Multi-modal person verification system based on face profiles and speech
read more
Citations
Automatic Person Verification Using Speech and Face Information
A review of multimodal biometric systems: Fusion methods and their applications
Multibiometric fusion strategy and its applications: A review
Fusion of palm-phalanges print with palmprint and dorsal hand vein
Noise compensation in a multi-modal verification system
References
Speaker identification and verification using Gaussian mixture speaker models
Hierarchical chamfer matching: a parametric edge matching algorithm
Profile Authentication Using a Chamfer Matching Algorithm
Related Papers (5)
Frequently Asked Questions (12)
Q2. What is the purpose of the paper?
During the training phase of the system, a 12-dimensional, 4- mixture GMM is computed for each speaker using parameters extracted from the speech signal.
Q3. What is the main structure of the system?
The system is made up of a Profile Verification System (PVS), a Speaker Verification System (SVS) and a Fusing and Classification Module (FCM).
Q4. How was the ROC curve generated for each experiment?
For each experiment, a Receiver Operating Characterstics (ROC) curve was generated by varying the decision threshold continuously.
Q5. What is the main idea of the paper?
As stated before, the system is made up of 3 modules:Speaker Verification SystemProfile Verification SystemFusing and Classification ModuleThe SVS used is based on the Gaussian Mixture Model (GMM) approach [1].
Q6. What is the method for outlier removal?
After outlier removal the values from the SVS are changedin polarity in order to make them compatible with the PVS, as this is required by the FCM.
Q7. How is the profile compared to the person whose identity is being claimed?
Using a GMM, belonging to the person whose identify is being claimed, a similarity measure is computed by averaging the log-likelihood of individual frames.
Q8. What is the ROC curve for the SVS?
Figure 5 shows the ROC curve with w = 1.A good way to evaluate the performance of a verification system is by computing the equal error rate (EER), where FA = FR, the success rate (SR), where 1 FA FRreaches a maximum, and the FR for an FA of 1%.
Q9. What is the way to remove outliers from the SVS?
For w = 0, only the SVS was used, while for w = 1 only the PVS was used, hence it can be seen that the SVS has better performance than the PVS.
Q10. What is the method of outlier removal?
For the SVS, an adequate method of outlier removal is by finding the median (m) and the deviation from the median m 2 (same as standard deviation, except substituting the median for the mean).
Q11. How do you get a profile shot?
Profile shots were obtained by manually finding theframes in head rotating sequences where the person is facing left and not wearing glasses.
Q12. What is the difference between the two experiments?
The SR at 10dB and lower of the combined system with w = 0:5 is better than with w = 0:33, hence there is a trade-off between lower performance at high SNRs versus more robust performance at low SNRs.