Hardware | Hardware

We present a system and method for non-contact vital signs monitoring (NCVSM) of multiple individuals, using a frequency-modulated-continuous-wave radar. By utilizing the sparse nature of the signals, we achieve accurate multi-person localization and NCVSM even in cluttered scenarios, where known techniques struggle. In addition, we developed a dedicated phantom that simulates human breathing, to enhance estimation performance and calibration. Our approach may ease the burden on medical staff and help prevent infectious diseases.

We present a dedicated hardware prototype, called modulo-ADC, that can handle high-dynamic-range (HDR) input signals by employing the non-linear modulo operator. The prototype executes the modulo operation on the HDR input signal and transforms the folded signal into bit sequences. In addition, we propose fast and robust unfolding algorithms, that operate at low sampling rates by formulating the problem as a sparse signal recovery problem. Through hardware results, we demonstrate that the proposed algorithm is fast, robust, and can recover high dynamic range signals at a low sampling rate.

We introduce a framework encompassing theory, algorithms, and hardware implementation for an asynchronous ADC designed to accommodate various signal classes. Operating without a global clock, it utilizes irregular sampling time intervals determined by integral crossings above a predefined threshold. This approach is referred to as IF-TEM (integrate and fire time encoding machine). IF-TEM's key advantage lies in eliminating the global clock, resulting in power savings and bit reduction. Our focus extends to structured signals, including Finite Rate of Innovation (FRI) signals and variable-pulse-width (VPW) forms like ECG, crucial in radar, ultrasound, and biomedicine. We establish that for such structured signals, sampling rates can be set below the Nyquist rate, achieving sub-Nyquist rates akin to synchronous sampling advancements.

Sampling and quantization play a crucial role in digital signal processing systems, allowing continuous amplitude signals to be represented with a finite number of bits. However, accurately representing signals necessitates a large number of quantization bits, resulting in significant cost, power consumption, and memory load. Task-based quantization allows to dramatically reduce the quantization rates by taking the task into account when designing the quantizer. An analog combiner is introduced prior to sampling and optimized to the task. This allows to quantize using a small number of bits with a simple scalar quantizer in each channel, without impairing performance.

High-resolution direction of arrival (DOA) estimation requires a large number of antenna elements which increases the computational cost, hardware complexity, and power requirements, which are critical resources in an automotive radar system. To balance between hardware complexity and resolution, we proposed a cognitive, scalable, sparse array selection technique based deep neural networks. In this demo, we present a design and implementation of a hardware prototype that demonstrate the proposed sparse antenna selection strategy. Through real-time experiments, we show that the proposed sparse selection method results in a 2-3 dB lower error compared to a typically employed random selection method.

Sub-Nyquist radar systems operate at lower sampling rates compared to the Nyquist rate and hence reduce the hardware cost and complexity. Sub-Nyquist systems uses the knowledge of the transmit pulse and the receive signal model to estimate the targets from lowrate samples. However, in practice, the pulse shape is often distorted and unknown at the receiver. Recently, a multiple-receiver based sub-Nyquist radar was proposed that estimates the targets without knowledge of the pulse shape. In this demo, we build a hardware prototype to demonstrate the proposed sub-Nyquist radar. We show that while operating at 10 times below the Nyquist rate the proposed two-receiver system with unknown pulse has comparable performance to a single-receiver sub-Nyquist system with a known pulse.

In this demonstration, we address the hardware complexity challenge in a single-cell multi-user MIMO system, focusing on the task of estimating the underlying channel. In order to reduce the number of receive RF chains, we present a hardware prototype implementing analog combining for RF chain reduction. The prototype consists of a specially designed configurable combining board as well as a dedicated experimental setup. We adopt the Kronecker channel model with known second-order statistics of the channel (i.e., transmit and receive side covariance matrices), and show that the optimal combiner corresponds to the first eigenvectors of the receive side covariance matrix. Afterwards, the channel estimation with reduced receive RF chains is realized following a Bayesian approach, by applying the minimum mean squared error (MMSE) channel estimator to the output of our proposed analog combiner. The experimental study, which focuses on channel estimation accuracy in MIMO channels, demonstrates that using the proposed prototype, the achievable channel estimation performance is within a small gap in a statistical sense from that obtained using a costly receiver in which each antenna is connected to a dedicated RF chain. This prototype is developed at the Weizmann Institute of Science and was demonstrated at International Conference on Acoustics, Speech and Signal Processing (ICASSP) at Brighton, UK in May 2019.

In this demonstration, the radar prototype is designed on the principles of a sub-Nyquist collocated multiple-input multiple-output (MIMO) radar. The setup allows sampling in both spatial and spectral domains at rates much lower than dictated by the Nyquist sampling theorem. We use frequency division multiplexing (FDM) to achieve the orthogonality of MIMO waveforms and apply the Xampling framework for signal recovery. The prototype also implements a cognitive transmission scheme where each transmit waveform is restricted to those pre-determined subbands of the full signal bandwidth that the receiver samples and processes. Real-time experiments show reasonable recovery performance while operating as a 4×5 thinned random array wherein the combined spatial and spectral sampling factor reduction is 87.5% of that of a filled 8 × 10 array

We present for the first time a design and implementation of a Xampling-based hardware prototype that allows sampling of radar signals at rates much lower than Nyquist. We demonstrate by real-time analog experiments that our system is able to maintain reasonable detection capabilities, while sampling radar signals that require sampling at a rate of about 30MHz at a total rate of 1Mhz, namely, at 1/30 of the Nyquist rate. Our board is based on a 4-channel crystal receiver. To evaluate the board we make use of National Instrument (NI) PXI equipment.

We demonstrate the first wideband sub-Nyquist receiver that can sample and process multiband signals at rates far below the Nyquist rate. Our prototype, referred to as the Modulated Wideband Converter (MWC), samples multiple narrowband transmissions at a rate proportional to the actual bandwidth occupation, without knowledge of the carrier positions. Our specific implementation uses a rate that is only 8% of the Nyquist rate. Various extensions to the MWC are also presented, among them collaborative processing, direction of arrival (DOA) estimation and modulated data reconstruction (PSK, OFDM). These prove that the MWC is both a viable and flexible solution for future communication systems.

We demonstrate a real-time fixed-point embedded implementation of the sub-Nyquist reconstruction algorithm in the modulated wideband converter. The embedded design enables, for example, fast spectrum sensing, in a time duration as low as several micro-seconds, which is essential to real-time cognitive radio applications. The system uses LabView and is implemented on an Altera Stratix III field-programmable gate array (FPGA) mounted on a Gidel PROCStar-III development board.