Naoki Fukaya

Estimation of Driving Phase by Modeling Brake Pressure Signals

It is important for a driver-assist system to know the phase of the driver, that is, safety or danger. This paper proposes two methods for estimating the drivers phase by applying machine learning techniques to the sequences of brake signals. One method models the signal set with a mixture of Gaussians, where a Gaussian corresponds to a phase. The other method classifies a segment of the brake sequence to one of the hidden Markov models, each of which represents a phase. These methods are validated with experimental data, and are shown to be consistent with each other for the collected data from an unconstrained drive.

A rear-end collision warning system for drivers with support vector machines

The paper proposes a rear-end collision warning system for drivers, where how to evaluate the risk of collision is critical. In the proposed system, the time-to-collision assuming a constant relative acceleration is employed as a subjective index of risk based on the collected data in an unconstrained drive and the discussion from a theory of visual control. The system produces an alert when it detects anomaly of the index using the one-class support vector machine, which is trained with the aforementioned data. Some cross-validation experiments show the effectiveness of the system.

Online multibody factorization based on Bayesian principal component analysis of Gaussian mixture models

An online multibody factorization method for recovering the shape of each object from a sequence of monocular images is proposed. We formulate multibody factorization problem of data matrix of feature positions as the parameter estimation of the mixtures of probabilistic principal component analysis (MPPCA) and use the variational inference method as an estimation algorithm that concurrently performs classification of each feature points and the three-dimensional structures of each object. We also apply the online variational inference method make the algorithm suitable for real-time applications.