Deep Learning-Aided OFDM Doppler Estimation with Low Latency for Future 6G-ISAC Systems
2026 IEEE Wireless and Microwave Technology Conference, WAMICON 2026, Florida, Amerika Birleşik Devletleri, 20 - 21 Nisan 2026, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Doi Numarası: 10.1109/wamicon68738.2026.11602019
- Basıldığı Şehir: Florida
- Basıldığı Ülke: Amerika Birleşik Devletleri
- Anahtar Kelimeler: Convolutional neural network, Doppler velocity estimation, frequency-domain correlation, latency, OFDM, range-Doppler map
- İstanbul Medipol Üniversitesi Adresli: Evet
Özet
Conventional sensing systems rely on a large number of pulses for accurate velocity estimation due to the difficulty of resolving Doppler shifts over short durations. However, as integrated sensing and communications (ISAC) emerges within sixth-generation (6G) systems targeting ultra-low latency, such long pulse sequences become impractical. To address this challenge, this paper proposes a convolutional neural network (CNN)-based approach that significantly reduces the required number of pulses for velocity estimation in a monostatic radar manner. Using the same parameters as prior work, the observation time is reduced from 20 ms to 0.2 ms while achieving an acceptable accuracy performance of the proposed method without modifying orthogonal frequency division multiplexing (OFDM) waveform structure. The results do not only indicate the proposed method's effectiveness but also prove the reliable impact of root-mean-square error (RMSE) investigation on accuracy analysis.