Speech frequency determination method - uses mathematical analysis of energy spectrum involving sum of peak frequencies

摘要

One of the problems in determining the frequencies in speech is that the vowels and other voiced sounds are modified by the physical characteristics and position of the speaker's mouth. Using conventional digital sampling techniques, an energy/frequency spectrum is obtained. Each of the peak frequencies, fn, is allocated a number n from 1 upwards. The base frequency is then calculated by solving the equation for the energy fuction Q(g) as the sum of values g from 1 to N of W (fn/g), where w is the peak value and g is a hypothetical scale factor.