Radar · Sensing · Signal processing

Millimeter-wave & micro-Doppler radar systems

My group works on 94 GHz W-band radar. At these frequencies the short wavelength provides high phase and Doppler sensitivity to small motions; we develop radar systems and micro-Doppler processing that converts motion-induced modulations into features for target classification, combining classical estimation with machine learning.

94 GHz radar front-end built from gold-plated WR-10 waveguide parts on an aluminum plate, with a horn antenna aimed down a linear track at a corner reflector. — The lab's 94 GHz front-end (WR-10 waveguide parts and a horn on an aluminum plate), aimed down a linear track at a corner reflector.

Applications

The same core sensor and processing chain addresses a range of published and ongoing remote-sensing problems:

Target classification & low-RCS detection: micro-Doppler signatures and machine-learning methods support classification of vehicles and other targets; ongoing work examines drone detection.
Remote metrology: velocity-independent vehicle-height estimation from the ratio of two Doppler components, for bridge-collision prevention (Sensors 2026).
Condition monitoring: multi-fault detection on wind turbines and rotating machinery from their micro-Doppler spectra.
Human sensing: remote sensing of human micro-motion, including speech-related vibration (published feasibility study, Remote Sensing 2025), fall detection, and limb-movement signatures; ongoing work targets contactless vital-signs monitoring (respiration and heart rate).

Method

Work spans the full chain: waveform and antenna choices, time–frequency (spectrogram) analysis, matched filtering and high-resolution estimation, and deep-learning classifiers, implemented and validated in MATLAB, with experimental data from real W-band hardware. Antenna front-ends are designed with the same full-wave tools used for the antenna work.

Radar cross-section (RCS) modelling

How strongly a target reflects, its radar cross-section (RCS), sets received power and detectability and can contribute aspect-dependent features for classification. I compute RCS from full-wave electromagnetic scattering and, for electrically large targets, physical-optics (PO) methods, in COMSOL and MATLAB.

COMSOL full-wave electromagnetic simulation of a plane wave scattering off a target, used to compute its radar cross-section (RCS). — Full-wave electromagnetic scattering off a target in COMSOL: the field solution from which the radar cross-section is extracted.

Monostatic RCS versus aspect angle for a rotating blade at 94 GHz (physical optics): the angular signature that underlies micro-Doppler classification of rotating targets.

Watch the rotating-blade RCS animation →

Simulating a quadcopter’s micro-Doppler at 94 GHz

Each rotor blade moves toward and away from the radar as it turns, so the echo carries Doppler shifts up to the blade-tip speed. In a short-window spectrogram this shows as vertical blade flashes, one each time a blade crosses perpendicular to the line of sight. In a long window the same signal splits into combs of helicopter rotor modulation (HERM) lines, spaced by the number of blades times the rotation rate.

Left: the four rotors seen from above, slowed 240 times; a blade turns orange when it crosses perpendicular to the line of sight. Right: 0.75 ms spectrogram. The flashes reach ±38 to ±45 kHz around the body line and repeat every 5.5 to 6.5 ms for each rotor.

Long-window spectrum of the simulated quadcopter: four interleaved combs of lines with spacings of 155, 164, 170 and 181 Hz, and the full-band spectrum bounded by the blade-tip Doppler. — HERM lines from a 0.5 s window of the same signal: four interleaved combs, one per rotor, spaced 155, 164, 170 and 181 Hz (two blades times the rotation rate).

The model sums point scatterers along rigid blades (CW radar, rotors at 4650 to 5430 rpm). Its tip Doppler, flash period and line spacing match the closed-form kinematics, a check internal to the model. Blade bending and W-band measurements are covered in our paper in Sensors 26, 4851 (2026).

77 GHz automotive radar behind a bumper

An automotive radar sits behind a plastic bumper, and the panel acts as a dielectric slab. When its thickness is a multiple of half a wavelength inside the plastic, the reflections from its two faces cancel and the wave passes with little loss. At other thicknesses part of the power returns toward the radar. At 77 GHz the wavelength in the panel (relative permittivity 2.82) is 2.32 mm, so the matched thicknesses are 1.16, 2.32 and 3.48 mm.

Left: the field of an 8-element array passing through the panel as its thickness sweeps from 1 to 4 mm. Right: transmitted and reflected power for a plane wave, COMSOL (points) and analytic (lines). At mismatched thicknesses up to 22% of the power reflects.

COMSOL field map at 77 GHz: a radar beam steered to 30 degrees passes through a 2.90 mm bumper panel, and part of it reflects back on the radar side. — COMSOL, beam steered to 30° through a 2.90 mm panel, a quarter wavelength off the 3.48 mm match. Part of the beam passes and part reflects specularly. In the far field the main beam drops by 1.1 dB and a reflected lobe appears at −5.6 dB relative to the beam peak without the panel. Values at equal feed currents, evaluated at 30°.

One-way loss versus panel thickness from 1 to 4 mm for three cases, COMSOL points on analytic lines, with a lower panel showing the difference between them. — One-way loss versus thickness: bare panel at 0° and 30°, and the panel with a metallic basecoat. Points: COMSOL unit cell; lines: analytic multilayer slab; largest difference 0.0003 dB. At the matched thickness the bare panel costs 0.30 dB, and the paint raises the loss to 2.7 dB with about 40% of the power reflected.

The model is 2D, for one polarization (electric field parallel to the panel), a flat panel and a single frequency. The panel and paint permittivities come from measurements of single samples in a Chalmers University of Technology diploma thesis (E. P. Emilsson, 2017). The agreement with the analytic slab verifies the numerical solution; the material data are a separate source of uncertainty.

94 GHz / W-bandMicro-Doppler Radar cross-section (RCS)EM scattering · physical optics Target classificationDrone detection Remote sensingCOMSOL · MATLAB · ML

Selected radar publications

Balal Y, Steinmetz N, Sherenzon A, Balal N. Bending-Aware Spectrogram Correlation for W-Band Quadcopter Detection: A Physics-Informed Non-Coherent Detection Framework. Sensors 26, 4851 (2026). doi:10.3390/s26154851
Steinmetz N, Magori E, Balal Y, Sudai YB, Balal N. A 94 GHz Millimeter-Wave Radar System for Remote Vehicle Height Measurement to Prevent Bridge Collisions. Sensors 26, 1921 (2026). doi:10.3390/s26061921
Steinmetz N, Balal N. Feasibility Study of Real-Time Speech Detection and Characterization Using Millimeter-Wave Micro-Doppler Radar. Remote Sensing 17, 91 (2025). doi:10.3390/rs17010091
Richter Y, Balal N. High-Resolution Millimeter-Wave Radar for Real-Time Detection and Characterization of High-Speed Objects with Rapid Acceleration Capabilities. Electronics 13, 1961 (2024). doi:10.3390/electronics13101961
Richter Y, Balal N, Pinhasi Y. Neural-Network-Based Target Classification and Range Detection by CW MMW Radar. Remote Sensing 15, 4553 (2023). doi:10.3390/rs15184553
Balal Y, Yarimi A, Balal N. Non-Imaging Fall Detection Based on Spectral Signatures Obtained Using a Micro-Doppler Millimeter-Wave Radar. Applied Sciences 12, 8178 (2022). doi:10.3390/app12168178
Balal Y, Balal N, Richter Y, Pinhasi Y. Time-Frequency Spectral Signature of Limb Movements and Height Estimation Using Micro-Doppler Millimeter-Wave Radar. Sensors 20, 4660 (2020). doi:10.3390/s20174660

Applied research and technical capabilities

Radar signal processing and micro-Doppler analysis
Target classification and performance evaluation
Radar cross-section and scattering simulation
MATLAB algorithm development and experimental-data analysis
Feasibility and proof-of-concept studies

Technical questions and research collaboration: nezahb@ariel.ac.il

← All research projects