Hilbert Transform for PPG Analysis
The Hilbert transform generates the analytic signal representation of PPG, providing instantaneous amplitude (envelope), instantaneous phase, and instantaneous frequency at every sample point — enabling continuous heart rate estimation with temporal resolution far exceeding FFT-based methods.
The Hilbert transform H{x(t)} computes the imaginary part of the analytic signal z(t) = x(t) + jH{x(t)}, from which instantaneous amplitude A(t) = |z(t)| and instantaneous phase φ(t) = arctan(H{x(t)}/x(t)) are derived. Instantaneous frequency f(t) = (1/2π)·dφ/dt provides a continuous estimate of the local oscillation frequency, which for PPG represents instantaneous heart rate at each sample point.
For PPG heart rate tracking, the Hilbert-derived instantaneous frequency provides higher temporal resolution than FFT (sample-by-sample vs. window-based) and naturally handles frequency transitions without the windowing artifacts inherent in STFT. However, Hilbert instantaneous frequency is meaningful only for narrowband signals. PPG must be bandpass filtered to the cardiac frequency band (0.5–4 Hz) before Hilbert analysis to avoid instantaneous frequency artifacts from waveform harmonics and noise.
The Hilbert-Huang Transform (HHT) combines EMD with Hilbert analysis: EMD first decomposes PPG into narrowband IMFs, then Hilbert analysis of the cardiac IMF provides physiologically meaningful instantaneous frequency. This adaptive approach avoids the fixed bandpass limitation of direct Hilbert analysis and provides robust instantaneous heart rate even during frequency transitions and non-stationary conditions.
Frequently Asked Questions
Does Hilbert transform work for real-time PPG?
The Hilbert transform is non-causal (requires future samples for exact computation). Causal approximations using FIR Hilbert filters (typically 31–127 taps) provide real-time instantaneous frequency with minimal delay, suitable for embedded PPG processing.
How does Hilbert instantaneous frequency compare to peak-based HR?
Hilbert provides continuous inter-peak frequency estimates useful for HRV analysis. Peak-based methods provide discrete beat-to-beat intervals. Hilbert is more sensitive to noise but provides higher temporal resolution between peaks.
What causes artifacts in Hilbert instantaneous frequency?
Broadband signals, sharp transients, and noise produce spurious high-frequency instantaneous frequency values. Bandpass filtering before Hilbert analysis and median filtering of the instantaneous frequency output mitigate these artifacts.