README.md

metadata

license: mit
language:
  - en
tags:
  - radar
  - signal-processing
  - micro-doppler
  - helicopter
  - classification
  - synthetic
  - iq-data
  - time-series
pretty_name: Micro-Doppler Signatures (Helicopter)
size_categories:
  - 100K<n<1M
task_categories:
  - tabular-classification
  - audio-classification

Micro-Doppler Signatures - Helicopter Classification Dataset

Synthetic IQ-sampled radar returns from three helicopter types, generated using a physics-based sinc micro-Doppler scattering model. Designed for benchmarking ML classifiers on rotating-blade target identification.

Dataset Description

What's in it

Two CSV files, each with 100,000 samples:

File	Contents
helicopter_microdoppler_dataset.csv	Baseline - fixed radar geometry, SNR in [5, 25] dB
helicopter_microdoppler_extended_dataset.csv	Extended - variable radar frequency (8-12 GHz), elevation angle (0-45 deg), bulk target velocity (+/-50 m/s)

Each row is one 0.5-second radar observation window at 1 kHz sampling rate (500 complex IQ samples per window), flattened as real and imaginary columns.

Target classes

Label	Helicopter	Blades	Rotor RPM	Blade length
2	Bell UH-1 Iroquois (Huey)	2	300-350	6.5-7.5 m
3	Aerospatiale Gazelle	3	360-400	4.5-5.5 m
4	Boeing AH-64 Apache / UH-60 Black Hawk	4	250-300	7.0-8.5 m

Signal model

The micro-Doppler return from the k-th blade is modelled as:

$$s_k(t)=L\cdot \text{sinc}\text{\hspace{0.17em}⁣}(\frac{2L}{\lambda }\cos {\varphi }_k\cos \beta )\exp \text{\hspace{0.17em}⁣}(j\frac{4\pi L}{\lambda }\cos {\varphi }_k\cos \beta )s\_k(t)\; =\; L\; \backslash cdot\; \backslash text\{sinc\}\backslash !\backslash left(\backslash frac\{2L\}\{\backslash lambda\}\backslash cos\backslash phi\_k\backslash cos\backslash beta\backslash right)\; \backslash exp\backslash !\backslash left(j\backslash frac\{4\backslash pi\; L\}\{\backslash lambda\}\backslash cos\backslash phi\_k\backslash cos\backslash beta\backslash right)$$

where phi_k(t) = omegat + theta_0 + 2pi*k/N_b is the instantaneous blade phase, L is blade half-length, lambda is radar wavelength, and beta is the elevation angle. AWGN is added to reach the target SNR.

Column Schema

label          - int {2, 3, 4}         helicopter class (number of main rotor blades)
snr_db         - float                 signal-to-noise ratio of this sample
n_blades       - int                   number of rotor blades
rpm            - float                 rotor revolutions per minute
blade_length_m - float                 blade half-length in metres
I_0 ... I_499  - float                 in-phase (real) IQ samples
Q_0 ... Q_499  - float                 quadrature (imaginary) IQ samples

Extended dataset additionally includes:

radar_freq_ghz - float                 radar carrier frequency (8-12 GHz)
elevation_deg  - float                 target elevation angle (0-45 deg)
velocity_ms    - float                 bulk target radial velocity (-50 to +50 m/s)

Intended Use

• Benchmarking classical and deep learning classifiers on radar micro-Doppler data

• • Evaluating robustness to noise (SNR sweep experiments)

  • • Research into quantum kernel methods on signal classification tasks

    • • Open-set recognition and out-of-distribution detection studies

      • Limitations

    • • Synthetic data only - real radar returns include ground clutter, multipath, and hardware-specific artefacts not modelled here

      • • Three helicopter classes only - does not cover fixed-wing aircraft, drones, or birds

        • • Monostatic radar geometry assumed

          • Related Repository

        • Code, notebooks, and full experimental pipeline:

        • bukac82/radar-microdoppler-ai

      • Citation

    • If you use this dataset in your research, please cite the dataset/software repository and/or the relevant papers below.

  • Dataset & Software Repository

  • @software{agnihotri2026microdoppler,
      author    = {Agnihotri, Vikas},
      title     = {Radar Micro-Doppler AI: End-to-End Helicopter Classification},
      year      = {2026},
      url       = {https://github.com/bukac82/radar-microdoppler-ai}
    }
    

    Related Papers

    Quantum ML on NISQ Hardware:

    @article{agnihotri2026quantum,
      author    = {Agnihotri, Vikas and Kaur, Jasleen and Kaushik, Sarvagya},
      title     = {Practical Evaluation of Quantum Kernel Methods for Radar
                   Micro-Doppler Classification on Noisy Intermediate-Scale
                   Quantum ({NISQ}) Hardware},
      journal   = {arXiv preprint},
      volume    = {arXiv:2601.22194},
      year      = {2026},
      url       = {https://arxiv.org/abs/2601.22194}
    }
    

    Radar-Based ATR Framework (foundational SVM/signal model):

    @article{agnihotri2020radar,
      author    = {Agnihotri, Vikas and Sabharwal, Munish},
      title     = {An Automatic Radar Based Aerial Target Recognition Framework},
      journal   = {Journal of Interdisciplinary Mathematics},
      volume    = {23},
      number    = {2},
      pages     = {321--333},
      year      = {2020},
      doi       = {10.1080/09720502.2020.1737377},
      url       = {https://doi.org/10.1080/09720502.2020.1737377}
    }
    

    Frequency Effects on Micro-Doppler (underpins extended dataset design):

    @inproceedings{agnihotri2019frequency,
      author    = {Agnihotri, Vikas and Sabharwal, Munish and Goyal, Vinay},
      title     = {Effect of Frequency on Micro-Doppler Signatures of a Helicopter},
      booktitle = {2019 International Conference on Advances in Big Data,
                   Computing and Data Communication Systems (icABCD)},
      year      = {2019},
      doi       = {10.1109/ICABCD.2019.8851024},
      url       = {https://doi.org/10.1109/ICABCD.2019.8851024}
    }
    

    License

    MIT - free to use for research and commercial purposes with attribution.