WhisperNetV2: SlowFast Siamese Network For Lip-Based Biometrics

TL;DR

WhisperNetV2 introduces a SlowFast Siamese network for lip-based biometrics, effectively capturing physiological and behavioral lip features, and achieves state-of-the-art accuracy by considering emotional variations during video acquisition.

Contribution

The paper proposes WhisperNetV2, a novel deep Siamese network with SlowFast architecture that improves lip-based biometric authentication by incorporating emotional variability and triplet loss.

Findings

Achieved an EER of 0.005 on CREMA-D dataset.

Outperforms most existing LBBA methods in accuracy.

Utilizes a dual-path SlowFast network to capture different lip features.

Abstract

Lip-based biometric authentication (LBBA) has attracted many researchers during the last decade. The lip is specifically interesting for biometric researchers because it is a twin biometric with the potential to function both as a physiological and a behavioral trait. Although much valuable research was conducted on LBBA, none of them considered the different emotions of the client during the video acquisition step of LBBA, which can potentially affect the client's facial expressions and speech tempo. We proposed a novel network structure called WhisperNetV2, which extends our previously proposed network called WhisperNet. Our proposed network leverages a deep Siamese structure with triplet loss having three identical SlowFast networks as embedding networks. The SlowFast network is an excellent candidate for our task since the fast pathway extracts motion-related features (behavioral lip movements) with a high frame rate and low channel capacity. The slow pathway extracts visual features (physiological lip appearance) with a low frame rate and high channel capacity. Using an open-set protocol, we trained our network using the CREMA-D dataset and acquired an Equal Error Rate (EER) of 0.005 on the test set. Considering that the acquired EER is less than most similar LBBA methods, our method can be considered as a state-of-the-art LBBA method.