Score normalization in multimodal biometric systems

By Anil K. Jain, Karthik Nandakumar, Arun A. Ross. 2005

2,272 citations recorded by the source corpus.

Multimodal biometric systems consolidate the evidence presented by multiple biometric sources and typically provide better recognition performance compared to systems based on a single biometric modality. Although information fusion in a multimodal system can be performed at various levels, integration at the matching score level is the most common approach due to the ease in accessing and combining the scores generated by different matchers. Since the matching scores output by the various modalities are heterogeneous, score normalization is needed to transform these scores into a common domain, prior to combining them. In this paper, we have studied the performance of different normalization techniques and fusion rules in the context of a multimodal biometric system based on the face, fingerprint and hand-geometry traits of a user. Experiments conducted on a database of 100 users indicate that the application of min-max, z-score, and tanh normalization schemes followed by a simple sum of scores fusion method results in better recognition performance compared to other methods. However, experiments also reveal that the min-max and z-score normalization techniques are sensitive to outliers in the data, highlighting the need for a robust and efficient normalization procedure like the tanh normalization. It was also observed that multimodal systems utilizing user-specific weights perform better compared to systems that assign the same set of weights to the multiple biometric traits of all users.

Subfields: Computer Vision and Pattern Recognition, Information Systems, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Information Systems

In Signal Processing

Citations in the corpus

Statistical Meta-Analysis of Presentation Attacks for Secure Multibiometric Systems (2016, methodology): “9 The matching scores are normalized in [0, 1] using the min-max technique [22], [45].”
Learning to Control the Specificity in Neural Response Generation (2018, methodology): “Next, we use the min-max normalization method (Jain et al., 2005) to obtain the NIRF value.”
Preprocessing methods for near‐infrared spectrum calibration (2020, methodology): “This has been widely used in machine learning and pattern recognition systems.(41) Considering that the overall linear adjustments will only lead to differences in the constant term of a linear model, the min-max normalization (MMN) used in this paper was carried out on the instance level, which is ”
Face Spoof Detection With Image Distortion Analysis (2015, methodology): “We have evaluated two types of score-level fusion schemes: sum and min rules [43].”
Fusion of Band Limited Phase Only Correlation and Width Centroid Contour Distance for finger based biometrics (2014, background): “consolidate the vein and geometry information at the feature level because the feature sets might be incompatible (Jain et al., 2005).”
Query-adaptive late fusion for image search and person re-identification (2015, background): “On the other hand, late fusion refers to fusion at score or decision levels [20, 8, 29, 9].”
Physical metallurgy-guided machine learning and artificial intelligent design of ultrahigh-strength stainless steel (2019, methodology): “The inputs and outputs were normalized by the z-score method [58], given by Eq.”
Multimodal fusion vulnerability to non-zero effort (spoof) imposters (2010, background): “Then, for imposter comparisons, pairs of two different subjects are chosen, one to represent all modalities for enrollment and the other to represent all modalities for authentication.”
Likelihood Ratio-Based Biometric Score Fusion (2008, methodology): “After an empirical evaluation, we found that the min-max normalization [7] is the best for the datasets used here.”
Comparison of Min-Max normalization and Z-Score Normalization in the K-nearest neighbor (kNN) Algorithm to Test the Accuracy of Types of Breast Cancer (2021, methodology): “= The new value from the normalized results Xnew X = Old value Max (X) = Maximum value in the dataset Min (X) = Minimum value in the dataset Z-score normalization is a method of normalization based on the mean (mean value) and standard deviation (standard deviation) of the data [18] [19].”
Hierarchical Fully Convolutional Network for Joint Atrophy Localization and Alzheimer’s Disease Diagnosis using Structural MRI (2018, methodology): “These patch-level feature representations were further concatenated and processed via z-score normalization process [65] to perform linear SVM-based classiﬁcation.”
Quality-augmented fusion of level-2 and level-3 fingerprint information using DSm theory (2009, neutral): “This algorithm generates a match score which is then normalized in the range of [0, 1] using score normalization technique [22].”
Double logistic curve in regression modeling (2010, background): “Such functions have been considered, for instance, in describing enzyme kinetics in biochemistry [1], fatigue profiling in physiology [3], activation processes in tri-state neural networks [21], score normalizing in biometrics and genetics [9,23], and global climate change evaluations [25].”
Discriminant Correlation Analysis: Real-Time Feature Level Fusion for Multimodal Biometric Recognition (2016, methodology): “Finally, in case of decision level fusion the final results of multiple classifiers are combined via techniques such as majority voting [1]–[3].”