Stadler R W, Rabinowitz W M
Research Laboratory of Electronics, Massachusetts Institute of Technology, Cambridge 02139.
J Acoust Soc Am. 1993 Sep;94(3 Pt 1):1332-42. doi: 10.1121/1.408161.
Microphone arrays with fixed (time-invariant) weights are directed at enhancing a desired signal from one direction (straight ahead) while attenuating spatially distributed interference and reverberation. Using the theory of sensitivity-constrained optimal beamforming [Cox et al., IEEE Trans. Acoust. Speech Sig. Process. ASSP-34, 393-398 (1986)], free-field arrays of head-sized extents were studied. The key parameters affecting array design and performance are the set of transfer functions from the target direction to each array microphone [H(f)] and the intermicrophone cross-spectral densities for isotropic noise [Szz(f)]. Design variables included the orientation of the array, the number, and [as motivated by Soede, Ph.D. thesis, Delft University of Technology (1990)] the directionality of the microphones within the array, and the complexity and robustness of the required processing. Performance was characterized by the broadband intelligibility-weighted directivity (gain against isotropic noise) and noise sensitivity (reflecting the array's sensitivity to uncorrelated noise, as well as device tolerances). For broadside orientation, a variety of arrays based on cardioid and hypercardioid microphones gave very similar performance. They can provide directivities of 7-8 dB with easily implemented weights (simple scalars). For endfire orientation, as Soede (1990) recognized, similar directivities result with weights based on analog gains and pure time delays. However, with weightings chosen independently for each frequency, directivities up to approximately 11 dB may be obtained, although the increased noise sensitivities of these arrays require practical evaluation. Because of sound diffraction, placement of arrays onto the head potentially impacts both their design and performance. In-situ measurements of H(f) and Szz(f) as well as simplified theoretical models are suggested to explore the optimization of head-mounted arrays.
具有固定（时不变）权重的麦克风阵列旨在增强来自一个方向（正前方）的期望信号，同时衰减空间分布的干扰和混响。利用灵敏度约束最优波束形成理论[考克斯等人，《IEEE声学、语音和信号处理汇刊》ASSP - 34，393 - 398（1986）]，对头部尺寸大小的自由场阵列进行了研究。影响阵列设计和性能的关键参数是从目标方向到每个阵列麦克风的传递函数集[H(f)]以及各向同性噪声的麦克风间互谱密度[Szz(f)]。设计变量包括阵列的方向、麦克风数量，以及[受代尔夫特理工大学Soede博士论文（1990年）启发]阵列内麦克风的指向性，还有所需处理的复杂度和鲁棒性。性能通过宽带可懂度加权指向性（相对于各向同性噪声的增益）和噪声灵敏度来表征（反映阵列对不相关噪声的灵敏度以及设备容差）。对于侧射方向，基于心形和超心形麦克风的各种阵列表现出非常相似的性能。它们可以通过易于实现的权重（简单标量）提供7 - 8 dB的指向性。对于端射方向，正如Soede（1990年）所认识到的，基于模拟增益和纯时间延迟的权重会产生相似的指向性。然而，当为每个频率独立选择权重时，尽管这些阵列的噪声灵敏度增加需要实际评估，但可以获得高达约11 dB的指向性。由于声音衍射，将阵列放置在头部可能会影响其设计和性能。建议进行H(f)和Szz(f)的现场测量以及简化的理论模型，以探索头戴式阵列的优化。
