الفهرس 
Summary
Level of speechOnly 14 pages are availabe for public view
 |  | from | 169 |  |  |
| | | from | 169 | | |
Abstract
Different acoustical measurements were preformed in fifteen different rooms of varied acoustical characteristics at the actual position of listeners. The relationships between these measurements and speech intelligibility scores were investigated to determine the most appropriate relationship. The best predictors of speech intelligibility were identified. The acoustical measurement include objective and architectural feature of the tested rooms. The objective acoustic measurements were: Articulation index (AI ), A weighting speech to noise level (S/N ), reverberation time ( RT ) for 60 dB decay, early to late sound energy ratio, (C50 ) in dB, Linear early to total sound energy ratio, (D50), useful to detrimental sound ratio (U50) in dB, calculated form early to late sound ratio.
signal to noise ratio in dB. On the other hand, the architectural acoustic measurement that considered in the room were: Volume to surface area ratio , Average sound absorption coefficient in the room, length width of the room and height of the ceiling. Subjective test was preformed in the tested room using listeners and especially Arabic rhythm Word lists to find speech intelligibility scores at the actual position of listeners.
Different analyses were carried out to find the relationship between speech intelligibility scores and each of individual and combined acoustic parameters to select the best model as a predictor of Arabic speech intelligibility in room. The preferred acoustic parameters of room that achieve highest degree of speech intelligibility were estimated which considered the acoustical design goals for ideal condition for speech in room. Fitting curve analysis and back propagation neural network used to find out the relations between speech intelligibility scores and acoustic and architectural parameters in room. Number of models was estimated to predict the effect of individual and combined acoustic parameters to speech intelligibility. The best models is neural network that used early to late sound ratio C50 or early to total sound ratio D50, reverberation time and Aweighting speech to noise ratio over octave band frequency from 125 to 4000 as input for learning. In room the recommended acoustical design criteria to achieve near perfect speech intelligibility can be stated as follows: 1?signal to noise ratio in dB <U+2265> 15,reverberation time RT60 < 0.65 sec and early to late sound energy in dB <U+2265> 5.5 or early to total sound energy ratio <U+2265> 0.79 2?Useful to detrimental sound ratio in dB <U+2265> 5.8 3?Articulation index <U+2265> .78 and reverberation time RT60 < 0.65 sec. 4?Signal to noise ratio in dBA <U+2265> 15, range of sound absorption coefficient at octave frequency from 125to 4000Hz for different values of volume to surface area ratio as follow: Volume to surface area ratio Sound absorption coefficient.