Decoding human activity accurately from wearable sensors can aid in applications related to healthcare and context awareness. The present approaches in this domain use recurrent and/or convolutional models to capture the spatio-temporal features from time-series data from multiple sensors. We propose a deep neural network architecture that not only captures the spatio-temporal features of multiple sensor time-series data but also selects, learns important time points by utilizing a self-attention mechanism. We show the validity of the proposed approach across different data sampling strategies on six public datasets and demonstrate that the self-attention mechanism gave a significant improvement in performance over deep networks using a combination of recurrent and convolution networks. We also show that the proposed approach gave a statistically significant performance enhancement over previous state-of-the-art methods for the tested datasets. The proposed methods open avenues for better decoding of human activity from multiple body sensors over extended periods of time. The code implementation for the proposed model is available at https://github.com/isukrit/encodingHumanActivity.

In this work, a novel Ultra-wideband –Multi-Input-Multi-Output Antenna Sensor (UMAS) probe is designed for the detection of the malignant cells in the breast. The Sensor probe has four radiating elements and it is operated within the 2.8 GHz to 20 GHz ultra-wide band range. Isolation between the radiating element is more than 20 dB. Further, three kinds of the breast phantoms (i.e. normal phantom, phantom with single and multiple tumors) are fabricated using tissue mimicking material. The electrical characteristics of the malignant cells are different from non-malignant cells of the breast. The S-parameter and Specific Absorption Rate (SAR) analysis are best approaches to detect the malignant cells in the breast. The UMAS sensing probe is embedded on the phantoms and S-parameters of the probe are recorded from the Vector Network Analyzer (VNA). Measured S-parameters of the probe for normal and malignant phantoms are differ from each other. The statistical machine learning concept of Principal Component Analysis (PCA) is also applied on the measured S-Parameters. Which exhibits clear detection of normal and malignant breast phantoms. Further verification is done by using Simulation based specific absorption rate (SAR) study of the phantom models for tumor detection. The obtained maximum SAR results are well differentiating the normal phantom.

A triple-band low radar cross-section (RCS) high isolation antenna is proposed for 24 GHz ISM band sensing and automotive radar applications. The proposed design consists of 2 × 2 patch array that acts as transmit and receive antennas. Low RCS and high isolation are achieved at 24 GHz by designing a metamaterial absorber (MA), which consists of a square ring with a resistor connected in its diagonal arm for the absorption of electromagnetic waves. The proposed MA shows near unity normalized impedance at 24.1 GHz with 90% absorptivity bandwidth of 1 GHz. An array of MA is placed in between and around the transmit/receive antennas to suppress surface current and reduce in-band RCS of the radar sensor. The -10 dB impedance bandwidths of the triple-band sensor antenna are 20.8 to 21.24 GHz, 23.94 to 24.55 GHz, and 27.18 to 27.5 GHz. The proposed sensor antenna achieves isolation of 34 dB between the transmit and receive ports, and peak RCS reduction of 11 dB, as compared to the reference antenna. The half-power beamwidth of the proposed sensor antenna is 38° for E-plane and 52° for H-plane at 24 GHz.

In this paper, an inset-fed dual-band patch antenna with truncated corners is presented. The ground plane of the antenna is loaded with D-shaped complementary split ring resonator (D-CSRR) structures to obtain dual-band resonance and size miniaturization. The overall electrical size of the antenna is 0.29λ₀ × 0.25λ₀ and patch size is 0.14λ₀ × 0.13λ₀, where λ₀ is evaluated at the lowest resonating frequency. The proposed planar antenna is low profile and does not need a dual feed or any external feeding mechanism for obtaining circular polarization. A size reduction of 66% is achieved as compared to the conventional patch antenna. The newly designed D-CSRR results in better miniaturization factor making proposed antenna suitable for modern wireless applications. The prototype of the antenna is fabricated and experimental findings are compared with electromagnetic and analytical results.