Scalogram

The vertical axis of a scalogram represents scale, which is inversely proportional to frequency. Small scales correspond to high frequencies (compressed wavelets) and capture rapid transients, while large scales correspond to low frequencies (stretched wavelets) and capture slow oscillations. This logarithmic relationship provides high temporal resolution at high frequencies and high frequency resolution at low frequencies, a direct consequence of the Heisenberg-Gabor uncertainty principle applied to wavelet analysis.

The scalogram displays the absolute value or squared magnitude of the complex CWT coefficients as a color intensity or contour map. High-magnitude regions (bright colors) indicate a strong correlation between the wavelet and the signal at that specific time and scale, effectively localizing energy bursts. This is distinct from a spectrogram, which plots power spectral density. Common colormaps like jet or parula are used, though perceptually uniform colormaps are preferred to avoid introducing visual artifacts that could be misinterpreted as signal features.

The local maxima of the scalogram's magnitude form continuous curves called time-frequency ridges. These ridges trace the instantaneous frequency of signal components and are critical for feature extraction in RF fingerprinting. By isolating the ridge associated with a transmitter's turn-on transient or a specific modulation-induced artifact, engineers can extract a one-dimensional signature from the two-dimensional representation. Algorithms like crazy climbers or simple peak-picking along the scale axis are used for automated ridge extraction.

The visual appearance and analytical utility of a scalogram are fundamentally determined by the chosen mother wavelet. A Morlet wavelet offers optimal joint time-frequency localization and produces a visually intuitive scalogram that closely resembles a spectrogram. In contrast, a Mexican Hat wavelet (Ricker wavelet) is better suited for detecting sharp discontinuities and peaks. For RF fingerprinting, the Generalized Morse Wavelets are often preferred because they provide precise control over time-frequency concentration, allowing the analysis to be tuned to match the expected transient duration of a specific device's impairments.

A scalogram's multi-resolution property is its key differentiator from a spectrogram. A spectrogram, derived from the Short-Time Fourier Transform (STFT), uses a fixed window length, resulting in a uniform time-frequency resolution grid. The scalogram, however, adapts its effective window length: it uses a short time window for high frequencies to precisely locate transients and a long time window for low frequencies to accurately resolve steady-state oscillations. This makes the scalogram superior for analyzing signals containing both short-duration, high-frequency bursts and long-duration, low-frequency drifts—a common characteristic of RF emissions.

Every scalogram exhibits edge effects near the boundaries of the time axis, where the wavelet extends beyond the signal's start or end. The region where these boundary effects are significant is demarcated by the Cone of Influence (COI) . Within the COI, coefficient magnitudes are artificially attenuated and unreliable for feature extraction. In RF fingerprinting, it is critical to either discard data within the COI or use signal extension techniques like zero-padding, symmetric extension, or predictive extrapolation to mitigate these artifacts before computing the scalogram.

The mathematical engine that generates a scalogram. The CWT convolves a signal with a scaled and translated mother wavelet across a continuous range of scales, producing a two-dimensional coefficient matrix. Unlike the STFT, the CWT uses a variable window length—short windows for high frequencies and long windows for low frequencies—providing multi-resolution analysis. The scalogram is simply the squared magnitude of these CWT coefficients.

The Fourier-based counterpart to the scalogram. A spectrogram visualizes the squared magnitude of the Short-Time Fourier Transform (STFT), plotting frequency against time. The key distinction is resolution: the spectrogram uses a fixed window size, resulting in a uniform time-frequency tiling. In contrast, a scalogram's time-frequency cells scale with frequency, offering superior temporal resolution for high-frequency transients and better frequency resolution for low-frequency oscillations.

The most common mother wavelet used to compute a scalogram for RF fingerprinting. It consists of a complex exponential carrier modulated by a Gaussian envelope. Its shape closely matches the transient and oscillatory nature of communication signals. The Morlet wavelet provides an optimal balance between time and frequency localization, making scalograms generated with it highly effective for visualizing transient signal analysis and subtle hardware impairments.

A post-processing technique that sharpens a scalogram by reassigning the CWT coefficients. Standard scalograms can suffer from energy smearing along the scale axis. The SST calculates the instantaneous frequency at each point and squeezes the energy to its true central ridge. This produces a highly concentrated, denoised time-frequency representation that makes individual signal components—and their subtle distortions—far easier to identify and track.

A continuous curve tracing the local maxima of a scalogram, corresponding to the instantaneous frequency of a signal component. In RF fingerprinting, ridge extraction is critical for isolating the precise frequency trajectory of a transmitter's turn-on transient or a specific modulation pulse. Deviations in these ridges from an ideal path reveal the unique hardware impairments that form the device's fingerprint.

A quadratic time-frequency distribution offering the highest possible resolution, computed by Fourier transforming the signal's instantaneous autocorrelation. While a scalogram provides a smoothed energy distribution, the WVD provides pinpoint concentration for linear chirps. However, for multi-component RF signals, the WVD suffers from severe cross-term interference—spurious artifacts that obscure the true signal components—making the scalogram a more robust choice for complex emitter analysis.