akutek.info

The www center for search, research and free sharing in acoustics

Welcome to the akuTEK voice acoustics section. ( Note: All linked separate windows are resizable)

The magic first seconds when a child is born, everyone present holds their breath, only waiting for its first vocal performance—the cry of life.  The importance of the sound from the human voice remains throughout life.

Corresponding to speech and singing, voice acoustics may be divided into Speech acoustics and Singing acoustics.

The sounds produced in speech and singing is a combination of vowels—tones produced by the vocal cords and filtered by the upper vocal tract, and consonants—transients and noise bursts produced and modulated by the upper vocal tract.

The acoustics of vowels and the temporal development of the power spectrum from the voice is often studied in a so called spectrogram, which is a common diagram resulting from voice analysis.

Examples:  [ i ] as in ‘heed’ (narrowband); [u] as in ‘mood’ (broadband), revealing the singer’s formant. Four vowel spectra aligned vertically (500k). Individual diagrams: [ i ] , [ æ ] , [ a ] , [ o ]. The most powerful vowel appears to be [ æ ].

Formants (formant frequencies) are a set of natural frequencies in the vocal tract, being unique physical properties associated with each vowel. As the tongue moves to different positions, the shape of the vocal tract changes, and thus the formant frequencies change. The formant frequencies appear as peaks in the power spectrum, having bandwidths in the range of a few hundred Hertz, and can most easily be identified in a spectrogram. The transition through a series of vowels leaves a sliding trace for each of the formant frequencies. In particular, a continuous sequence of whispered vowels leaves a clear spectrogram, since the sound source does not have harmonics disturbing the picture.

Since the tongue can be moved continuously between different positions, the formant frequencies changes continuously within their individual frequency ranges. Every language has a unique set of tongue positions (and movements) and thus a set of formant frequencies (and frequency sweeps) associated with each voiced sound. The lowest formant ( 1. formant) and the second lowest formant (2. formant) have quite wide frequency ranges, 300-1200Hz for 1. formant and 800-3000Hz for 2. formant. These are often referred to as moving formants, in contrast to the fixed formants in the resonating bodies, air-columns or cavities of some musical instruments. Higher order of formants have considerable narrower range. Though each formant can change its frequency continuously, it can not change independently of the other formants. Rather, they appear in certain formant combinations.

The time varying frequency content in fluent speech can be studied in a speech spectrogram (Thanks to Dr Bill Jones for this brilliant example). Voiced consonants often adopts the formant structure of neighboring vowels, or form transitions between the formant structure before and after the consonant. The spectrogram will show less high frequency content if the vocal tract is partly obstructed during the consonant sound, e.g. by l, r and the nasal sounds m and n. Un-voiced fricatives like f and s cover a very broad frequency band without formant structure or low frequency content. Plosives will typically be proceeded by a blank part in the spectrogram, as the pressure build-up requires a stop in the air-flow.

A special case of vocal sound may occur in the transition between [ e ] - [ æ ] - [ a ], namely a phenomenon called glottal whistling, apparently triggered by a quasi-harmonic constellation of the four lowest formant frequencies, e.g. 850—1700—2550—3400Hz, as can be seen in this window (resizable).

Link to external sources:

Voice acoustics: An introduction

Speech Acoustics (slides show—click the full screen button there for larger image)

 

 

Voice acoustics

Chest radiation

While most of the voice sound energy radiates from the head— basically from mouth and nose– a large fraction of the low frequency sound of vowel sounds in deep male voices radiates from the chest. Measurements have proved examples of as much as 78% of the sound power of the fundamental tone radiating from the chest of a bass singer while humming on an E flat, see the paper Sound radiation from the chest of bass singers from SMAC 93 (Stockholm Music Acoustics Conference 1993) (SMAC93 proceedings: Content).

Formants and Harmonics

During speech or singing, formant  frequencies and harmonic frequencies may change rapidly. While formant frequencies change as vowels change, the harmonic frequencies change proportionally to the pitch of the voice, i.e. the fundamental frequency. The combined effect of formants and harmonics can be studied in a glissando over the voiced vowel [ i ]. In harmonic singing (i.e. harmonic chant , overtone singing), a trained singer tune narrow-banded formants to emphasize individual harmonics, being able to create polyphony in three different ways:

1. Changing the formant frequencies (vowels) while harmonics are constant (constant pitch)

2. Changing the harmonics (changing pitch) while formant frequency (vowel) is constant

3. Combining 1. and 2.

Sound example: Harmonic chant by David Hykes (You Tube, opens in separate window)

on site