Artificial Software Complex "Artificial Head". Part 2 Evaluation of Speech Intelligibility in Classrooms

Abstract

Experimental studies of the use of the developed hardware and software tool “Artificial Head” for two-channel estimation of the intelligibility of the speech distorted by reverberation have been performed in this paper. The peculiarity of this complex is that it contains electro-acoustic equipment of different quality, including household appliances with mediocre quality features.In the first stage of this evaluation, the response of the room to the test signal is recorded. The test signal was based on the maximum-length sequence (mls) which is a periodic two-level signal of length 216, which corresponds to a signal length of 1.49 s at a sampling rate of 44.1 kHz. This mls-sequence was repeated 17 times during radiation and averaging the last 16 bursts of the cross-correlation estimates have been made to increase the signal-to-noise ratio by 12 dB in the calculation of the room impulse response.The calculation of the cross-correlation function of the response and the test signal was performed in the second stage, while the correction of the frequency characteristic of the measuring path was performed by it dividing by the amplitude-frequency characteristic of the loudspeaker-microphone subsystem. To improve the accuracy of the calculations, the result of such division was multiplied by the regularization factor in the form of a Hann window.In the third stage, the modulation coefficients were calculated according to the Schroeder formula, using the room impulse response estimate. In the fourth, last step, speech intelligibility was evaluated by a modulation or formant-modulation method. The results of the speech intelligibility evaluation in two lecture rooms of small and medium size showed that the speech intelligibility in the middle of the room can be less than that near the wall of the room. These results are in line with the results of previous studies, where speech intelligibility has been evaluated by objective and subjective methods for another lecture room. It should be noted that although the estimates of the widely used C50 coefficient are consistent with the estimates of speech intelligibility, the phenomenon of increasing speech intelligibility near the walls of the room is more pronounced when applying the speech intelligibility estimates.The results presented in this paper indicate that the usefulness of early reflections in the room is different in different parts of the room. This important fact must be taken into account both in the design and the reconstruction of the lecture rooms.