resolution, compact size and low cost. This paper presents improved target detection and tracking techniques for moving objects with impulse-radio UWB (IR-UWB) radar in a short-range indoor area. This is achieved through signal-processing steps, such as clutter reduction, target detection, target localization and tracking. In this paper, we introduce a new combination consisting of our proposed signal-processing procedures. In the clutter-reduction step, a filtering method that uses a Kalman filter (KF) is proposed. Then, in the target detection step, a modification of the conventional CLEAN algorithm which is used to estimate the impulse response from observation region is applied for the advanced elimination of false alarms. Then, the output is fed into the target localization and tracking step, in which the target location and trajectory are determined and tracked by using unscented KF in two-dimensional coordinates. In each step, the proposed methods are compared to conventional methods to demonstrate the differences in performance. The experiments are carried out using actual IR-UWB radar under different scenarios. The results verify that the proposed methods can improve the probability and efficiency of target detection and tracking.

Recent developments in the fields of smartphones and wireless communication technologies such as beacons, Wi-Fi, and ultra-wideband have made it possible to realize indoor positioning system (IPS) with a few meters of accuracy. In this paper, an improvement over traditional fingerprinting localization is proposed by combining it with weighted centroid localization (WCL). The proposed localization method reduces the total number of fingerprint reference points over the localization space, thus minimizing both the time required for reading radio frequency signals and the number of reference points needed during the fingerprinting learning process, which eventually makes the process less time-consuming. The proposed positioning has two major steps of operation. In the first step, we have realized fingerprinting that utilizes lightly populated reference points (RPs) and WCL individually. Using the location estimated at the first step, WCL is run again for the final location estimation. The proposed localization technique reduces the number of required fingerprint RPs by more than 40% compared to normal fingerprinting localization method with a similar localization estimation error.