Technology
Time-Frequency Analysis
Quick fact
A spectrogram, a visual output of time-frequency analysis, can reveal the exact words spoken in a recording even when the signal is buried in noise, which is why it is used in forensic audio analysis.
Why this is interesting
When you listen to a song, you hear notes changing over time—how can a single mathematical tool capture both the pitch (frequency) and its rhythm (time)?
Read the full explanation
Understanding Time-Frequency Analysis
Imagine you have a recording of a person speaking. Traditional frequency analysis (Fourier transform) tells you the overall mix of frequencies in the whole recording, but it loses the order of words. Time-frequency analysis lets you see how the frequency content changes moment by moment. Think of a musical score: time runs left to right, and each note's pitch rises or falls. The most common tool is the Short-Time Fourier Transform (STFT), which chops the signal into small overlapping windows and applies the Fourier transform to each window. Stacking these results over time produces a spectrogram—a graph with time on the x-axis, frequency on the y-axis, and intensity (color) representing amplitude. It is like watching the signal's 'fingerprint' evolve.
A deeper explanation
The core challenge in time-frequency analysis is the trade-off imposed by the uncertainty principle: you cannot know both the exact time and exact frequency of a signal simultaneously. High time resolution (short windows) gives blurry frequency details, while high frequency resolution (long windows) smears time events. Different methods navigate this trade-off: the STFT uses a fixed window length, while the wavelet transform uses variable window sizes—short for high frequencies, long for low frequencies—mimicking how the human ear processes sound. This makes time-frequency analysis vital for analyzing non-stationary signals (which change over time), such as speech (formants shift), music (instrument attacks), earthquakes (arrival of different waves), and radar echoes (Doppler shifts). By revealing how energy distributes over time and frequency, it turns raw data into actionable insight, enabling applications like speech recognition, noise cancellation, and medical diagnostics (e.g., EEG seizure detection).