The envelope curve (often called simply the envelope) of a signal is a smooth curve that outlines the extremes of the waveform – it connects the peaks (or the troughs) and describes how the overall amplitude of the signal varies over time. In essence, the envelope captures the shape of the signal's energy, ignoring the rapid oscillations inside.
Think of it like the outline of a mountain range against the sky – you don't see every tree and rock, but you see the grand contour. Similarly, the envelope of an audio signal tells you when the sound is loud, when it fades, and how its intensity changes, without being distracted by every single vibration.
Figure 1: A 200 Hz sinusoidal carrier (blue) amplitude-modulated by a slowly varying envelope (red, dashed). The envelope smoothly follows the peaks of the waveform, revealing the loudness contour.
Mathematically, the envelope \( e(t) \) of a signal \( x(t) \) is most commonly obtained using the analytic signal via the Hilbert transform. For any real-valued signal \( x(t) \), the analytic signal is:
where \( \mathcal{H}\{\cdot\} \) is the Hilbert transform. The envelope is then the magnitude (absolute value) of this complex analytic signal:
For simpler cases, the envelope can also be approximated by rectification (taking the absolute value) followed by a low-pass filter – a method often used in analogue audio circuits. Another common approach is to detect the peaks by interpolation, especially in digital synthesis.
The envelope is crucial because it separates the "fast" content (carrier frequency, timbre) from the "slow" content (loudness, articulation, dynamics). In music, the classic ADSR (Attack, Decay, Sustain, Release) model is nothing but a piecewise-linear envelope that shapes the volume of each note.
How the envelope is used
- Audio synthesis and effects: Envelopes control the amplitude of each note (ADSR), the filter cutoff, or the modulation depth.
- Amplitude modulation (AM) and demodulation: In radio, the envelope of the modulated signal carries the original audio message.
- Speech processing: The envelope of speech reveals stress patterns, rhythm, and is used in voice activity detection and speech enhancement.
- Dynamic range compression: Audio compressors detect the envelope to reduce gain when the signal gets too loud.
- Music information retrieval: Envelope features help in beat tracking, genre classification, and onset detection.
- Biomedical signals: In ECG and EEG analysis, the envelope helps extract the modulation of neural or cardiac rhythms.
In the time domain, the envelope is often slow-varying compared to the carrier. For example, in the plot above, the carrier is 200 Hz (oscillating 200 times per second), while the envelope changes at about 1–2 Hz – much slower. This separation of time scales is the foundation of many signal processing techniques.
It is important to distinguish the amplitude envelope from the instantaneous frequency – the envelope tells you how much energy is there, while instantaneous frequency tells you where that energy is located in frequency. Together, they form the two components of the analytic signal: magnitude (envelope) and phase (from which frequency is derived).
One of the most elegant properties of the Hilbert transform is that it yields a canonical envelope – it is the minimal envelope that is both smooth and non-negative, and it is orthogonal to the instantaneous phase. In practice, though, the Hilbert transform is sensitive to noise and works best for narrowband signals. For broadband signals (like speech), other methods like the Teager–Kaiser energy operator or filter-bank envelopes are used.
In modern audio processing, the envelope is often computed in mel-frequency bands – this gives a set of envelope signals, each corresponding to a different perceptual frequency region. These are the foundation of features like modulation spectrograms and reverberation time estimation.
Despite its power, the envelope is not a complete description – it discards the fine structure of the waveform, which is essential for pitch perception and timbre. For instance, two signals with the same envelope but different fine structure can sound completely different (e.g., a violin vs. a flute playing the same note). That is why many audio systems combine envelope and spectral features.
The envelope curve is one of the most intuitive yet mathematically rich concepts in signal processing. It bridges the gap between the physical waveform and our perceptual experience of loudness and dynamics. From vintage analogue synthesizers to cutting‑edge deep learning models for audio, the envelope remains a fundamental tool for understanding and shaping sound.