聲音特質與文字的情緒向性對於聲音偏好度之影響

Abstract

此研究探討中文母語者對於聲音的好感度判斷，其中聲音特質與說話內容為兩大討論重點。過去研究顯示，音質特色在西方言語溝通中，為衡量好感度的關鍵因素。本研究旨在分析聲音特質對中文講者的影響，並且深入討論說話內容的正負向情緒是否會影響聽者對於說話者的印象。再者，我們分析聽者在判斷其聲音好感度上的重要音質特質，分別在中性、正向、負向說話內容上的表現。研究參與者皆為年輕的中文母語者，每位參與者皆需參與聲音好感度判別的測驗。此研究播放錄音室錄製的不同人聲，共有180種（3種文字內容 X 5種句子 X 6位錄音者 X 2種性別）。參與者每一次只會聽到一個說話者的聲音，接著就必需直覺的回答他們對於此聲音的好感程度，作答方式是使用「五點李克特量尺」(five-point Likert scale) 來做出程度上的分別。實驗結束後，參與者需憑藉方才作答之印象，回答關於聲音特質及文字情緒的相關問題。本研究發現聲音特質及文字內容皆會影響中文聽者對於中文說話者的好感程度。首先，正向、中性及負向說話內容有明顯的差異。其中，正向內容最廣為喜愛，其次為中性內容，再來為負向語句。根據參與者的實驗後問答，正向文字會增加聲音的好感度。接者，在眾多聲音特質(acoustic features)中，頻率擾動度擾動度 (jitter)是衡量聲音喜好的最關鍵因素。無論在正向或負向內容中，頻率的穩定性越高，此聲音就很容易被評斷為好聽。另外，本研究也發現音質影響力與內容息息相關。舉例來說，對於中性內容，穩定度是最重要的，但對正向內容，音高的表現更為重要。最後，我們討論了各種音質特質如何影響著中文的聲音好感度。頻率擾動度(jitter)及音量擾動度(shimmer)皆對好感度有負面的作用，而諧噪比 (HNR) 、平均音高 (pitch mean) 和音高全距 (pitch range) 皆有正面的作用。此實驗也證實了聲音好感度不僅和聲音特質相關，其聲音呈現是否符合說話者的內容也是判斷要素之一。

This study investigates two major issues on Mandarin voice preference: the effect of voice quality and text emotions. Previous research has pointed out the importance of voice quality in Western languages. This study therefore explores the impact of voice quality (e.g., pitch mean, pitch range, jitter, shimmer, HNR, and duration) in Mandarin, and put forward to the investigation of text emotions (neutral, positive, and negative). Finally, we examine whether listeners rely on different acoustic information while making their preference judgment for different textual emotions.To examine the influence of voice quality and text emotion on voice preference, this thesis includes one voice preference task. We had a group of native Mandarin speakers rate a variety of voices with their personal preference. Materials were 180 stimuli (5 experimental speech x 3 text emotions x 6 Mandarin speakers x 2 genders). The participants were instructed to mark their voice preference on a five-point Likert scale. After the experiment, listeners would answer a few questions about voice qualities and text emotions.Our results showed that both voice quality and text emotions demonstrated their influence on voice preference. Concerning the content of the utterance, positive utterances are more favored than those with neutral and negative content. From statistical analysis and qualitative response, it was suggested that lexical positivity is likely to enhance listeners’ preference for a voice. As for the acoustic influence, jitter is found to be the most important acoustic cue. Regardless of content, the amount of jitter stands as the most influential factor on voice preference. Besides, we also found that acoustic influence is associated with the content. Listeners may rely on different acoustic cues while facing different content. For instance, vocal stability is the most crucial factor in neutral content, while pitch performance is more essential for positive utterance. Finally, we further discuss the acoustic influence in Mandarin. Increasing jitter and shimmer was found to be detrimental to Mandarin voice preference, while that of HNR, pitch mean, and pitch range can greatly enhance one’s personal preference for a voice. We therefore conclude that both influences of voice and text should be taken into account regarding the discussion on voice preference. A preferred voice should be defined with a suitable acoustic performance on the corresponding text valence.