Indoor human activity recognition using high-dimensional sensors and deep neural networks

Abstract

Many smart home applications rely on indoor human activity recognition. This challenge is currently primarily tackled by employing video camera sensors. However, the use of such sensors is characterized by fundamental technical deficiencies in an indoor environment, often also resulting in a breach of privacy. In contrast, a radar sensor resolves most of these flaws and maintains privacy in particular. In this paper, we investigate a novel approach towards automatic indoor human activity recognition, feeding high-dimensional radar and video camera sensor data into several deep neural networks. Furthermore, we explore the efficacy of sensor fusion to provide a solution in less than ideal circumstances. We validate our approach on two newly constructed and published data sets that consist of 2347 and 1505 samples distributed over six different types of gestures and events, respectively. From our analysis, we can conclude that, when considering a radar sensor, it is optimal to make use of a three-dimensional convolutional neural network that takes as input sequential range-Doppler maps. This model achieves 12.22% and 2.97% error rate on the gestures and the events data set, respectively. A pre-trained residual network is employed to deal with the video camera sensor data and obtains 1.67% and 3.00% error rate on the same data sets. We show that there exists a clear benefit in com-Baptist Vandersmissen

Smart Living: An Interactive Control System for Household Appliances

“…The average operation time for both appliances is similar, around 30 seconds, even though the evaluation scenario for controlling the TV has two more steps than for the fan. Compared with recent work which usually examines only parts of the overall process, especially in terms of concept, interface, cost, image capture, and image process [20]- [26], the interactive control system is innovative since there are 20 equations in the study that combines the concepts of image processing, Fast Fourier Transform, motion detection, kNN, and modified DT from YCbCr color recognition for human skin to ultimate gesture recognition. Its successful demonstration of the integration of algorithms needed for solving the image detection and pattern recogni- tion problems is beneficial to academic.…”

“…For example, the process starts with Equation ( 16) where the input (independent variable) is set to the normalized distance measured from edge to center and the output (dependable variable) is set to Gesture 0 or 6. For the rest equations (17)(18)(19)(20), the DT mechanism is similar and the process stops when all inputs are recognized.…”

Improved Chan Algorithm Based Optimum UWB Sensor Node Localization Using Hybrid Particle Swarm Optimization

“…Estimate the mapping link between TDOA's and the relevant objectives. Vandersmissen et.al [27] proposed Chan algorithm and the Taylor algorithm are two regular approaches for solving nonlinear positioning equations for radiolocation in a two-dimensional (2-D) space. As a result, Chan approach has a low computational complexity and good accuracy in high signal-to-noise ratio (SNR) and Gaussian noise environments improved accuracy of node localization due to algorithm hybridization and discussed various methods for recognizing indoor human activity involved a combination of video-camera and radar sensors combined with a convolutional neural network.…”

Using Smart Virtual-Sensor Nodes to Improve the Robustness of Indoor Localization Systems

“…Aiming at recognizing indoor human activity using high-dimensional sensors and deep neural networks, a fusion of video-camera and radar sensors by a three-dimensional convolutional neural approach was explored in [ 24 ]. In contrast, [ 25 ] applied radio sensors to fingerprint-based device-free (DF) WiFi indoor localization that coped with noisy channel-state information.…”