System Prototyping
Overview (hardware implementation in waveform design, 5G/6G, IoT, AI, MIMO, Security)
1. 2023: irSinc shaped NOFS signal waveform for 6G: New waveform design for 6G (Nature Communications Engineering).
2. 2022: Over-the-air ISAC multiuser-MIMO for 6G: Integrated sensing and communications:ISAC (IEEE Open journal, MobiCom).
3. 2022: Deep learning UnfoldingDecNet detector: Unfolding neural network based signal detection (IEEE TWC journal).
4. 2021: Over-the-air Security: Waveform-defined security (IEEE IoT journal).
5. 2020: AI-sensing over-the-air: Wavelet based classification on non-orthogonal signals (IEEE GLOBECOM).
6. 2019: AI-sensing over-the-air: Deep learning non-cooperative waveform communications (IEEE VTC).
7. 2019: Hybrid analog-digital multiuser-MIMO beamforming: Hybrid analogue and digital beamforming (IEEE IoT journal).
8. 2018-2019: Next-generation 6G IoT waveform: Signal precoding (IEEE IoT journal, IEEE INFOCOM, IEEE PIMRC); Extend signal coverage (IEEE GLOBECOM); Data rate enhancement (IEEE PIMRC); Double data rate (IEEE PIMRC); Double connected IoT devices (IEEE IoT journal).
9. 2018: Waveform-AI: Neural network compression for non-orthogonal signal detection (IEEE WCNC).
10. 2017-2018: 5G-SDR platform: USRP self-interference cancellation signal communications over the air (IEEE VTC); Dual USRP setup for the coexistence of 4G and 5G signals (IEEE CSNDSP); Pre-commercialization 5G USRP platform design and over-the-air testing (IEEE PIMRC).
11. 2016: mm-wave 60 GHz platform: 60 GHz non-orthogonal signal waveforms are delivered at 3.75 Gbit/s through 250 meters multimode fiber over 3 meters wireless link (IEEE JLT journal).
12. 2015: 5G compressed-CA platform: Bandwidth compressed carrier aggregation wireless signal waveform transmission using SDR transceiver and Spirent VR5 channel emulator (IEEE TVT journal).
13. 2015: 4G-SDR platform: LTE/LTE-A experimental testbed. (IEEE ICC).
14. 2014-2015: Optical fiber platform: Direct detection optical testbed to transmit non-orthogonal signals at 10 Gbit/s (IEEE PTL journal); Dual polarization coherent detection optical testbed to transmit non-orthogonal signals at 24 Gbit/s (IEEE PTL journal).
15. 2013: FPGA based detector: A real-time MIMO/SEFDM detector working at 1.06 Gbit/s was designed and implemented on a Xilinx Vertex-6 FPGA chip (IEEE ICT).
Massive MIMO Platform
irSinc Shaped NOFS Signal
Integrated Sensing and Communications (ISAC)
Hybrid Analog-Digital Multiuser MIMO Beamforming
Waveform Design for Spectrally Efficient IoT
Feature: Fast-OFDM non-orthogonally compresses sub-carriers leading to 50% spectral bandwidth saving.
Benefits:
Double the number of connected devices (users) without occupying extra resources.
Double the data rate without occupying extra resources.
Improve device power efficiency via waveform scheduling.
Extend signal coverage through constructive channels.
Extend signal coverage since lower coding rate is allowed.
Achieve the same BER performance.
Maintain similar computational complexity.
Spectrally Efficient Bandwidth Compression Carrier Aggregation
Feature: Spectrally efficient FDM (SEFDM) non-orthogonally compresses sub-carriers leading to (1-a)x100% spectral bandwidth saving. '0<a<1' is the bandwidth compression factor, which determines the bandwidth saving ratio. The special case, Fast-OFDM has a=0.5.
Benefits:
Improve data rate via tuning waveforms instead of higher order modulation schemes.
Type-I signal maintains the same data rate but saves (1-a)x100% spectral bandwidth.
Type-II signal maintains the same spectral bandwidth but improves data rate by (1/a-1)x100%.
Suppression of inter-modulation distortion (IMD) in optical access networks.
Robust to fiber non-linearity in long-haul optical fiber transmission.
Over-The-Air 60 GHz mmWave Signal Transmission
Filtering Non-Orthogonal Signal Waveform Design for Non-Interference Cognitive Radio
Feature: Nyquist-SEFDM reserves the benefits of SEFDM and further cuts out-of-band power leakage.
Benefits:
Low out-of-band power leakage enables smooth signal coexistence.
Reduced occupied spectral bandwidth.
Half-Sinc Signal for Energy Efficient IoT
Feature: Half-Sinc saves 50% spectral bandwidth for single-carrier waveform.
Benefits:
Double the number of connected devices (users) without occupying extra resources.
Double the data rate by packing two Half-Sinc waveforms without occupying extra resources.
Robust to frequency offset in frequency hopping scenarios due to the protection gap.
Deep Learning Neural Network Topology Design for Signal Detection
Background:
IoT devices require low power consumption, thus simple signal processing.
Neural network is much simpler than typical communications algorithms such as maximum likelihood (ML), sphere decoding (SD).
AI+IoT (AIoT) will be a competitive combination.
AIoT doesn’t focus on an entire communication system but the receiver side signal detection. Therefore, time-variant wireless fading channels do not affect machine learning training and prediction accuracy.
Waveform detection is highly related to signal waveform features. Therefore, the correlation of signal waveform and neural network topology plays an important role and a joint design is more efficient.
Deep Learning/Machine Learning Based Intelligent Signal Sensing
Deep Learning Adversarial Attack Resistance Secure Communications:
Waveform-Defined Security (WDS) Framework
Optical Fiber Non-Orthogonal Signal Transmission Experiment
