The Spectrograph & Spectral Analysis

There are two main kinds of voice analysis performed by the spectrograph, wideband (with a bandwidth of 300-500 Hz) and narrowband (with a bandwidth of 45-50 Hz).

Wideband Spectrograms
Wideband (a.k.a. 'broadband') spectrograms, when used for normal speech with a fundamental frequency of around 100-200 Hz, will pick up energy from several harmonics at once and add them together, as seen below. The Fo (fundamental frequency) can be determined from the graphic by counting the number of individual vertical lines per unit time. Also, the frequencies and relative strengths of the first two formants (F₁ and F₂) are visible as dark, rather blurry concentrations of energy.

The wide bandwidth in this type of analysis allows for excellent time resolution; you can see the energy peaks from each individual vibration of the vocal folds in the graph. However, the wide bandwidth also means that individual harmonics cannot be singled out by this kind of spectrogram. This means that for the wideband spectrograph, frequency components within a 300-500 Hz bandwidth are not easily distinguished.

Narrowband Spectrograms
The narrowband spectrogram has different strengths; it is able to pick out each individual harmonic, unlike the wideband spectrogram, but its time resolution is not good enough to isolate each individual cycle of vibration, and the formant structure of the sound is not rendered as clearly as with a wideband analysis. Note the dark horizontal stripes, representing each harmonic, in the graphic below. Also note that the large clusters of formant energy which we saw in the wideband spectrogram are not present:

Although the spectrograph is often used in research, it can also serve as a helpful tool in the voice clinic to provide feedback as part of a voice therapy or training program.

Tutorials Homepage