Design of a Multimodal Emotion Analysis System Based on Speech and text in Business Negotiations

With the advancement of globalization, business negotiations are no longer limited to simple exchange of interests, and the recognition and management of emotions play an increasingly important role in negotiations. Traditional emotion judgment methods rely on the experience and intuition of negotiators and have great limitations. In order to improve the accuracy and real-time performance of emotion analysis, this study designed and implemented a multimodal emotion analysis system based on speech and text. By fusing speech signals and text data, the system can recognize and analyze the emotional state of both parties in real time, providing data support for the adjustment of negotiation strategies and communication optimization. In terms of speech emotion recognition, Mel-frequency cepstral coefficients (MFCC) and deep learning models are used; in text emotion analysis, sentiment dictionaries and deep learning techniques are combined. Experimental results show that the system performs well in multimodal data fusion, can effectively improve the accuracy of emotion analysis, and significantly improve the efficiency and success rate of business negotiations. However, the system still faces challenges such as noise interference and diversity of emotional expression in practical applications. In the future, it is necessary to further optimize the model and algorithm to enhance its cross-cultural adaptability and robustness.