Deep-Learning-Based Chirp Signal Detection in Asynchronous Pulse Code Multiple Access for Massive IoT Applications
Student Contest:
Yes
Affiliation Type:
Academia
Keywords:
Low Power Wide Area Networks, Massive Internet of Things (Massive IoT), Wireless Communication, Asynchronous Pulse Code Multiple Access (APCMA)
Abstract:
Low-power wide-area networks (LPWANs) support long-range and low-power IoT communication but suffer from packet losses due to overlapping transmissions. Asynchronous Pulse Code Multiple Access (APCMA) has been proposed as a pulse-based solution for massive IoT, showing better performance than conventional LPWA schemes. To enhance receiver sensitivity and extend communication range, Chirp Spread Spectrum (CSS) has been introduced into APCMA, forming CSS-APCMA. Conventional detection methods apply dechirping and FFT followed by thresholding, which is sensitive to tuning under low SNR. This paper proposes a threshold-free detection method for APCMA using machine learning. Chirp signal features are extracted via Principal Component Analysis (PCA) and classified using a Support Vector Machine (SVM). We implemented a USRP-based APCMA system and constructed a training dataset using wired transmission under high attenuation. Evaluation in a wireless environment demonstrates that our method achieves comparable decoding performance to threshold-optimized conventional methods, while eliminating the need for manual tuning in low SNR conditions.
Track ID:
5.8
Track Name:
Japan-Korea Joint Special Session on Intelligent Systems & Applications for Future Networked Society
