When encoding, index information is output indicating a set of coefficients, from among sets of predetermined coefficients corresponding to each sample position, for minimising the sum, with regards to each sample position, of the error between values of samples and values obtained by multiplying quantization values of the samples by coefficients corresponding to the position of each sample. When decoding, a plurality of values corresponding to an input vector quantization index is obtained and made into decoding values corresponding to the plurality of sample positions. Using sets made up of predetermined coefficients corresponding to the plurality of sample positions and showing the input index information, values corresponding to the sample positions are obtained by multiplying the decoding values by the coefficients, and then output.