A Kalman Filter-Enhanced Image Reconstruction Technique for Privacy-Preserving Human Action Recognition Using mmWave Radar Smart Sensors
Keywords:
Human Action Recognition (HAR), Kalman filter, mmWave radar, Privacy-preserving sensing, Sparse-to-dense heatmap reconstructionAbstract
Millimeter-Wave (mmWave) radar provides a privacy-preserving alternative to RGB cameras for Human Action Recognition (HAR). However, its sparse and noisy point clouds often limit classification accuracy. Existing temporal filtering methods typically apply Kalman filters only to centroid tracking, rather than full image reconstruction. This paper proposes a Kalman filter-enhanced image reconstruction framework that integrates a full state-space model into the sparse-to-dense heatmap generation pipeline. Raw point clouds are first preprocessed using adaptive DBSCAN clustering. A Kalman filter is then employed to estimate a dense 64×64 heatmap based on constant-velocity dynamics. To efficiently manage the high-dimensional 4096-state vector, the study assume pixel-wise independence (a diagonal covariance matrix), reducing computational complexity from O (N³) to O(N) per update. The filtered points are subsequently projected onto a 2D grid using Gaussian kernel density estimation. Finally, a lightweight CNN-LSTM network performs action classification. Evaluated on the RadHAR dataset (five activities, ten subjects) using a Texas Instruments IWR1443 radar and an NVIDIA Jetson Nano, the proposed method achieves 94.78% accuracy. This represents a substantial improvement of 13.43 percentage points over vanilla reconstruction (81.35%) and 5.05 percentage points over centroid-only Kalman filtering (89.73%). The approach also demonstrates strong robustness at low frame rates (86.37% at 5 fps vs. 44.38%) and under occlusions (SSIM of 0.732 vs. 0.418). Privacy evaluations confirm zero successful face recognition and only chance-level re-identification performance (8.2%). Notably, the Kalman filter executes in just 12.8 ms per frame on the Jetson Nano, enabling real-time inference at up to 19 fps. By integrating a full state-space Kalman filter into the mmWave radar heatmap reconstruction process, the proposed framework significantly enhances accuracy, occlusion resilience, and temporal coherence, while maintaining strong privacy protection and satisfying strict edge-device real-time constraints.
References
Y. Yang, P. Hu, J. Shen, H. Cheng, Z. An, and X. Liu, “Privacy-preserving human activity sensing: A survey,” High-Confidence Computing, vol. 4, no. 1, Mar. 2024.
A. B. Chan, Zhang-Sheng John Liang and N. Vasconcelos, “Privacy preserving crowd monitoring: Counting people without people models or tracking,” 2008 IEEE Conference on Computer Vision and Pattern Recognition, Anchorage, AK, USA, 2008, pp. 1–7.
W. Mucha and M. Kampel, “Beyond privacy of depth sensors in active and assisted living devices,” Proceedings of the 15th International Conference on PErvasive Technologies Related to Assistive Environments, Jul. 2022, pp. 425–429.
Y. Soni, M. Goswami, N. P. Shetty, and Dhiraj, “Millimeter-wave radar for intelligent sensing: A comprehensive review of techniques, applications, and challenges,” Computers and Electrical Engineering, vol. 128, Dec. 2025.
C. Zhao et al., “mm-Fall: Practical and robust fall detection via mmWave signals,” in IEEE Transactions on Mobile Computing, vol. 24, no. 9, pp. 8747–8760, Sept. 2025.
H. Kong, C. Huang, J. Yu and X. Shen, “A survey of mmWave radar-based sensing in autonomous vehicles, smart homes and industry,” in IEEE Communications Surveys & Tutorials, vol. 27, no. 1, pp. 463-508, Feb. 2025.
A. Naser, A. Lotfi and J. Zhong, “Multiple thermal sensor array fusion toward enabling privacy-preserving human monitoring applications,” in IEEE Internet of Things Journal, vol. 9, no. 17, pp. 16677–16688, Sept., 2022.
Texas Instruments, Texas Instruments IWR1443 Single Chip 76-GHz → 81-GHz mmWave Sensor Evaluation Module mmWave Sensor Evaluation, RS Stock No.: 235-5150.
R. E. Hail, P. Mehrjouseresht, D. M. M.-P Schreurs, and P. Karsmakers, “Radar-based human activity recognition: A study on cross-environment robustness,” Electronics, vol. 14, no. 5, Feb. 2025.
C. Zhu, J. Yang, P. Deng, J. Lin, L. Huang and H. Lin, “STPointNet for Human Action Recognition in MmWave point clouds,” 2024 27th International Conference on Computer Supported Cooperative Work in Design (CSCWD), Tianjin, China, 2024, pp. 550–555.
X. Wang, Y. Chen, F. Wang, K. Zheng, and Y. Zong, “Mpct-Lstm: A lightweight human activity recognition model for 3d point clouds in Millimeter-Wave Radar,” Digital Signal Processing, vol. 164, Sep. 2025.
Y. Wu, F. Fioranelli, and C. Gao, “RadMamba: Efficient Human Activity Recognition through radar-based micro-doppler-oriented mamba state-space model,” arXiv, 2025.
J. Yang et al., “mmDEAR: mmWave point cloud density enhancement for accurate human body reconstruction,” 2025 IEEE International Conference on Robotics and Automation (ICRA), Atlanta, GA, USA, 2025, pp. 11227–11233.
J. Fan, H. Rao and X. Wang, “From radar to depth: Multi-modality co- learning for high-resolution human pose estimation,” 2024 IEEE International Conference on Smart Internet of Things (SmartIoT), Shenzhen, China, 2024, pp. 248–253.
S. Kato, P. P. Wang, T. Fujihashi and A. Markham, “Heatmap-to-SMPL multi-view radar transformer for multi-person 3D pose estimation,” ICASSP 2026 - 2026 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Barcelona, Spain, 2026, pp. 20302–20306.
J. Pegoraro and M. Rossi, “Real-time people tracking and identification from sparse mm-Wave radar point-clouds,” in IEEE Access, vol. 9, pp. 78504–78520, 2021.
G. Liu, C. Chang, S. -H. Fang, H. -C. Wu and K. Yan, “Novel hybrid machine-learning technique for robust indoor multisubject tracking using mmWave radar,” in IEEE Internet of Things Journal, vol. 13, no. 8, pp. 15929–15942, Apr. 2026.
F. Ghura, K. Poon and U. Ahmad, “Evaluation of detection and tracking approaches for people counting using mmWave radar,” 2025 8th International Conference on Signal Processing and Information Security (ICSPIS), Dubai, United Arab Emirates, 2025, pp. 1–5.
E. Ahmad, Y. He, Z. Luo and J. Lv, “A hybrid long short-term memory and Kalman Filter model for train trajectory prediction,” in IEEE Transactions on Intelligent Transportation Systems, vol. 25, no. 7, pp. 7125–7139, July 2024.
A. R. Pitombeira-Neto, D. S. de França, L. A. Cruz and J. F. A. de Macedo, “Human Activity Recognition via a categorical Kalman Filter using inertial sensors,” in IEEE Sensors Journal, vol. 26, no. 9, pp. 13723–13737, May 2026.
J. Yan, Y. Wang, C. Xu and D. Liu, “OG-PCL: Efficient sparse radar point cloud processing for Human Activity Recognition,” ICASSP 2026 - 2026 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Barcelona, Spain, 2026, pp. 20137–20141.
Z. Gu, X. He, G. Fang, C. Xu, F. Xia, and W. Jia., “Millimeter wave radar-based Human Activity Recognition for healthcare monitoring robot,” arXiv, 2024.
L. Zhang, J. Zhang, H. Shi, L. Gao, X. Hu and R. Chen, “Adaptive point cloud clustering algorithm for practical roadside MmWave radar systems,” 2024 IEEE 99th Vehicular Technology Conference (VTC2024-Spring), Singapore, Singapore, 2024, pp. 1–5.
R. B. Rusu and S. Cousins, “3D is here: Point Cloud Library (PCL),” 2011 IEEE International Conference on Robotics and Automation, Shanghai, China, 2011, pp. 1–4.
M. Ester, H.-P. Kriegel, J. Sander, and X. Xu, “A density-based algorithm for discovering clusters in large spatial databases with noise,” in Proceedings of the Second International Conference on Knowledge Discovery and Data Mining, Aug. 1996, pp. 226–231.
R. E. Kalman, “A new approach to linear filtering and prediction problems,” Journal of Basic Engineering, vol. 82, pp. 35–45, Mar. 1960.
G. Evensen, “The Ensemble Kalman Filter: theoretical formulation and practical implementation,” Ocean Dynamics, vol. 53, pp. 343–367, Nov. 2003.
B. W. Silverman, Density Estimation for Statistics and Data Analysis, 1st ed. Boca Raton, FL, USA: Routledge, 1998.
H. E. Rauch, F. Tung, and C. T. Striebel, “Maximum likelihood estimates of linear dynamic systems,” AIAA Journal, vol. 3, no. 8, pp. 1445–1450, May. 2012.
S. Ioffe and C. Szegedy, “Batch normalization: Accelerating deep network training by reducing internal covariate shift,” Proceedings of the 32nd International Conference on Machine Learning - Volume 37, Jul. 2015, pp. 448–456.
Y. Wang, H. Liu, K. Cui, A. Zhou, W. Li and H. Ma, “m-Activity: Accurate and real-time Human Activity Recognition via millimeter Wave Radar," ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Toronto, ON, Canada, 2021, pp. 8298–8302.
H. Cui, S. Zhong, J. Wu, Z. Shen, N. Dahnoun, Y. Zhao, “MiliPoint: A point cloud dataset for mmWave radar,” arXiv, 2023.
Student, “The Probable Error of a Mean,” Biometrika, vol. 6, no. 1, pp. 1–25, Mar. 1908.
I. Zakariyya, L. Tran, K. B. Sivangi, P. Henderson, and F. Deligianni, “Differentially Private Integrated Decision Gradients (IDG-DP) for Radar-based Human Activity Recognition,” arXiv, Nov. 2024.
A. Deep, S. S. Sandha, L. Garcia, and M. Srivastava, “RadHAR: Human activity recognition from point clouds generated through a millimeter wave radar,” Proceedings of the 3rd ACM Workshop on Millimeter-wave Networks and Sensing Systems, Oct. 2019, pp. 51–56.
F. Jin, A. Sengupta and S. Cao, “mmFall: Fall detection using 4-D mmWave Radar and a hybrid variational RNN AutoEncoder,” in IEEE Transactions on Automation Science and Engineering, vol. 19, no. 2, pp. 1245–1257, April 2022.
J. Zhang et al., “A survey of mmWave-based human sensing: Technology, platforms and applications,” in IEEE Communications Surveys & Tutorials, vol. 25, no. 4, pp. 2052–2087, Fourthquarter 2023.
A. Soumya, C. K. Mohan, and L. R. Cenkeramaddi, “Recent advances in mmWave-radar-based sensing, its applications, and machine learning techniques: A review,” Sensors, vol. 23, no. 21, Jan. 2023.
