Sound Spectrum

When you hear someone produce a single vowel, you don't just hear a single frequency. In reality, you hear a combination of waves, each with its own frequency. One of the best ways to visualize this combination of frequencies is to use a sound spectrum. The sound spectrum is often used in phonetic analysis. While some issues pop up when using a sound spectrum for real speech analysis, understanding the definitions of the parts of a sound spectrum can help you understand phonetics as a whole.

Sound Spectrum Definition

Understanding the sound spectrum requires a few key definitions.

The definition of a sound spectrum relies on the concepts of simple and complex waves.

Any complex, repeating wave can be broken down into a series of simple waves. Each simple wave is called a component of the complex wave.

Fig. 1 - All of these simple waves combine to make one complex wave.

You can think of a sound spectrum as a list of all the simple waves that make up a given complex wave.

The Spectrum of a Sound Wave

The spectrum of a sound wave displays frequency on the x-axis and amplitude on the y-axis.

Amplitude is often measured in decibels (dB). Relative amplitude, i.e., how loud one sound is compared to another sound, is most relevant for linguistic analysis.

The above components are all part of the spectrum of a sound wave. So how do we build the sound spectrum for a complex wave? Here are the 4 main steps:

• Break the complex wave down to its simple wave components.

• Measure the frequency of each component.

• List out each component's frequency in a row from lowest to highest.

• Plot this list of frequencies by each wave's relative amplitude.

Luckily, linguistic analysis programs do this work automatically. However, understanding how a program generates a spectrum can help you interpret the spectrum for linguistic analysis. Now let's take a look at frequency in more detail:

Frequency on the Sound Spectrum

The sound spectrum shows the frequency of each component of the complex wave. Frequency is typically measured in hertz (Hz), meaning cycles per second.

The frequency of a complex wave is the number of times the entire complex wave pattern repeats in one second. This is the lowest-frequency component of the wave, also called the fundamental frequency.

On a spectrum, the fundamental frequency is the first complete visible peak. If the first peak on the spectrum has a frequency of 100 Hz, then the fundamental frequency of the entire wave is 100 Hz.

Fig. 2 - The first full "peak" visible on the spectrum (shown by the arrow) is the wave's fundamental frequency.²

Each peak to the right of the fundamental frequency represents a higher-frequency component.

Sound Spectrum Analysis

Sound spectrum analysis can be conducted for a variety of purposes. For example, some scientists conducted sound spectrum analysis on baby birds to compare bird songs both before and after they slept. They found that the baby birds were learning the songs in their sleep!

Here are a few things that are helpful to keep in mind when using a sound spectrum for speech analysis.

• The "peaks" on the spectrum represent the components of the glottal source wave.

• The harmonic frequencies of the glottal source wave components are whole number multiples of the fundamental frequency.

• The wavy amplitude pattern on the spectrum represents vowel formants.

The terms highlighted in bold may be tricky to understand, so let's unpack these:

When you produce a vowel, you push air past the vocal folds in your throat, causing them to vibrate. This vibration is the glottal source wave. Every peak on the sound spectrum of a vowel is a component of the glottal source wave.

Once you know the fundamental frequency of the glottal source wave, you can easily calculate the rest of the harmonic frequencies. A harmonic frequency is a whole number multiple of the fundamental frequency. For example:

The sound spectrum of a vowel looks "wavy" at the top. The high points on this wavy pattern represent frequency ranges with higher amplitude. These high-amplitude ranges are the vowel's formants.

Formants help identify the posture of the vocal tract during different vowels. For example, formant frequencies can distinguish [a] from [i].

Sound Spectrum Issues

Some issues can arise when using a sound spectrum for real speech analysis. The biggest issue is that sound spectra in real speech are more difficult to read than sound spectra in synthesized speech.

A synthesized sound wave can have a neat, clean spectrum with clear and intelligible components. A recording of human speech, on the other hand, consists of imprecise waves and often contains background noise that confuses the program.

Fig. 3 - The smaller peaks on this spectrum represent background noise and imperfections in the speaker's voice.²

The sound spectrum of a natural human vowel tends to look jagged and rough compared to the smooth, uniform spectrum of a computer-synthesized vowel. This is because a human speech recording often comes with background noise, and the human voice doesn't produce even pure tones.

In the spectrum of a human recording, only the largest peaks represent the harmonics of the glottal source wave. The rest are the result of background noise and imperfections in the speaker's voice.

A recording of human speech will never provide a completely pure sound. However, the sound spectrum still serves as a useful tool for speech analysis.