How can a picture reveal the changing sounds of speech?

A recording of speech can become a picture of changing sound frequencies. A spectrogram places time horizontally and frequency vertically. Its shade or color represents signal strength at each time–frequency location under a stated scale. A dark region is not a physical dark patch in the speaker’s throat, and vertical position is not loudness.

The analysis takes a short window of recorded samples, estimates its frequency content and repeats the operation as the window moves. Stacking these spectra in time reveals changing acoustic structure. The window introduces a tradeoff: a longer duration can distinguish nearby frequencies more finely, while a shorter duration can localize rapid changes more precisely. The display is an analysis of a recording, not a direct image of moving organs.

Voiced speech often contains harmonics spaced approximately by the vocal-fold repetition frequency, F0. The vocal tract shapes their strengths through resonances called formants. F1 and F2 name the first two prominent resonance regions; they are not simply the first and second harmonics of F0. Changing tongue and lip configuration can move formants while a similar repetition rate continues.

The illustration uses invented resonance tracks at 500 and 1,500 Hz that later shift to 700 and 1,200 Hz. It also shows separate harmonic guides spaced by 100 or 200 Hz. These guides distinguish source spacing from resonance locations; they are not measured speech or outputs of an implemented short-time transform. Real vowel interpretation depends on speaker, accent and context, so these numbers do not uniquely label words or speakers. A spectrogram makes acoustic evidence visible, but interpreting language still requires more than reading a colored stripe.