| 摘要 |
PURPOSE:To enable encoding which is small in quality degradation unless the encoding is impossible below a permissible noise level when information is less than a permissible amount by providing a quantized noise distribution which minimizes the loudness of a noise approximately when a voice is converted into a frequency range and then encoded. CONSTITUTION:An encoding means 1 and an information amount calculating means 2 are provided. Namely, an encoding means 1 performs the encoding so performed that Dm is nearly max(DWi(Si/Vi)<2/3>,Mi/Vi), where (i) is the number of a divided section that an (m)th frequency component Xm belongs to when a frequency component is divided at a ratio proportional to critical band width, Vi the number of frequency components in the divided section (i), Wi and Si the critical band width and spread spectrum nearby the divided section (i), Mi is a total permissible noise level assigned to the divided section (i), and D an area value. Then an information amount calculating means 2 finds the amount of information which is required for the encoding by the encoding means 1. Consequently, the encoding which is small in quality degradation is performed. |
