Reduction of Distance Calculation Time by Feature Dimension Compression for Person Identification with Quasi Homomorphic Encryption

DOI

Bibliographic Information

Other Title
  • 準同型暗号を用いた人物対応付けのための特徴量次元圧縮による距離計算時間の削減

Abstract

In this research, we have investigated encryption methods and feature extraction methods for person identification in order to reduce the distance calculation time in person identification with quasi homomorphic encryption. We have confirmed that by reducing the number of dimensions from 1536 to 512, the calculation time for person identification can be reduced by about 16%, or about 1 ms, compared to the original calculation time, with almost no degradation in the identification accuracy.

Journal

Details 詳細情報について

Report a problem

Back to top