結合詞意分析之語音情緒辨識系統之研究

A Study of Applying the Spoken Word Semantic Analysis to the Speech Emotion Recognition System

楊岫瑿(Hsui-yi Yang)

指導教授 ： 包蒼龍

大同大學/電機資訊學院/資訊工程學系所/碩士(2011年)

摘要

近年語音情緒辨識是許多研究者所關注的議題，過去的研究多以短語句的語音訊號 來做分析辨識，但在日常生活中，人與人之間的對話大部分都是以長語句的型式呈 現，因此，電腦系統在這方面的辨識率就顯得不盡理想。為了提高長語句的辨識率， 以及更貼近人類對說話者的語音情緒之判定，本論文提出在語音情緒辨識系統中增 加對長語句字詞的詞意分析，以提升系統對長語句中情緒的辨識率。透過分析實驗 結果，本實驗所提出的辨識系統的辨識率相較於單獨用語音情緒辨識系統所得到的 辨識率提升8%。

關鍵字

語音情緒辨識 ； 詞意分析

並列摘要

In recent years, the speech emotion recognition is one of the active topics in speech signal processing as well as human emotion research. The majority of corpus used in the speech emotion recognition researches is based on short corpus. However, in the daily human conversation, what we used are almost long sentences. Consequently, the accuracy of speech emotion recognition is low when we apply it in the real life situation. In order to improve the emotion recognition rate for long sentences and be more close to the real feeling of emotion perceived by human, we propose a method which combines the semantics of spoken word and the emotion recognized from the speech signal. The results
indicated that by combining the spoken word semantics, we can increase the recognition accuracy by 8% for various scripts as compared with those using speech emotion recognition only.

並列關鍵字

analysis the spoken word semantics ； speech emotion recognition

參考文獻

[1] J.H. Yeh, Emotion Recognition from Mandarin Speech Signals, Master Thesis,

[2] A. Ortony and T. J. Turner, “What’s Basic about Basic Emotions,” Psychological Review, pp. 315-331, 1990

[4] X. Jin and Z. Wang, “An Emotion Space Model for Recognition of Emotions in

Spoken Chinese,” First International Conference on Affective Computing and

[8] Y.H. Chang, Emotion Recognition and Evaluation of Mandarin Speech Using