To provide a device and a method for vector-quantizing and decoding a sound signal.

The vector quantizing device is composed of a code book group. a KLT part, first and second selection parts and a transmission part. The code book group is composed of a plurality of code books storing a code vector against a sound signal obtained by KLT. The code books are classified by KLT domain statistical values. The KLT part converts the inputted sound signal into a KLT domain. The first selection part selects the optimum code book from the code books based on a peculiar value set against a covariance matrix of the input sound signal obtained by KLT. The second selection part selects the optimum code vector based on distortion of the respective code vectors filled on the selected code book and the sound signal converted into the KLT domain by the KLT part.