A brainwave emotion estimation device based on time-frequency domain features and language cues

Abstract

This invention provides a method and apparatus for estimating emotion based on time-frequency domain features and language cues, belonging to the field of computer technology. The method includes: segmenting, normalizing, and filtering the input raw EEG signal, retaining signals in a preset frequency band, and generating a time-spectrum map through continuous wavelet transform; extracting time-frequency feature vectors based on the time-spectrum map using a preset deep network; calculating the differential entropy of the preset frequency band based on the raw EEG signal to generate a spatiotemporal feature vector; embedding language cues into a text description reflecting the subject's emotional state using a large language model; fusing the time-frequency feature vectors and the spatiotemporal feature vectors, modally aligning the fused features with the language cues embedding, and outputting the emotion classification result through a classifier. This invention can capture key patterns in the EEG signals of subjects with different emotions, significantly improving the accuracy, objectivity, and reliability of the assessment.

Images