An Embedded ZigBee Identification System Based on LMS Filter

ZigBee technology, which is widely used in low-power and low-data-rate applications, encounters challenges in accurate device recognition. This paper presents a deep learning-based approach for device identification utilizing the physical radio frequency fingerprint (RFF) of ZigBee devices. To enhance performance in cross-scenario identification, artificial noise and simulated channel impairments were introduced for data augmentation. A Least Mean Squares (LMS) filter was employed for estimating channel coefficients, specifically designed for channel impulse response simulation. The proposed methodology was implemented and validated on the cost-effective LPA3588 platform using a convolutional neural network (CNN) based on differential constellation trace figures (DCTF). Experimental results demonstrate the effectiveness of this approach in improving the performance of deep learning models for device recognition under real-world conditions.