Constructing Better Partial SumsBased on Energy- MaximumCriterion forFastEncoding ofVQ

Inavector quantization (VQ)framework, oneof thekeyproblems initspractical applications istheencoding speed. InordertospeedupVQ encoding process, itismost important toavoidcomputing unnecessary k-dimensional (k-D) realEuclidean distances fortheobviously unlikely candidate codewords. Themean,thevariance andthetwopartial sumsofa k-Dvector havealready beenproposed astheeffective features in theprevious worksinordertorealize arejection totheunlikely codeword byusing just alittle computational cost. Itisclear that themeanandthevariance ofak-Dvector areconstant butthe twopartial sumsofak-Dvector arenotconstant depending on howtheyhavebeenconstructed. Therefore, howtoconstruct two better partial sumsforfastVQ encoding becomesimportant. Instead ofusingfixed thefirst halfvector andthesecondhalf vector criterion thathasbeenintroduced intheprevious works, thispaperproposes a new energy-maximum criterion to construct twobetter partial sumsforak-Dvector. Mathematical analysis andexperimental results confirmed thattheproposed criterion ismuchmoreeffective forfast VQ encoding compared tothefixed criterion usedintheprevious works. Inaddition, itis veryeasytousetheenergy-maximum criterion inpractice.