3D Landmark-Guided Facial Expression Recognition

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing facial expression recognition systems often fail to accurately and efficiently utilize facial landmark detection, leading to suboptimal performance in classification tasks.

Innovation Solution

Integrate three-dimensional morphable model (3DMM) techniques for facial landmark detection within neural networks to enhance the accuracy of facial expression recognition by focusing on relevant image regions, without increasing processing time or power consumption.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If facial landmark detection is integrated into neural networks for facial expression recognition, then classification accuracy is improved, but system complexity increases

Engineering Contradiction:
Improvefacial expression classification accuracyVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent combines facial landmark detection and facial expression recognition into a single integrated neural network model. The network simultaneously processes input images to detect facial landmarks (key points on the face) and recognize expressions, merging two previously separate functions into one unified system. This integration improves classification accuracy by allowing the model to leverage landmark information directly during expression recognition without requiring separate processing stages.

Inventive Principle:
Principle #5Merging (Combining)

2Measurement precision

If three-dimensional morphable model techniques are used for landmark detection, then relevant image regions are focused on, but processing time increases

Engineering Contradiction:
Improvelandmark detection precisionVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent extracts and utilizes only the essential landmark feature information from the three-dimensional morphable model technique, rather than processing the complete 3D model data. By selecting and focusing on specific landmark points that are most relevant for facial expression recognition, the system achieves precise landmark detection while avoiding the computational overhead of processing all 3D model components, thus maintaining efficiency.

Inventive Principle:
Principle #2Taking out (Extraction)

3Measurement precision

If facial landmark detection is implemented in neural networks, then expression recognition precision is improved, but computational power consumption increases

Engineering Contradiction:
Improveexpression recognition precisionVSAvoidcomputational power consumption
Core Design Contradiction:
Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The patent applies local quality by focusing computational resources on specific landmark regions of the face that are most informative for expression recognition, rather than uniformly processing the entire face image. The neural network is designed to pay particular attention to key facial regions (such as eyes, eyebrows, and mouth areas) where landmark variations most strongly indicate emotional expressions. This localized processing approach improves recognition precision while reducing overall computational power consumption by avoiding unnecessary processing of less relevant facial regions.

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP4298616B1Facial expression recognition
Publication Date: 2026.03.25 QUALCOMM INC
  • EP4298616B1 patent drawingFigure 1
  • EP4298616B1 patent drawingFigure 2
  • EP4298616B1 patent drawingFigure 3A

AI summary

Systems and techniques are provided for facial expression recognition. In some examples, a system receives an image frame corresponding to a face of a person. The system also determines, based on a three-dimensional model of the face, landmark feature information associated with landmark features of the face. The system then inputs, to at least one layer of a neural network trained for facial expression recognition, the image frame and the landmark feature information. The system further determines, using the neural network, a facial expression associated with the face.