Alcántara J I, Dooley G J, Blamey P J, Seligman P M
Human Communication Research Centre, University of Melbourne, Australia.
Audiology. 1994 Jan-Feb;33(1):15-27. doi: 10.3109/00206099409072951.
The effects on speech perception in noise of dynamic filtering with bandpass filters centred at the first formant (f1) and second formant (f2) frequencies were evaluated with four normally hearing listeners. Multitalker babble was added to the speech signal with signal-to-noise ratios of -5 to -15 dB, chosen to reduce intelligibility to about 50%. The combined signal was then filtered with two-pole programmable bandpass filters centred at f1 and f2 under the control of a real-time speech processor. The f1 and f2 frequencies were estimated from the speech signal before noise was added to avoid hardware processing errors. Closed set vowel and consonant tests (using 11 /h/vowel/d/ and 12 /a/consonant/a/ stimuli), the Consonant-Nucleus-Consonant Monosyllabic Word Test and the Bamford-Kowal-Bench Sentence Test were carried out for three filter bandwidths (3/4, 1/3 and 1/6 octave) and for unprocessed speech in noise. The processing produced a small significant improvement for vowels in all three processed speech conditions and for monosyllables at the broadest filter setting compared to the unprocessed speech condition. There was no significant effect on consonants. A small negative effect was observed for sentences at the narrowest filter setting.
利用四个听力正常的受试者，评估了以第一共振峰（F1）和第二共振峰（F2）频率为中心的带通滤波器进行动态滤波对噪声中言语感知的影响。将多说话者嘈杂声添加到语音信号中，信噪比为-5至-15dB，选择该信噪比是为了将可懂度降低到约50%。然后，在实时语音处理器的控制下，用中心频率为F1和F2的两极可编程带通滤波器对组合信号进行滤波。在添加噪声之前，从语音信号中估计F1和F2频率，以避免硬件处理错误。针对三种滤波器带宽（3/4、1/3和1/6倍频程）以及噪声中的未处理语音，进行了闭集元音和辅音测试（使用11个/h/元音/d/和12个/a/辅音/a/刺激）、辅音-元音-辅音单音节词测试以及班福德-科瓦尔-本奇句子测试。与未处理的语音条件相比，在所有三种处理后的语音条件下，处理对元音产生了小的显著改善，并且在最宽的滤波器设置下对单音节词也有改善。对辅音没有显著影响。在最窄的滤波器设置下，观察到对句子有小的负面影响。
