Spark image

Physics of the human voice

How do two peoples voices differ from each other? In general notion they carry different frequencies and different overtones are mixed in their voices. That is the quality of sound, but I need something more. I need a definite unique feature by which I can recognize that this voice is say his or hers. That feature must be measurable. Is there any formula that exists or is there any way out?


Firstly you seem to have grasped all the important differences between one voice and another.
There will be differences in:
(a) Overall pitch
(b) The overtones (harmonics) within this. Remember that if two people sing the same note that the actual frequency heard may not actually be there. The perceived sound is made up of the sum of the harmonics within it.
(c) The distribution of energy within these overtones
(d) The "envelope" of the sound - in other words how quickly the sound begins and ends. You can see this very well when considering the differences between different musical instruments.
(e) The "speed" at which the word is delivered
(f) The vibrato within the word this is possible even for the spoken word but is very obvious in singing
(g) Control of the sound by the use of the diaphragm

I know of no actual formula for the reconstruction of the sound made by a given voice. We have done some work on it at school using a simple mixing desk but it was all trial and error.
I have an excellent article on the Acoustics of the Singing Voice (Scientific American Johan Sundberg, March 1977). I know its singing but the principle of speech must be somewhat similar.

The digital analysis and reconstruction of the voice is more accurate that looking at an analogue trace on an oscilloscope but the sampling rate is of vital importance.

© Keith Gibbs 2011