On Features Obtained by Insertion of White Noise into Intermittently Removed Intervals of Speech Signals
Manabu Ishihara*, and Jun Shirataki**
* Department of Electronic Engineering, Polytechnic Unviersity, 4-1-1, Hashimotodai, Sagamihara, Kanagawa, 229 Japan
** Department of Electrical and Electronic Engineering, Kanagawa Institute of Technology, 1030, Shimoogino, Atsugi, Kanagawa, 243-02 Japan
In this study, a signal was synthesized by removing a speech signal at a certain uniform interval and inserting noise into those signal–absent parts. An auditory experiment was conducted to make clear how humans can hear such synthesized signals. In other words, the relationship between the size of noise and the intensity of signal sound and the relationship between the size of noise and clearness degree were made clear. On the basis of the result of the experiment, in case the size of the white noise inserted is smaller than OdB, a degree of sentence comprehension of over 90 percent is obtained as long as the removed intervals amount to around 60 to 50 percent. In this case, the degree of sentence comprehension is seen to have improved by over 30 percent, in view of the fact that the single syllable comprehension is around 50 to 60 percent. Starting with the region where the removed intervals exceed 50 percent, the degree of sentence comprehension goes down sharply, but this is considered to be due to an effect of the insertion of the white noise. On the basis of the results of this experiment, one of the auditory characteristics to be realized by a digital circuit was made clear.