摘要 |
PROBLEM TO BE SOLVED: To prevent from erroneously selecting a high encoding rate due to an effect when an ambient noise with large hourly fluctuation in power is inputted in a voiceless section. SOLUTION: When a voice input part 1 receives an input voice, a short-term power operator 2 operates the power of the input voice at every prescribed time unit. An ambient noise power estimator 3 estimates the power of the ambient noise superimposed on the input voice, and a rate selection threshold value operator 4 operates a power threshold value group for selecting a voice encoding rate by using the estimated result. A power comparator 5 compares the power obtained by the short-term power operator 2 with the threshold value group obtained by the rate selection threshold value operator 4 to select one of suitable rate from plural voice encoding rates. An ambient noise characteristic estimator 6 estimates the characteristic of the ambient noise superimposed on the input voice, and lowers forcedly the output value of the short-term power operator when the ambient noise is provided with the large hourly fluctuation in power. |