Regression-based index assignment algorithms
Jen-Der Day, L. Thomas · 2002
The paper looks at index assignment algorithms which seek to minimize channel distortion on noisy binary symmetric channels for equiprobable scalar and vector quantizations. An eigenspace index assignment algorithm (EIA) for a scalar quantization model is proposed which depends on a regression calculation and a sorting algorithm. Then, a vector eigenspace index assignment algorithm (VEIA) which extends the EIA for scalar quantization to vector quantization is proposed. The proposed algorithms are compared with the Binary Switch Algorithm (BSA) on a voice digitization in North American Telephone Systems CCITT and the first-order Markov-Gauss codebooks. In terms of CPU time and signal-to-noise ratio performances, they are shown to be fast and effective.