Abstract
Gesture recognition is key to enabling natural human-computer interactions. Existing approaches based on wireless sensing focus on accurate identification of arm gesture types. It remains a challenge to recognize and profile the details of arm gestures for precise interaction control. In addition, current approaches have strict positioning requirements between radars and users, making them difficult for real-world deployment. In this article, we adopt the multisensor approach and present gesture-radar—a dual Doppler radar-based gesture recognition and profiling system, which can capture subtle arm gestures with less positioning or environmental dependence. Gesture-radar uses two vertically placed Doppler radars to collect complementary sensing data of gestures, based on which cross-analysis can be performed for gesture recognition and profiling. Specifically, we first propose a two-stage classification model and enhance the signal proximity matching method by applying constraint functions to the DTW algorithm, aiming to improve the accuracy of gesture type recognition. Afterward, we establish and analyze unique features from the time-frequency spectrogram, which can be used to characterize in-depth gesture details, e.g., the angle or range of an arm movement. Experimental results show that gesture-radar achieves up to 93.5% average accuracy for gesture type recognition, and over 80% precision for profiling gesture details. This proves that the proposed approach is viable and can work in real-world environments.
Introduction
While touchless gesture recognition using Wi-Fi or single-radar systems has advanced significantly, it faces two major hurdles for real-world deployment: environmental sensitivity and a lack of granularity. Wi-Fi-based systems are often unstable due to channel interference, while single-radar setups suffer from "blind spots"—they can only easily detect motion moving directly toward or away from the sensor (radial velocity). Furthermore, most existing systems are "coarse-grained," meaning they can identify that a wave occurred but cannot determine the exact angle or range of that movement. This prevents the precise control needed for complex interfaces, such as fine-tuning light intensity or scrolling through detailed menus.
System Architecture and Methodology
The Gesture-Radar system utilizes two 24GHz K-LC2 short-range Doppler radar sensors placed vertically: a "front" radar and a "side" radar. This configuration ensures that the "tangential" component (invisible to one radar) becomes the "radial" component (visible to the other), effectively eliminating sensing blind spots.
The system pipeline consists of four main phases:
- Signal Acquisition: Capturing dual-channel In-phase (I) and Quadrature (Q) signals.
- Preprocessing: Utilizing low-pass filtering and Discrete Wavelet Transform (DWT) to denoise and compress the data.
- Two-Stage Classification:
- Stage I (Phase Discrimination): The system uses phase unwrapping on the I/Q signals to determine if a gesture is approaching or departing, acting as a fast filter to narrow the search space.
- Stage II (Signal Matching): It employs Restricted Dynamic Time Warping (R-DTW) combined with a k-Nearest Neighbor (kNN) classifier. R-DTW constrains the matching path to specific boundaries (linear, rhombic, or parabolic), reducing computation and preventing illogical signal matches.
- Quantitative Profiling: By converting time-amplitude signals into time-frequency spectrograms, the system calculates the ratio of signal intensities between the two radars to estimate the specific angle (θ) of the arm movement.
Experimental Evaluation
The researchers recruited 10 volunteers to perform two types of data collection: standard gestures and non-standard "free-style" movements. The standard dataset included 12 distinct arm movements performed 10 times each, totaling 1,200 samples.
Key Results
- Recognition Accuracy: The system achieved an average accuracy of 93.5% across the 12 gesture types, significantly outperforming baseline models like Random Forest or SVM.
- Profiling Precision: For non-standard gestures, the system estimated movement angles with a mean error of less than 15°, achieving over 80% precision in quantitative profiling.
- Robustness: The system maintained performance over 90% in multi-person environments with up to 4 people, provided non-target individuals remained relatively still within 3 meters of the sensors.
Conclusion and Future Work
Gesture-Radar demonstrates that a dual-radar approach can provide the granularity required for sophisticated human-computer interaction. While the system is robust, future work will address multi-user interference from large-scale movements and explore microsecond-level hardware synchronization. Additionally, the team plans to implement transfer learning to make the system more "plug-and-play" across diverse physical environments without extensive recalibration.
Summary updated