Vector quantization of speech waveforms

Weighted Parzen windows for pattern classification

“…This mapping quantizes the transmission data. Reference [1] is an excellent source for further details if necessary.…”

Voronoi Projection-Based Fast Nearest-Neighbor Search Algorithms: Box-Search and Mapping Table-Based Search Techniques

“…1 Thus, the main complexity of algorithms based on the full Voronoi diagram arises from the preprocessing stage in constructing the Voronoi diagram of N input points which has an ⍀(N K/2 ) worst case time and storage, in addition to the cost required in organizing the Voronoi diagram for point location. While the basic computation of the Voronoi diagram invariably faces this intrinsic exponential complexity, the use of the diagram for fast nearest-neighbor search has further practical difficulties since the known search (pointlocation) algorithms which use the Voronoi diagram in higher-dimensional subdivisions have similar exponential dependence on dimension in their average and worst-case query time performance bounds and preprocessing and storage costs which grow exponentially with dimension [11][12][13]47].…”

“…Determine CЈ(x) ϭ 5c i : x ʦ B i 6: the candidate set of codevectors whose boxes B i contain the test vector x. 1 A Voronoi diagram in K-dimensions is a cell complex made up of convex polytopes which consist of faces of dimensions varying from 0 to K, such as vertex (0-dimension), edge (1-dimension), region (2-dimension), facet ((K Ϫ 1)-dimension), and cell (Kdimension). In general, a face in the cell complex is called a k-face, (0 Յ k Յ K), if it is of dimension k. Thus, vertex, edge, region, facet and cell are 0-face, 1-face, 2-face, (K Ϫ 1)-face, and K-face, respectively.…”