To solve such problem that the tone quality is degraded by improperly selecting a short block length or missing a quick rise of a signal in a device which divides an audio signal into blocks to encode it.
The audio signal is divided into segments, and signal levels of segments are calculated by a segment signal level calculator 10, and a quick rise of the signal is detected in a signal level change detector 30 by the change of signal levels of segments, and a quantization noise in segments preceding the segment where a quick rise of the signal has been detected is estimated by a quantization level estimator 40, and a masking level based on preceding segments is estimated in masking level estimators 50 and 51 to 5M by segments in which the quantization noise has been estimated, and the device is controlled by a block length decision device 70 so that the time of a block for encoding of the audio signal may be shortened in the case that the quantization noise is larger than the masking level.
USAMI AKIRA
IIMURA KATSUHIKO