Customer call emotion real-time analysis method and device based on deep learning, storage medium and computer equipment

By integrating multimodal information from acoustics, speech rate, and text semantics through deep learning, and utilizing cross-modal attention mechanisms and an end-to-end architecture, the accuracy and real-time performance of emotion analysis in customer service calls have been addressed. This enables instant emotion feedback and strategy adjustment, improving the precision and timeliness of service interactions.

CN122090881APending Publication Date: 2026-05-26CHENGDU LUYI TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202610016588.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-01-07
Publication Date
2026-05-26

AI Technical Summary

Technical Problem

Existing customer service call sentiment analysis technologies rely on single-modal analysis, which makes it difficult to accurately identify customers' true emotions in complex contexts. Furthermore, they lack real-time processing capabilities, resulting in high analysis latency and an inability to adjust communication strategies in a timely manner.

Method used

By integrating multimodal information from acoustics, speech rate, and text semantics through deep learning, and dynamically fusing features using a cross-modal attention mechanism, combined with an end-to-end real-time processing architecture, we can acquire and analyze multidimensional emotional states in customer service calls in real time.

Benefits of technology

It enables real-time output of customer emotional state during calls, allowing customer service personnel to adjust communication strategies promptly, significantly improving the accuracy and timeliness of service interactions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122090881A_ABST
    Figure CN122090881A_ABST
Patent Text Reader

Abstract

The invention discloses a customer call emotion real-time analysis method and device based on deep learning, a storage medium and computer equipment, and the method comprises the steps: obtaining a voice stream of a customer service call in real time, and converting the voice stream into a real-time text; extracting customer continuous voice segments from the voice stream, and extracting customer text content from the real-time text; based on a deep learning feature extraction model, extracting an acoustic feature sequence and a speech speed feature sequence from the continuous speech segments, and extracting a semantic feature sequence from the client text content; inputting the acoustic feature sequence, the speech speed feature sequence and the semantic feature sequence into a multi-modal emotion recognition model to obtain a multi-dimensional state vector representing the emotion of the customer, dynamically determining an intermediate vector of each feature sequence by the multi-modal emotion recognition model through a cross-modal attention mechanism, and performing feature fusion through a gating fusion unit to obtain a multi-dimensional state vector representing the emotion of the customer; performing noise suppression processing on the fusion feature vector through a noise suppression unit; and determining the real-time emotional state of the customer according to the multi-dimensional state vector.
Need to check novelty before this filing date? Find Prior Art