Cognitive Demand Recognition Modeling Based on Speech Signal Processing

Prediction of cognitive needs using voice signals is an important subject in information science, management science and sociology, and it is also a very potential research direction. The paper is about the regression prediction of cognitive needs by using speech signal modeling. Its purpose is to find a new cognitive needs assessment model that enables rapid, accurate and non-intrusive assessment of individuals, making up for the lack of timeliness and limited application scope. We obtained the voice data of 240 participants and their cognitive needs scale evaluation results. Then, we use principal component analysis to reduce the dimension of speech features, and use support vector machine, random forest and neural network models to predict cognitive needs. We find that the neural network model not only has the highest prediction accuracy, but also has the most uniform distribution of prediction errors. Through the research, we put forward our own suggestions and prospects for the future research direction of cognitive needs.