Amos Y. Johnson

Gait recognition using static, activity-specific parameters

A gait-recognition technique that recovers static body and stride parameters of subjects as they walk is presented. This approach is an example of an activity-specific biometric: a method of extracting identifying properties of an individual or of an individuals behavior that is applicable only when a person is performing that specific action. To evaluate our parameters, we derive an expected confusion metric (related to mutual information), as opposed to reporting a percent correct with a limited database. This metric predicts how well a given feature vector will filter identity in a large population. We test the utility of a variety of body and stride parameters recovered in different viewing conditions on a database consisting of 15 to 20 subjects walking at both an angled and frontal-parallel view with respect to the camera, both indoors and out. We also analyze motion-capture data of the subjects to discover whether confusion in the parameters is inherently a physical or a visual measurement error property.

A novel sequence representation for unsupervised analysis of human activities

Formalizing computational models for everyday human activities remains an open challenge. Many previous approaches towards this end assume prior knowledge about the structure of activities, using which explicitly defined models are learned in a completely supervised manner. For a majority of everyday environments however, the structure of the in situ activities is generally not known a priori. In this paper we investigate knowledge representations and manipulation techniques that facilitate learning of human activities in a minimally supervised manner. The key contribution of this work is the idea that global structural information of human activities can be encoded using a subset of their local event subsequences, and that this encoding is sufficient for activity-class discovery and classification. In particular, we investigate modeling activity sequences in terms of their constituent subsequences that we call event n-grams. Exploiting this representation, we propose a computational framework to automatically discover the various activity-classes taking place in an environment. We model these activity-classes as maximally similar activity-cliques in a completely connected graph of activities, and describe how to discover them efficiently. Moreover, we propose methods for finding characterizations of these discovered classes from a holistic as well as a by-parts perspective. Using such characterizations, we present a method to classify a new activity to one of the discovered activity-classes, and to automatically detect whether it is anomalous with respect to the general characteristics of its membership class. Our results show the efficacy of our approach in a variety of everyday environments.

Using similarity scores from a small gallery to estimate recognition performance for larger galleries

We present a method to estimate recognition performance for large galleries of individuals using data from a significantly smaller gallery. This is achieved by mathematically modelling a cumulative match characteristic (CMC) curve. The similarity scores of the smaller gallery are used to estimate the parameters of the model. After the parameters are estimated, the rank 1 point of the modelled CMC curve is used as our measure of recognition performance. The rank 1 point (i.e.; nearest-neighbor) represents the probability of correctly identifying an individual from a gallery of a particular size; however, as gallery size increases, the rank 1 performance decays. Our model, without making any assumptions about the gallery distribution, replicates this effect, and allows us to estimate recognition performance as gallery size increases without needing to physically add more individuals to the gallery. This model is evaluated on face recognition techniques using a set of faces from the FERET database.

Predicting large population data cumulative match characteristic performance from small population data

Given a biometric feature-space, in this paper we present a method to predict cumulative match characteristic (CMC) curve performance for a large population of individuals using a significantly smaller population to make the prediction. This is achieved by mathematically modelling the CMC curve. For a given biometric technique that extracts measurements of individuals to be used for identification, the CMC curve shows the probability of recognizing that individual within a database of measurements that are extracted from multiple individuals. As the number of individuals in the database increase, the probabilities displayed on the CMC curve decrease, which indicate the decreasing ability of the biometric technique to recognize individuals. Our mathematical model replicates this effect, and allows us to predict the identification performance of a technique as more individuals are added without physically needing to extract measurements from more individuals.

Temporal integration of multiple silhouette-based body-part hypotheses

A method for temporally integrating appearance-based body-part labelling is presented. We begin by modifying the silhouette labelling method of Ghost (Haritaoglu, Harwood, and Davis. 1998); that system first determines which posture best describes the person currently and then uses posture-specific heuristics to generate labels for head, hands, and feet. Our approach is to assign a posture probability and then estimate body part locations for all possible postures. Next we temporally integrate these estimates by finding a best path through the posture-time lattice. A density-sampling propagation approach is used that allows us to model the multiple hypotheses resulting from consideration of different postures. We show quantitative and qualitative results where the temporal integration solution improves the instantaneous estimates. This method can be applied to any system that inherently has multiple methods of asserting instantaneous properties but from which a temporally coherent interpretation is desired.

Relationship between identification metrics: expected confusion and area under a ROC curve

The mathematical relationship between the expected-confusion metric and the area under a receiver operating characteristic (ROC) curve is derived. Given a limited database of subjects and an identification technique that generates a feature vector per subject, expected confusion is used to predict how well the feature vector will filter identity in a larger population. Related is the area under a ROC curve that can be used to determine the probability of correctly discriminating between subjects given the feature vector. These two measures have different connotations, but we show mathematically and verify experimentally that a simple transformation can be applied to the expected confusion to find the probability of incorrectly discriminating between subjects, which is the complement of the area under a ROC curve. Furthermore, we show that as a function of the number of subjects, this transformed expected-confusion measure converges more quickly than direct calculation of the area under a ROC curve.