We based our experiments on the assumption that the essential cause of noise disturbance should be judged from the observation of the speech distortion.
The vowel articulation loss by noise is ultimately due to the timbre distortion which causes the so-called mishearing, and the consonant articulationt loss is ultimately caueed by the masking effect of noise which results in difficult hearing. But generally speaking, the vowels are stronger against noise than the consonants and as for the magnitudes of noises which appear in the ordinary telephone lines we can put
vowel articulation=0,95÷1
simple sound articulation
=vowel arti.×Consonant arti.
÷consonant articulation
Consequently the disturbing effect of noise to the telephone speech may be considered chiefly as the maskibg effect, the effect of the timbre distortion being negligible.
By the feeble noise of 1000 cycles the consonants are invaded nearly in the same manner as when they are exposed to the attenuation stimulus. But by the strong noise of 1000 cycles the consonants are encroached as when they are influenced by the higher-frequency cutting stimulus. This fact forces us to conclude that the out-line curves of the masking area of noises upon the sensational area depend on the strength of noises.
In case of the lower frequency addition (200, 500 cycles) the vowel E is liable to be heard as U and in case of the middle frequency addition (1000, 2000 cycles) the vowel U is often heard as I.
The proximity of I to U is augmented by the two means, the one is middle frequency addition and the other is higher frequency cutting (employing the lowpass filter).
In like manner we can increase the proximity of E to U by employing the lower frequency addition or by cutting the higher frequency range.
Thus, as far as the vowel distortion is concerned, the higher frequency cutting process is almost equivalent to the lower frequency noise addition.
The fact that the inarticulation of the consonants is largely due to the masking effect of noise can be inferred from the phenomenon of the drop of consonants, which we can also observe in the experiments of the attenuation stimulus. In this phenomenon we see the ccinsonants combined with the vowel I is very apt to drop and the consonants combined with A or O vowel is dropped rarely.
In these experiments we neglected the relation between the pitch of the speech sounds and the frequency of noise. We remember only one case in which the vowel E pronounced by a certain girl accorded admirably with the 1000 cycles noise and sounded more powerfully and brightly than when the vowel was free from noise.
It seems that the trembling beats which are formed between the noise and the speech vowel have the tendency to modify unvoiced consonants (preceeding the vowel) into voiced consonants.
