Related Experiment Videos
[Retrieval of faces from features described by eyewitnesses]
K Adachi1, S Watanabe, A Suzuki
1National Research Institute of Police Science, Tokyo.
Shinrigaku Kenkyu : the Japanese Journal of Psychology
|April 1, 1992
Summary
A novel computer technique identifies faces from eyewitness descriptions. The feature sampling without replacement model demonstrated superior accuracy in facial recognition tasks.
Area of Science:
- Computer Science
- Forensic Science
- Psychology
Context:
- Eyewitness testimony is crucial in legal investigations.
- Accurate facial identification from descriptions is challenging.
- Existing facial recognition methods have limitations.
Purpose:
- To develop and evaluate a computer technique for facial identification based on eyewitness feature descriptions.
- To compare the effectiveness of three different probabilistic models for facial retrieval.
Summary:
- The study proposes a computer-aided facial identification system using eyewitness feature descriptions.
- Three models were developed: judgment through all features, feature sampling with replacement, and feature sampling without replacement.
- The feature sampling without replacement model yielded the highest accuracy in retrieving the correct face from a database.
Impact:
- Enhances the accuracy and efficiency of facial identification in forensic investigations.
- Provides a more reliable tool for law enforcement agencies.
- Contributes to improving the justice system through better eyewitness identification procedures.