ABSTRACT
The key mission of ambient diagnostics is to record, convert, compress, and discover patterns from massive continuous data streams. This chapter focuses on data transformation, which includes general feature descriptions and visualization of patterns. We discuss such key questions as
• How can we represent sequential features in a vector? • How can we transform spatiotemporal features to a matrix? • How can we transform features to a shape? • How can we convert a shape into numbers? • How can we represent features in a frequency domain? • How can we decompose an image into tiny self-similar features? • How can we reduce the dimensions of a feature space? • How can we automatically measure image quality?
