To attain aimed subjective quality and to achieve a low bit rate.

Identifier α of a voice section and a non-voice section, an average power ratio β of the both sections, and a linear prediction gain γ are calculated from an input voice. Entry of the α, β, γ is performed, a table 3 is referred, relation among α, β, γ, anyone of encoding devices A, B, C, a bit rate, and subjective quality are stored in the table 3, an encoding device having aimed subjective quality and the lowest bit rate is selected an input voice is encoded by the selected encoding device (anyone of A, B, C), the encoded information and the selected information are multiplexed and outputted.