Facial Recognition Using Joint Bayesian Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing facial recognition methods based on joint Bayesian probability models face inefficiencies due to the need for extensive comparisons between feature vectors in large face image databases, leading to high calculation burdens and long processing times, limiting their applicability in fast recognition scenarios.
Innovation Solution
The method involves selecting k candidate vectors from a face image database using a joint Bayesian probability matrix, where vectors include components [v·A,v·B·vT], and determining a matching vector with the shortest cosine distance, allowing for efficient facial recognition by pre-calculating and storing medium-level feature interactions offline, reducing online calculations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a to-be-matched face feature vector is compared with all vectors in a face image database using the joint Bayesian probability model, then recognition accuracy is improved, but calculation burden increases and processing time is extended
Solution Approach 1:
The patent segments the face image database into multiple clusters, where each cluster contains a subset of face vectors. Instead of comparing the to-be-recognized face vector with all vectors in the database, the system first identifies which cluster the face belongs to and then performs detailed comparison only within that cluster. This segmentation reduces the number of comparisons from N (total database size) to M (cluster size), where M << N, thereby improving processing speed while maintaining recognition accuracy.
Solution Approach 2:
The patent performs preliminary clustering of the face image database offline before the actual recognition process. The database is pre-organized into clusters based on facial features, and this cluster structure is stored for quick reference. During online recognition, the system only needs to perform a simple cluster assignment followed by limited vector comparisons, rather than exhaustive comparisons with all database vectors. This preliminary organization significantly reduces computational burden during real-time recognition.
2Productivity
If a joint Bayesian probability matrix is learned offline and decomposed into submatrices A and B, then recognition efficiency is improved, but device complexity increases
Solution Approach 1:
The patent decomposes the joint Bayesian probability matrix P into two submatrices A and B, where P = A × B. This segmentation allows the system to perform lighter computations during online recognition by using only the pre-computed submatrices, rather than manipulating the full probability matrix. The decomposition transforms a complex matrix operation into simpler, pre-prepared components that can be efficiently applied during recognition.
Solution Approach 2:
The patent performs the computationally intensive task of learning and decomposing the joint Bayesian probability matrix offline, before the actual recognition process. The submatrices A and B are pre-calculated and stored, so that during online recognition, the system only needs to perform simple matrix multiplications with these pre-computed components. This preliminary action shifts the computational burden from the online recognition phase to the offline training phase, improving real-time recognition efficiency.
Data Source
AI summary
The present invention provides a method and an apparatus for determining an identity identifier of a face in a face image, and a terminal. The method includes: obtaining an original feature vector of a face image; selecting k candidate vectors from a face image database; selecting a matching vector of the original feature vector from the k candidate vectors; and determining, an identity identifier that is of the matching vector. In embodiments of the present invention, a face image database stores a medium-level feature vector formed by means of mutual interaction between a low-level face feature vector and autocorrelation and cross-correlation submatrices in a joint Bayesian probability matrix. The medium-level feature vector includes information about mutual interaction between the face feature vector and the autocorrelation and cross-correlation submatrices in the joint Bayesian probability matrix, so that efficiency and accuracy of facial recognition can be improved.