TACKLING THE CROWDED RADIO FREQUENCY (RF) SPECTRUM USING DEEP LEARNING Krishna Karra, March 28 2018 Real-Time AI Systems
INTRODUCTION • KickView creates real-time AI systems. • Intelligent Multi-Sensor Analytics (IMSA) • kvMotif - Our intelligent multi-sensor software stack • Collect, process and learn from networks of sensors in real-time • Directly connect AI to the physical world • Real-time AI Processing at Network Scale - Telco, IoT, Industrial, Smart Cities, Security and Defense • Real-time collection, processing, streaming analytics • Network optimization and prediction • Intelligent video and multi-sensor processing • Custom datasets, algorithms and training 2
MOVING TOWARDS INTELLIGENT RF SYSTEMS Expert RF Systems, while powerful, do not scale to the IoT time ! Optimal Sample " Time Symbol Duration Time Domain 4 Band-Edge Spectral 5 Frequency component Domain $ % $ & Example of demodulation of Bluetooth signal to extract frequency # " unique header information Symbol Rate '(()*+,- $/0-1+-2ℎ Unique cyclostationary features of digitally modulated signals can be extracted for signal detection and discrimination Tools and theory from DSP can aid and inform the development of AI-powered Intelligent RF systems The RF spectrum is increasingly crowded as devices and networks compete for frequency bands 3
DEEP LEARNING APPLIED TO RF DATA The RF modality presents unique challenges • In-Phase/Quadrature data representation • RF signals contain information about both magnitude and phase • Massive data rates • Many modern communication signals are wideband in order to transmit data at a higher rate (e.g. WiFi 802.11ac is 160 MHz wide) • Datasets • Lack of multiple, comprehensive, freely available datasets compared to other domains (e.g. ImageNet, KITTI, Coco) 4
DEEP LEARNING APPLIED TO RF DATA Signal Detection from Spectrograms • Frequency • RF spectrum usage and interference detection • Time kv Prelude • KickView created the Signal Processing with DIGITS tutorial for NVIDIA Deep Learning Institute • Actual RF signal environment at KickView office 5
DEEP LEARNING APPLIED TO RF DATA OFDM Signal Detection - WiFi IEEE 802.11g (1) Orthogonal Frequency Division Multiplexing (OFDM) is a digital multi-carrier modulation scheme that is employed in many fielded CP OFDM Symbol CP OFDM Symbol CP OFDM Symbol systems; WiFi, cable systems (e.g. DOCSIS 3.1) and cellular networks (e.g. 4G, 5G) either currently deploy or are moving towards OFDM as the PHY-layer standard. t 0 80 samples 80 samples 80 samples t 1 80 samples 80 samples 80 samples t 2 80 samples 80 samples 80 samples 80 samples + 80 samples + 80 samples 80-pt DFT Channelization transforms the raw time-domain Depiction of OFDM carriers in the frequency OFDM signal into a complex signal image domain for IEEE 802.11g, a common WiFi standard 6
DEEP LEARNING APPLIED TO RF DATA OFDM Signal Detection - WiFi IEEE 802.11g (2) Pre-processing maximizes the structure in the Generalizes to real-world RF environments data for feature extraction using deep and does not require explicit time/ learning approaches frequency synchronization 1.1 1 0.9 0.8 PROBABILITY OF DETECTION (PD) 0.7 0.6 0.5 802.11g WiFi 0.4 Bluetooth 0.3 Noise (Full-band) Noise (5MHz Sub-band) 0.2 Noise + BT (Full-band) 0.1 Noise + BT (5MHz Sub-band) Live RF collection at 0 -8 -6 -4 -2 0 2 4 6 8 KickView office SIGNAL TO NOISE RATIO (SNR) 7
DEEP LEARNING APPLIED TO RF DATA kvPrelude RF Dataset Generation Tools Example snapshots Large, representative datasets are required for KickView WiFi Dataset deep learning approaches to be successful OFDM • ~20,000 I/Q snapshots • OFDM (IEEE 802.11g WiFi) • Bluetooth • Noise RF kvPrelude Noise Dataset • Complex signal images • Output of channelizer Bluetooth • Includes time/frequency errors Our patent pending kvPrelude tools can automatically generate labeled RF datasets collected over-the-air for any modern Stay tuned for the public release of our WiFi communications protocol dataset along with others on our website 8
DEEP LEARNING APPLIED TO RF DATA Intelligent Spectrum Monitoring and Anomaly Detection Receiver collection parameters (e.g. dwell time, receiver bandwidth, frequency overlap) all affect a system’s ability Frequency to monitor and evaluate RF bands of interest. am Frequency Receiver Frequency Time Overlap Dwell Time Receiver Maximum Bandwidth Bandwidth Signal Fits into Overlap Region DOCSIS 3.0 cable modem signal interference and anomaly detection using a combination of DSP and deep learning 9
DEEP LEARNING APPLIED TO RF DATA Generalized Modulation Recognition Symbol Mapping We are conducting research to develop a generalized modulation recognition & ⃗ & ⃗ ' " B A recognition scheme using deep learning. # ⃗ $ ! " Complex Sampling ⃗ $ instance(s) C D We train a deep neural network over a sampling of possible combinations of ' "(* waveform trajectories for a short number of observed symbols. ' "(+ & ⃗ ' " = A ' "() Depicted ' "() = C Symbol $ ⃗ ' "(* = B Sequence Key benefits include fast fixed inference ' "(+ = A Trajectory through I/Q space time, robustness to co-channel interference, vs. time for a QPSK and no need for explicit time/frequency modulated signal (4 symbols, synchronization. 5 samples per symbol) 10
DEEP LEARNING APPLIED TO RF DATA Space Applications - RF Emitter Geolocation Differential Delay Differential Doppler Cross Ambiguity Function computes raw TDOA/ FDOA measurements across a sensor pair Θ = TDOA/FDOA estimates Θ # KV-Geo = Satellite Ephemerides Ψ Engine # ⃗ = Geolocation Estimates Ψ $ Deep neural network learns a manifold that maps all ⃗ $ = GDOP possible TDOA/FDOA measurements and sensor configurations to potential geolocation solutions 11
INTELLIGENT MULTI-SENSOR ANALYTICS kv Prelude: Real-Time Intelligent Spectrum Monitoring • kv Prelude is a scalable software solution for real- time RF data acquisition and intelligent spectrum monitoring • Automated spectrum monitoring and analysis • Container-based architecture • Processing at the edge, fog, or cloud • Common software framework for applications • Supports multiple receivers and sample rates • Real-time stream processing and visualization • Automated training data collection and labelling 12
Real-Time AI Systems See a live RF demo at the ADLink Booth (207) Contact Us: krishna.karra@kickview.com 13
Recommend
More recommend
