We present an approach to recognizing faces with varying appearances which also considers the relative probability of occurrence for each appearance. We propose and demonstrate extending dimensionality reduction using locally linear embedding (LLE)

No abstract

No abstract

Fundamental to the generation of 3D audio is the HRTF processing of acoustical signals. Unfortunately, given the high dimensionality of HRTFs, incorporating them into dynamic/interactive virtual environment and gaming applications is computationally very demanding. This greatly limits the performance of such applications that incorporate real-time 3D audio. This paper examines the application of data reduction models to HRTFs. In particular, the locally linear Isomap, Locally Linear Embedding (LLE), and the globally linear Principal Components Analysis (PCA) dimensionality reduction tools are applied to the MIT HRTF dataset. Our motivation is to project the inherently highdimensional space inherent in HRTF measurements onto a lower dimensionality such that they can be incorporated into interactive virtual environments and gaming applications.

No abstract