294 BELL SYSTEM TECHNICAL JOURNAL 



such substitutes so that they may be used with confidence in general 

 testing and physical measurement of telephone transmitters and re- 

 ceivers. The requirements which an artificial voice and ear must 

 meet to insure proper performance are outlined below. 



Requirements for an Artificial Voice 

 An ideal artificial voice must be able to reproduce human speech 

 without introducing any change in frequency, amplitude or directivity 

 over the entire range of intensities possible in speech. Furthermore, 

 it should react to changes in acoustic load in the same manner as does 

 the human voice under ordinary conditions. Such requirements, of 

 course, must be reduced to a more specific and practical form to 

 enable the construction of a physical piece of apparatus. 



It is useful to consider the artificial voice as consisting of two 

 parts, the mouth and the source of power. Practical considerations 

 point immediately to the use of some form of electro-acoustic trans- 

 ducer for the mouth, and any of several sources of electrical power. 

 In the production of speech there is required as a source of power either 

 a high quality transmitter or a phonograph record and reproducing 

 system. For purposes of physical measurement and analysis, a 

 source of single-frequency power, such as an oscillator, is needed. 



If it be assumed that the frequency composition of the actual human 

 voice is automatically included with either of the sources of speech 

 mentioned, and that proper frequency weighting will be introduced in 

 the single-frequency source when desired by means of suitable electrical 

 networks, the practical requirements for the artificial mouth may 

 then be stated specifically as follows: 



(1) It should introduce no amplitude distortion within the range of 



speech frequencies. 



(2) It should be capable of delivering an acoustic output without non- 



linear distortion over the range of intensities possible for the 

 human speaking voice. 



(3) The distribution of the sound field about the mouth at every 



frequency and distance should be the same as that of the human 

 mouth. 



(4) The introduction of objects such as transmitters in the sound 



field should react on the output of the mouth and distort the 

 field in the same way as they do when introduced in the field 

 of the human mouth. 



(5) It should be completely specifiable and reproducible as well as 



constant in performance. 



