Sound and Perception

Tutorial 1

Waveform: x-axis represents time, y-axis represents amplitude Quiet sounds: everything is flat, doesn’t change much from the center line Loud sounds: the pattern moves far away from the center line (both up and down) Spectrogram: a lot more information than we can see in the waveform! We can see differences between different vowel sounds that we can’t easily see in the waveform The spectrogram has different grayscale shades from white to black Different speech sounds show different patterns of where areas of dark gray appear on the spectrogram

Psychoacoustics

Perception of objective reality is subjective

What you perceive depends on how you sense

Eg 5 five blind men might perceive an elephant as wall, ladder etc

Physical character of sound

Speech is a type of airflow

From airflow to sound

Light travels from point A to point B Sound propagates from point A to point B

Sound: propagation of air pressure fluctuations

Air pressure fluctuations = a wave!

This pattern of compression (high pressure zones: air particles pressed together) and rarefaction (low pressure zones: air particles spread apart) can be modeled as a wave

Air pressure fluctuations propagate to your ear

Three basic dimensions of speech perception

Loudness: “quantity of sound”

Relation between objective and subjective loudness

Matching pressure to loudness

Weber’s Law

The increase in stimulus size necessary to produce a just noticeable difference (JND) is a constant fraction of the stimulus size: ∆S ∝ S ∆S = k · S k = “Weber fraction” JND for perception of loudness ≈ 12%

JND= just noticeable difference

Logarithms: the decibel (dB) scale to the rescue!

Relationship between objective loudness and dB

The goal is met: objective change = subjective change 0 db = just audible 100 db = loud 1 db change = JND for loudness

Matching dB to loudness

Relation between objective and subjective pitch

Timbre relationship between sound and colour

Spectrum to spectrogram

We can view speech over time (speech isn’t frozen in time, after all!) by putting spectra from different time points together in temporal order. This method of viewing speech is called a spectrogram.

A spectrogram provides three dimensions of information: time, frequency, and amplitude. Amplitude is represented by degrees of light-dark gray: the darker the peak, the greater the amplitude.

Spectrum → spectrogram (cont.)