Human monitoring applications in indoor environments depend on accurate human identification and activity recognition (HIAR). Single modality sensor systems have shown to be accurate for HIAR, but there are some shortcomings to these systems, such as privacy, intrusion, and costs. To combat these shortcomings for a long-term monitoring solution, an interpretable, passive, multi-modal, sensor fusion system PRF-PIR is proposed in this work. PRF-PIR is composed of one software-defined radio (SDR) device and one novel passive infrared (PIR) sensor system. A recurrent neural network (RNN) is built as the HIAR model for this proposed solution to handle the temporal dependence of passive information captured by both modalities. We validate our proposed PRF-PIR system for a potential human monitoring system through the data collection of eleven activities from twelve human subjects in an academic office environment. From our data collection, the efficacy of the sensor fusion system is proven via an accuracy of 0.9866 for human identification and an accuracy of 0.9623 for activity recognition. The results of the system are supported with explainable artificial intelligence (XAI) methodologies to serve as a validation for sensor fusion over the deployment of single sensor solutions. PRF-PIR provides a passive, non-intrusive, and highly accurate system that allows for robustness in uncertain, highly similar, and complex at-home activities performed by a variety of human subjects.

Passive Infrared (PIR) sensors are commonly used in indoor applications to detect human presence. PIR sensors detect human presence by detecting the change in infrared radiation across the polarity of the sensor. Due to this, PIR sensors are unable to accurately detect stationary human subjects, which results in false negatives. In the pursuit of creating a low-cost solution for detecting stationary occupants in a closed space, the novel approach to mount a PIR sensor on a moving platform was developed (MI-PIR). This approach was developed for the system to artificially induce the motion that is necessary for stationary human detection. Utilizing the raw analog output of the PIR sensor and an artificial neural network (ANN), the closed space was accurately classified for room occupancy, the number of occupants, the approximate location of the human targets, and the differentiation of targets. This novel approach provides the advantages of a utilizing a single PIR sensor for human presence detection, while eliminating the major known drawback to this type of sensor. Scanning the room using a PIR sensor also allows for an expanded field of view (FoV) and a simpler deployment, in comparison to other approaches using a PIR sensor. Finally, MI-PIR expands the functionalities of PIR sensors by using an ANN to detect various other occupancy parameters. The experimental results show that the system can detect room classification with 99% accuracy, 91% accuracy in occupancy count estimation, 93% accuracy in relative location prediction, and 93% accuracy in human target differentiation. These results show promise for an application of tracking and monitoring an at-risk patient in an indoor setting.

Deep learning has shown to be capable of learning features from complex datasets such as in the case of biometric authentication. Biometric authentication relies on unique biological qualities to verify a user's identity against a group of potential adversaries for security purposes. In this paper, a deep learning dependent biometric authentication system based on chest motion data captured by a passive infrared (PIR) sensor, CM-PIR, is proposed. PIR sensors are cheap, commercial-off-the-shelf (COTS) components that are dependent on motion across its field of view (FoV) for accurate detection of human subjects. Thus, CM-PIR utilizes deep learning for the accurate detection of stationary human subjects, as well as for biometric authentication. CM-PIR collects chest motion data from sixteen human subjects in nine different home locations. Coefficients of Fourier transform (FFT), discrete wavelet transform (DWT), and the absolute value of an acceleration filter calculated from the raw voltage PIR values were selected as the optimal features for input to the deep learning models for biometric authentication. Preprocessing of these features included a threshold voltage range, normalization, and finally principal component analysis (PCA). CM-PIR is 94% accurate in stationary human presence detection and 75% accurate for biometric authentication using a RNN with a 90 second window size. This work provides a promising solution for stationary human detection and biometric authentication using a PIR sensor in a real-world setting.