Source-linked AI summary
OFDM-Based Neuromorphic Wake-Up Signal
Gaspard Goupy, Elias Raninen, Dani Korpi
TL;DR
Conventional OFDM waveforms do not naturally provide the spike-like inputs needed for neuromorphic wake-up detection. The paper introduces an OFDM-compatible neuromorphic wake-up signal and event-driven detector; simulations show higher detection probability than OOK baselines with lower computational cost.
Problem
Neuromorphic wake-up detection requires spike-like inputs, which conventional OFDM waveforms do not naturally provide.
Source: §I · p. 1Method
The nWUS encodes a wake-up signature in OFDM amplitude peaks, while a noncoherent detector converts received peaks to spikes and correlates only on spike events.
Source: §I · p. 1Results
Mean detection probability improves by 18.6 and 3.3 percentage points versus comparable- and tenfold-duration OOK baselines, with 2.5× and 43.5× fewer counted operations, respectively.
Source: §I · p. 1Takeaways & Limitations
The sparse waveform concentrates energy into fewer active samples, reducing correlation cost, while retaining OFDM compatibility for neuromorphic wake-up radios.
Source: §VII · p. 6
Abstract
from arXiv · showhide
Wake-up radios (WURs) reduce energy consumption by continuously monitoring the channel for a wake-up signal (WUS) while the main receiver remains asleep. Neuromorphic computing offers a promising approach to ultra-low-power WUS detection through sparse, event-driven processing. However, realizing these benefits requires spike-like inputs, which conventional orthogonal frequency-division multiplexing (OFDM) waveforms do not naturally provide. To enable neuromorphic WUS detection in OFDM-based systems, we introduce a neuromorphic wake-up signal (nWUS) that can be generated by a standard OFDM transmitter. The proposed design superposes frequency-domain vectors with linear phase progressions across the OFDM subcarriers, producing a time-domain waveform with sparse amplitude peaks at predefined positions. These peaks encode the wake-up signature in a spike-like representation that facilitates neuromorphic processing at the receiver. Building on this structure, we then propose a noncoherent neuromorphic detector that encodes received peaks into spikes and performs sparse correlation only upon spike events. Simulations over a 3GPP urban macro channel show that the proposed approach outperforms on-off keying (OOK) baselines in detection probability while achieving up to 43.5x lower computational cost. During idle periods, event-driven processing further reduces correlation operations by 86% relative to nWUS reception.
I. INTRODUCTION
Neuromorphic wake-up radios can process sparse spike events, but conventional OFDM waveforms lack the spike-like inputs they require. This paper introduces an OFDM-compatible wake-up signal and detector, then evaluates detection and computational cost against OOK baselines.
I. INTRODUCTION
- Neuromorphic processing represents information as discrete binary spikes and computes in a sparse, event-driven manner, which can reduce computational cost and energy use.Conventional OFDM waveforms lack this spike-like structure, limiting their direct use with neuromorphic wake-up radios.Source: §I · p. 1
- The nWUS uses spike-like amplitude peaks at configurable positions to encode a wake-up signature for efficient neuromorphic processing.The design is compatible with a standard OFDM transmitter.Source: §I · p. 1
- The proposed noncoherent detector encodes received peaks as spikes and performs weighted-template sparse correlation only on spike events.Fewer generated spikes during idle periods further lower detector activity.Source: §I · p. 1
- 18.6 and 3.3 percentage points higher mean detection probability, with 2.5× and 43.5× fewer counted operations, versus comparable- and tenfold-duration OOK baselines.During idle periods, correlation operations are further reduced by 86% relative to nWUS reception.Source: §I · p. 1
A. Low-Power Wake-Up Signals
Low-power wake-up signals should support noncoherent detection with minimal receiver processing, while remaining compatible with standard OFDM transmitters. The background and system model describe the neuromorphic waveform mismatch and the time-domain transmission and reception setting.
A. Low-Power Wake-Up Signals
- OOK is commonly used for wake-up signals, but sequential ON/OFF symbols can produce long signals and dense receiver processing; combining energy concentration with sparsity to lower processing costs had not been investigated.Neuromorphic studies mainly considered OOK, while IR-UWB offers sparse impulses but is not compatible with conventional OFDM transmitters.Source: A. Low-Power Wake-Up Signals · p. 2, B. Neuromorphic Wake-Up Radios · p. 2
B. Neuromorphic Wake-Up Radios
- Neuromorphic wake-up radios need spike-like inputs, but conventional OFDM waveforms do not naturally provide them.The section identifies OFDM-based sparse, spike-like waveform design for neuromorphic detection as unexplored to the authors' knowledge.Source: B. Neuromorphic Wake-Up Radios · p. 2
III. BACKGROUND
- The paper's notation uses one-based vector-entry and FFT-bin indices, zero-based detection lags, and j for the imaginary unit.Source: §III · p. 2
A. System Model
- The system model maps a WUS onto an OFDM symbol, applies an IFFT to subcarrier coefficients, and searches for the known signal in the received time-domain baseband waveform.It omits the cyclic prefix for simplicity and models reception through a time-varying finite-impulse-response channel with additive noise.Source: A. System Model · p. 2
B. Problem Formulation
The primary task is identifying a target wake-up signal among valid signals with different signatures. Detection is evaluated per attempt over a lag search, with a separate noise-only false-alarm setting also considered.
B. Problem Formulation
- The primary hypothesis test distinguishes the target WUS from a valid non-target WUS with a different signature, using the maximum score across searched lag windows.The decision threshold is calibrated for per-attempt detection and false-identification probabilities.Source: B. Problem Formulation · p. 2
- A separate noise-only hypothesis supports evaluation of the per-attempt false-alarm probability.Under this hypothesis, the received signal consists of noise.Source: B. Problem Formulation · p. 3
C. On-Off Keying Baseline
The OOK baseline represents a wake-up signature as binary ON/OFF intervals and detects it by correlating normalized bit-energy measurements with the target sequence.
C. On-Off Keying Baseline
- An OOK-WUS uses a binary signature, with each bit represented by an active ON or inactive OFF interval spanning Lb time samples.Its waveform has length Ltx = NLb samples.Source: C. On-Off Keying Baseline · p. 3
- The receiver integrates squared amplitude over sliding Lb-sample intervals to form an energy sequence, then assembles bit-energy vectors for each lag.Source: C. On-Off Keying Baseline · p. 3
- For each lag, the receiver mean-centers and unit-normalizes the bit-energy vector, correlates it with the target OOK sequence, and applies the detection decision.Source: C. On-Off Keying Baseline · p. 3
IV. NEUROMORPHIC WAKE-UP SIGNAL
The nWUS represents a wake-up signature as sparse time-domain peaks while retaining compatibility with conventional OFDM transmitters. Its frequency-domain coefficients are chosen to place peaks at predefined sample positions.
A. Signature Definition
- An nWUS signature is a length-F sequence with Ns spikes at predefined indices, separated by at least Δmin.The signature is fixed and known to the WUR; different WURs can be assigned different signatures.Source: A. Signature Definition · p. 3
B. Waveform Construction
- Linear phase progressions across active subcarriers align contributions at each target sample, and summing the per-spike vectors forms one OFDM symbol.The coefficients can equivalently be obtained by taking an F-point FFT of the spike sequence and retaining active-band bins.Source: B. Waveform Construction · p. 3
- The OFDM IFFT produces peaks at the predefined positions; with all F subcarriers active, it exactly reconstructs the spike sequence, while D < F broadens peaks into Dirichlet-kernel pulses.A sufficiently large Δmin helps keep peaks distinguishable; DC nulling adds deterministic distortion.Source: B. Waveform Construction · p. 3
- The nWUS changes only the frequency-domain coefficients supplied to the OFDM modulator, so it requires no transmitter changes.A cyclic prefix is optional; when included, peak centers should lie outside the copied prefix samples.Source: B. Waveform Construction · p. 3
C. Information Capacity
An nWUS can encode information through both the number and positions of its spikes, subject to a minimum-spacing constraint.
C. Information Capacity
- One nWUS symbol can carry at most ⌊log2 Nseq⌋ bits, where Nseq counts feasible spike sequences with fixed Ns and minimum spacing Δmin.Feasibility requires F ≥ 1 + (Ns − 1)Δmin; usable capacity may be lower because of shifted-signature ambiguity or channel impairments.Source: C. Information Capacity · pp. 3–4
V. NEUROMORPHIC DETECTOR
The detector converts received amplitude peaks into sparse spike events before further processing. Its fixed-threshold encoder trades detection robustness against the number of generated spikes and requires threshold calibration.
V. NEUROMORPHIC DETECTOR
- The neuromorphic detector first converts received amplitude peaks into spike events, creating a sparse representation for subsequent processing.This is intended to reduce subsequent computational cost.Source: A. Spike Encoding · p. 4, §V · p. 4
A. Spike Encoding
- A fixed-threshold level-crossing encoder compares each received sample with η, with η selected so the empirical mean spike count under H1 matches budget K.The threshold must detect nWUS peaks while avoiding excessive spikes during idle periods; K > Ns can improve robustness.Source: A. Spike Encoding · p. 4
- The proposed nWUS and correlator are encoder-agnostic, but this encoder relies on absolute amplitude levels and requires η calibration before operation.Adaptive encoders could reduce calibration requirements and improve robustness to changing conditions.Source: A. Spike Encoding · p. 4
B. Sparse Correlator
The sparse correlator scores spike-encoded windows against a weighted template for each possible arrival lag. Its weights decay between expected spike positions to tolerate local timing distortions.
B. Sparse Correlator
- Because only active spike positions contribute, correlator computation scales with the number of spikes rather than the window length.The maximum score across the lag search is retained for detection.Source: B. Sparse Correlator · p. 4
- The template assigns each expected spike a maximum weight that decays toward a minimum value until the next spike position.The decay constant controls the trade-off between temporal robustness and selectivity.Source: B. Sparse Correlator · p. 4
- Weight decay tolerates local timing distortions such as band-limiting and multipath smearing, so detection can remain discriminative despite missed, shifted, or extra spikes.The lag search handles global delay, distinct from the template's local tolerance.Source: B. Sparse Correlator · p. 4
VI. SIMULATION RESULTS
The simulations compare the nWUS with two OOK-WUS baselines under a shared urban-macro channel and evaluation protocol. The setup uses both a comparable-airtime baseline and a tenfold-longer baseline.
A. Setup
- The nWUS uses one 66.67 µs OFDM symbol with five randomly positioned spikes, at least 20 samples apart, and an encoder budget of 14 spikes per observation.Its FFT size is 1024 with 870 active subcarriers; parameters were selected by random search on independent data.Source: A. Setup · p. 4
- The OOK baselines use 1024 bits over 666.67 µs and 42 bits over 68.36 µs, respectively, to provide tenfold-longer and comparable-airtime comparisons.The 42-bit baseline was chosen to approximately match nWUS signature capacity; both use the same sampling rate as nWUS.Source: A. Setup · pp. 4–5
- All waveforms are evaluated at equal average transmit power in a 3.5 GHz SISO 3GPP urban-macro channel with complex AWGN and 60 searched lags.Channel realizations randomize user location, height, indoor/outdoor status, and speed; no additional random timing offset is introduced.Source: A. Setup · p. 5
- Detection thresholds are calibrated to a target false-identification probability of 10^-2, and reported results average five random target-signature seeds.The nWUS encoder threshold is calibrated separately to match its mean spike budget before evaluation.Source: A. Setup · p. 5
B. Detection Performance
Under equal average transmit power, the nWUS improves detection probability over both OOK baselines. Its sparse correlator also reduces counted operations, particularly during noise-only periods.
B. Detection Performance
- At equal average transmit power, nWUS gains 18.6 pp in mean Pd over similar-duration 42-bit OOK-WUS and 3.3 pp over tenfold-longer 1024-bit OOK-WUS.With the threshold calibrated to Pfi = 10−2, nWUS reaches the target Pd near −8 dB, versus near 0 dB for 1024-bit OOK-WUS; 42-bit OOK-WUS remains below target throughout the evaluated SNR range.Source: Figure 2 · p. 5, B. Detection Performance · p. 5
- Under equal average transmit power, nWUS reaches the target detection probability at approximately −8 dB SNR, while 42-bit OOK-WUS remains below that target throughout the evaluated SNR range.The decision threshold is calibrated to achieve Pfi = 10−2.Source: B. Detection Performance · p. 5, Figure 2 · p. 5
C. Computational Cost
- 2.5× and 43.5×: at comparable airtime and information capacity, the 42-bit OOK-WUS requires 2.5× the nWUS counted operations; the 1024-bit baseline requires 43.5×.The cost estimates count encoder preprocessing and correlation operations; they are coarse proxies for computational cost and receiver energy because memory access and implementation-dependent costs are omitted.Source: Table I · p. 6, C. Computational Cost · p. 6
- 86% fewer correlation operations: during noise-only periods, the nWUS correlator performs approximately 86% fewer operations than under target reception.The OOK correlator cost remains the same regardless of received activity.Source: C. Computational Cost · p. 6
VII. CONCLUSION
The nWUS concentrates energy into sparse, localized temporal peaks while retaining OFDM compatibility, supporting detection and lower correlation cost. Practical realization still requires hardware investigation, including assessment of peak effects on power amplifiers.
VII. CONCLUSION
- Sparse, localized active samples concentrate transmit energy, improve detectability, and reduce correlation cost; event-driven processing further reduces cost during idle periods.The approach retains OFDM compatibility and may also benefit existing neuromorphic WUR architectures that use spike-based WUS signatures.Source: §VII · p. 6
- The impact of high amplitude peaks on power-amplifier operation remains unassessed, and practical realization requires further study of hardware constraints and integration.Multiplexing nWUS with user data on unused subcarriers is also identified as future work to improve spectral efficiency.Source: §VII · p. 6