1 Definition and purpose

A scalogram is a graphical display of a signal’s time-frequency content, usually produced from a wavelet transform. It shows how signal strength, energy, or power varies across both time and scale, which is often interpreted as a frequency-like quantity. By presenting localized changes in a two-dimensional form, a scalogram helps reveal features that may be hidden in a conventional time plot.

1.1 Basic concept

At its simplest, a scalogram maps a one-dimensional signal onto a surface or image whose axes represent time and scale or frequency. Areas of stronger color or higher intensity indicate larger wavelet coefficients, suggesting that the signal contains more energy at those points. This makes the display useful for signals that change rapidly or irregularly.

1.2 Relationship to wavelet analysis

Scalograms are closely associated with wavelet analysis, especially the continuous wavelet transform. The transform compares a signal with shifted and dilated copies of a chosen wavelet, producing coefficients that describe similarity at each time and scale. The scalogram is commonly constructed from the magnitude or squared magnitude of these coefficients.

1.3 Use in signal interpretation

In practice, scalograms are used to interpret nonstationary signals, meaning signals whose statistical properties change over time. They can highlight short-lived events, repeated structures, and changes in oscillatory behavior. As a result, they are widely applied in scientific measurement, engineering diagnostics, and biomedical analysis.

2 Mathematical background

2.1 Time-frequency representation

A scalogram belongs to the broader family of time-frequency representations. Unlike a simple Fourier spectrum, which summarizes a signal over the full observation interval, a time-frequency display preserves information about when particular components occur. This is especially valuable for signals that contain brief pulses, drifting frequencies, or intermittent oscillations.

2.2 Continuous wavelet transform

The most common basis for a scalogram is the continuous wavelet transform, which expresses a signal as the result of correlation with wavelets of different sizes and positions. Each transform coefficient reflects how well the signal matches the wavelet at a specific time and scale. Because the wavelet is localized, the method captures both duration and characteristic pattern.

2.2.1 Wavelet coefficients

Wavelet coefficients are numerical values obtained from the transform. Large coefficients indicate a strong correspondence between the signal and the wavelet at a given location. In a scalogram, these coefficients are often converted into a visual intensity measure, such as absolute value or squared magnitude.

2.2.2 Scale and frequency

In wavelet analysis, scale describes the width of the wavelet. Small scales correspond to narrow wavelets and are associated with higher-frequency content, while large scales correspond to broader wavelets and lower-frequency behavior. Although scale is not identical to frequency, it is often converted to an equivalent frequency axis for display and interpretation.

2.3 Energy or power mapping

A scalogram is commonly formed by plotting the energy or power associated with the wavelet coefficients. This can be done by taking the magnitude squared of the transform, which emphasizes stronger features and reduces the effect of sign changes. The result is an image that resembles a heat map of signal activity across time and scale.

3 Construction of a scalogram

3.1 Signal acquisition

The process begins with a measured or recorded signal, such as data from a sensor, microphone, biomedical electrode, or seismic instrument. The signal is sampled over time and may be preprocessed to remove offsets, noise, or artifacts. Accurate acquisition is important because the scalogram reflects the quality of the input data.

3.2 Choice of wavelet function

A wavelet function must be selected before computation begins. Different wavelets emphasize different patterns, such as sharp spikes, smooth oscillations, or compact bursts. The choice depends on the signal characteristics and the features the analyst wants to detect.

3.3 Computation of coefficients

Once the wavelet is chosen, the continuous wavelet transform is computed over a range of times and scales. This produces a matrix of coefficients that can be analyzed numerically or displayed visually. Dense sampling of scales yields a more detailed image, though at greater computational cost.

3.4 Visualization methods

The coefficient matrix is displayed as an image or contour plot, with colors or shades representing amplitude or power. The resulting view may be scaled linearly or logarithmically depending on the application. Proper visualization helps make subtle variations easier to interpret.

3.4.1 Color mapping

Color mapping assigns different colors to different coefficient magnitudes. Dark or cool colors may represent low values, while bright or warm colors indicate strong activity. The exact palette is a matter of design, but it should preserve contrast and avoid misleading visual effects.

3.4.2 Axis conventions

Time is usually placed on the horizontal axis, while scale or frequency appears on the vertical axis. Some displays invert the vertical axis so that higher frequencies appear near the top, matching common spectral conventions. Labels and units are important for reading the figure correctly.

4 Features of scalograms

4.1 Time localization

One major feature of a scalogram is precise timing information. It can show when a sudden event begins, how long it lasts, and whether it repeats. This temporal localization is one of the main reasons wavelet-based displays are preferred for transient signals.

4.2 Frequency localization

A scalogram also reveals which scales or frequencies are active at a given moment. This helps distinguish broad low-frequency trends from narrow high-frequency oscillations. While localization is not perfect at all scales, the method offers a flexible compromise between time and frequency detail.

4.3 Amplitude intensity patterns

The intensity pattern in a scalogram often reflects the relative strength of different signal components. Bright ridges, patches, or bands may correspond to dominant oscillations or bursts of energy. Analysts use these patterns to track changes in signal structure across an observation period.

4.4 Multiscale structure

Scalograms naturally display the same event at multiple scales. A single physical occurrence may appear as a short feature at high frequency and a broader structure at low frequency. This multiscale view can expose hierarchical patterns that are difficult to detect with single-resolution methods.

5 Applications

5.1 Mechanical and vibration analysis

In mechanical systems, scalograms are used to study vibrations, rotating machinery, and structural responses. They can help identify impacts, bearing faults, resonances, and changes in operating conditions. Because many mechanical signals are intermittent and nonstationary, wavelet displays are especially useful.

5.2 Biomedical signal processing

Biomedical signals often contain brief, irregular, and overlapping features. Scalograms provide a convenient way to inspect these patterns and support tasks such as classification, event detection, and artifact recognition.

5.2.1 Electrocardiography

In electrocardiography, scalograms can emphasize the timing and shape of cardiac components. They may assist in detecting arrhythmias, abnormal beats, or noise contamination. The method is valuable when short-lived changes are more important than long-term averages.

5.2.2 Electroencephalography

In electroencephalography, scalograms are used to study brain rhythms, bursts, and transient events. They can show how oscillatory activity changes during sleep, stimulation, or seizure-like episodes. Their ability to present time-varying frequency content makes them useful in neurological analysis.

5.3 Seismology and geophysics

Seismic signals often contain waves arriving at different times and frequencies. Scalograms can help identify arrivals, distinguish phases, and observe local changes in energy. They are useful for examining short events embedded in longer background fluctuations.

5.4 Communications and acoustics

In communications and acoustics, scalograms are applied to modulated signals, speech, machinery noise, and environmental sound. They can reveal bursts, chirps, and changing spectral envelopes. This makes them helpful in feature extraction and pattern recognition tasks.

6 Interpretation

6.1 Identifying transients

Transient events appear in a scalogram as localized regions of elevated intensity. These may correspond to impulses, sudden onsets, or abrupt changes in waveform shape. Their compact appearance makes them easier to spot than in a standard time trace.

6.2 Detecting bursts and anomalies

Repeated bursts often form clusters or ridges in certain scale ranges. Unusual shapes, isolated hot spots, or unexpected scale patterns may indicate anomalies. Analysts use these cues to compare normal and abnormal behavior in recorded signals.

6.3 Comparing signal events across scales

Because the display contains multiple scales at once, it allows direct comparison of how the same event behaves at different resolutions. Some features remain visible over a wide band of scales, while others are confined to a narrow region. This comparison can help determine whether a signal component is broad, fine-grained, persistent, or brief.

7 Advantages and limitations

7.1 Advantages over Fourier-based methods

Compared with Fourier-based methods, scalograms preserve time information. This is a major advantage when the signal is not stationary and when the timing of events matters. They also offer flexible resolution, giving better detail for some features than a single fixed-frequency view.

7.2 Resolution trade-offs

A scalogram must balance time and frequency detail. Narrow wavelets improve time localization but reduce frequency precision, while wider wavelets do the opposite. This trade-off is inherent to time-frequency analysis and affects how the display should be interpreted.

7.3 Sensitivity to wavelet choice

The appearance of a scalogram depends strongly on the chosen wavelet. A wavelet well matched to the signal can highlight important structures, while a poor match may obscure them. For that reason, interpretation often involves testing several wavelet types or parameter settings.

8.1 Spectrograms

A spectrogram is another time-frequency display, usually based on the short-time Fourier transform. It also shows how spectral content changes over time, but it uses fixed window sizes rather than scale-dependent wavelets. Spectrograms and scalograms are often compared because both visualize evolving frequency content.

8.2 Wavelet transforms

Wavelet transforms are the mathematical foundation of many scalograms. They include both continuous and discrete forms, each suited to different analysis tasks. The continuous form is commonly used for visualization, while discrete forms are often used for compact data representation and algorithmic processing.

8.3 Time-frequency analysis tools

Scalograms are one of several tools used to study signals whose properties vary over time. Other methods include Wigner-type distributions, reassigned time-frequency plots, and adaptive decomposition techniques. Each approach offers its own balance of interpretability, precision, and computational complexity.