////■; soixns (»/• sriiuii 589 



forms ol ilu- >|utrli soiiiuls rctiuiriil inorr pririsc determination, aiici 

 iiuieiil ri'siarrli in llie art of ti-itplioiiy has emphasizc<l this need. 

 Thinraphiial reeortls of speech sounds, whicli form ,i siippltincnt to 

 I he present paper, are contributions to tiiis stu(i> . 



I 



N(ii|.. i)N nil-; (.'ii.\K.V( TKRisTic Frequencies ok Speech 



Spctch is, in itself, a sound wave — a succession of condensations 

 and rarefactions in the air. For the purposes of this study we are 

 not |)rimarily concerned with the mechanism of production, nor with 

 the processes of perception of speech, though it may be necessary 

 to digress to inquiries of this kind, in their bearing on certain charac- 

 teristics of speech. We are interested primarily in what can be 

 learnetl from the records of the speech vibrations themselves. 



Speech stiiuids are complex, that is, they are composites of simple 

 sounds, each component having a particular frequency, amplitude, 

 phase and duration. Considering speech in the mass, we find its 

 energy- distributed among frequencies from 75 to above 5,000 cycles 

 with the larger part of this energy contained in the region below 1,000 

 cycles. This distribution is shown approximately in Fig. 1 taken 

 from reference (8g); the limitation on these data being that the measur- 

 ing apparatus was not sufliciently sensitive to measure the speech 

 energy as.sociattxl with frequencies higher than 5,000 cycles. Inas- 

 much as the energy of speech resides largely in the vowel sounds, the 

 curve in Fig. 1 can also be taken as applying to the average distribu- 

 tion in the vowel sounds. The energy distribution diagram is of 

 fundamental importance in the physical study of speech sounds; it 

 reveals at once the frequencies of large energy content which are 

 characteristic. For each vowel sound, there is a distinctive energj' 

 fretiuency diagram. 



The consonant soun<ls present a difficult problem because of the 

 small amount of energy associated with them. Most of our knowledge 

 of the consonant sounds is qualitative: for example Fletcher (refer- 

 ence 8h) who studied the nature of speech by the method of testing 

 articulation when different frequency ranges are eliminated shows 

 that for t^vo fricative or sibilant consonants 5 and s, there are essential 

 frequency components which lie above 5,000 cycles. The character- 

 istic frequencies of the consonant sounds are usually only part of the 

 whole story; these sounds are richer in transients, and clearly less 

 periodic in their nature than the vowel sounds. And in between 

 the two broad classes of consonant and vowel sounds there is a group 



