Cognitive Load, as an indicator of pressure on working memory during task performing, attracts more and more research interests in recent years. By correctly measuring cognitive load levels, the system can adjust task procedure to maintain the cognitive load in an acceptable range; therefore, the subject can execute tasks more accurately and efficiently. Among many different cognitive load measuring approaches, speech-based measurement is effective due to its non-intrusive nature and possibility of online measurement. Most existing research on speech-based cognitive load measurement is based on manually extracted features, which prevent practical use. In this paper, some potential speech features, such as rate of pauses and rate of pitch peaks are investigated and proved to be effective. All feature extraction is based on automatic algorithm.
展开▼