Course contentsShow
← Machine Learning and Deep Learning
Lesson 2438 of 3,53853. Speech and Audio ProcessingFree lesson

Spectrograms: Visualizing Audio

Creating and interpreting spectrograms as 2D representations showing frequency content over time.

Spectrograms: Visualizing Audio

You'll learn how spectrograms transform audio from a one-dimensional signal into a two-dimensional "picture" that reveals both frequency and time information.

What Is a Spectrogram?

A spectrogram is a visual representation of audio that shows how the frequency content of a signal changes over time. Think of it as a heat map: the x-axis represents time, the y-axis represents frequency (pitch), and the color intensity shows the amplitude (loudness) of each frequency at each moment.

From STFT to Spectrogram

Remember the Short-Time Fourier Transform (STFT) you learned earlier? It splits audio into overlapping windows and computes the frequency spectrum for each window. A spectrogram is simply a 2D visualization of all those STFT results stacked together:

  • Each vertical slice represents one window's frequency spectrum
  • Brighter/warmer colors indicate stronger frequency components
  • Darker/cooler colors show weaker or absent frequencies

Why Spectrograms Matter

Unlike raw waveforms that only show amplitude over time, spectrograms reveal patterns in frequency space. This is crucial because:

  • Human speech has distinctive frequency patterns (formants)
  • Musical notes appear as horizontal bands at specific frequencies
  • Noise, silence, and different sounds have visually distinct "signatures"
  • Deep learning models (like CNNs) can process spectrograms as images

Think of it like this: a waveform is like reading sheet music as a single line of numbers, while a spectrogram is like seeing the full musical score with all instruments visible at once.

Key Takeaway: Spectrograms convert audio into 2D time-frequency representations by visualizing STFT outputs, making frequency patterns visible and enabling image-based deep learning approaches for audio tasks.