Psychological research points out that nonverbal cues play an important role in interpersonal communication and information judgment. As an important nonverbal clue, does voice sentiment play an important role in information transmission in the capital market? Can this clue help investors judge the management's "duplicity"? this paper first constructs a signaling game model to analyze the voluntary information disclosure behavior of listed companies. Using the audio data of the earnings conference call held in 2013-2016 by Listed Companies in China, this paper quantifies the voice sentiment of executives, and empirically tests the role of the executives’ voice sentiment in the signaling game framework. The results show that: (1) the "good" companies are more inclined to hold and disclose the earnings conference call and send "signals" to the market, but the "bad" companies also try to mix in (2) voice sentiment contain incremental information beyond tone(text) sentiments, which can predict the future performance of the company. (3) Investors are more sensitive to negative sentiments (4) During the Q&A interaction or under-perform than expectation, the pressure will lead to more negative sentiments released by executives. This study shows that non-verbal voice sentiment has a unique information content, which can improve the cost of "false signals" sent by "poor" companies and help to achieve "separation equilibrium". This paper enriches the relevant research of voluntary information disclosure, and proves that voice sentiment plays an important role in the capital market. It also provides a useful reference for how to use new dimension data to cross research in the financial field under the background of big data.