以腦波測量複雜場景中聲音對於人類感知的顯著性

EEG-based Measurement of Auditory Saliency in a Complex Context

黃薰誼(Xun-Yi Huang)

指導教授 ： 林文杰

國立交通大學/資訊學院/多媒體工程研究所/碩士(2015年)

https://doi.org/10.6842/NCTU.2015.00744

摘要

聽覺顯著性是幫助人類在特定的聽覺環境提取相關信息的重要機制。前注意顯著性(或稱自下而上顯著性)會在人類聽覺系統自動發生，發生時人也不需要注意聽聲音刺激。在行為的實驗方法中，受試者會被要求聽到兩個聲音後，指出一個較突出(顯著)的聲音，這個方法常用於衡量或評估聲音聽覺的顯著性;然而，行為的實驗方法卻不能真實地反映前注意顯著性，因為受試者的決策過程往往涉及到一些更複雜的認知過程和行為反應。因此，發展一個更好的前注意聽覺顯著性測量方法是必要的。此研究中，我們提出了以腦電圖為基礎的方法，它能與行為實驗方法的效益互補，提供更詳細的分析並更深入了解聽覺顯著性。我們使用了兩種類型的神經反應來觀察受試者的聽覺感知反應：不匹配負波（MMN）來觀察前注意聽覺處理，以及P3a反映注意力轉移的程度。為了展示我們的方法，我們以腦電波實驗測量聲音的MMN和P3a，實驗包括兩種現實生活的場景，大眾捷運系統（MRT）與小客車艙內。研究結果顯示，MMN和P3a能提供更多的信息探索自下而上的顯著性。我們也相信，我們的研究從三個面向的提供聲音顯著性分析新的見解，分別為前注意，注意，與行為的顯著性這三個面向。

關鍵字

聲音顯著性 ； 下而上的顯著性 ； 腦電波儀 ； 不匹配負向波

並列摘要

Auditory saliency is an important mechanism for helping human extract relevant information in a particular auditory environment. Pre-attentive/bottom-up saliency, which occurs automatically in human auditory system, does not require humans to pay attention to audio stimuli. Behavioral studies, in which subjects are asked to determine a more salient sound after hearing two sounds, is commonly applied to measure or evaluate auditory saliency; however, behavioral studies may not truly reflect pre-attentive saliency because the subjects’ decision process often involves some complicate cognitive processes and behavioral responses only provide limited observations. Thus, a better measurement method for pre-attentive auditory saliency is necessary. In this study, we propose an electroencephalography-based approach that can complement behavioral studies to provide a more elaborate analysis of auditory saliency. We used two types of neural responses to detect subjects’ auditory perceptual responses: mismatch negativity (MMN) for pre-attentive processing and P3a for attention shifting. We demonstrate our method with an EEG experiment that measures MMN and P3a of measured sounds in two realistic-like scenarios, including carriage in a massive rapid transportation (MRT) and cabin in a car. The results of this study exhibit that MMN and P3a provide more detailed information for deeply exploring bottom-up saliency. We also believe that our study could provide novel insights at the view of three perspectives of saliency : pre-attentive, attentive, and behavioral saliency.

並列關鍵字

auditory saliency ； bottom-up saliency ； electroencephalography (EEG) ； mismatch negativity

參考文獻

[1] V. Duangudom and D. V. Anderson, "Using auditory saliency to understand complex auditory scenes," in 15th European Signal Processing Conference (EUSIPCO), 2007, pp. 1206-1210.

[2] F. Tordini, A. S. Bregman, J. R. Cooperstock, A. Ankolekar, and T. Sandholm, "Toward an improved model of auditory saliency," in Proceedings of the International Conference on Auditory Display (ICAD2013), 2013.

[4] C. Kayser, C. I. Petkov, M. Lippert, and N. K. Logothetis, "Mechanisms for allocating auditory attention: an auditory saliency map," Current Biology, vol. 15, pp. 1943-1947, 2005.

[5] M. Elhilali, J. Xiang, S. A. Shamma, and J. Z. Simon, "Interaction between attention and bottom-up saliency mediates the representation of foreground and background in an auditory scene," PLoS biology, vol. 7, p. e1000129, 2009.

[6] V. Duangudom and D. V. Anderson, "Identifying salient sounds using dual-task experiments," in Applications of Signal Processing to Audio and Acoustics (WASPAA), 2013 IEEE Workshop on, 2013, pp. 1-4.