Enhanced Performance of Multimodal Biometric Systems by Confidence Estimation
| Type of publication: | Idiap-Com |
| Citation: | sutapa-com-03-05 |
| Number: | Idiap-Com-05-2003 |
| Year: | 2003 |
| Institution: | IDIAP |
| Address: | Martigny, Switzerland |
| Note: | submitted as a project report for postgraduate course in speech and language engineering, 2003 under the supervision of Prof. Herve Bourlard |
| Abstract: | The main scope of this project is to identify the best method of confidence estimator whose performance could be reliable in comparison to multimodal fusion alone. To do that, three alternative approaches to prediction confidence estimation are presented and compared. Among the three methods, the first one is the Gaussian hypothesis, the second one is the non parametric and the third one is the proposed distance method, which will be discuss later on in section 4.3 The three techniques are tested and compared on two different types of fusion (face and speech) methods namely multi layer perceptrons (MLP) and support vector machine (SVM) and the basic modalities were speech and face. |
| Userfields: | ipdmembership={learning}, language={English}, |
| Keywords: | |
| Projects |
Idiap |
| Authors | |
| Added by: | [UNK] |
| Total mark: | 0 |
|
Attachments
|
|
|
|
|
Notes
|
|
|
|
|
