摘要 |
PROBLEM TO BE SOLVED: To provide a voice analyzing device in which an adverse effect caused by noise components is eliminated and a stable analysis is conducted even though the noise components are changed in the middle of a sound interval. SOLUTION: A microphone 1 receives voices of a speaker and a microphone 2 takes in background noise. Analyzing circuits 3 and 4 extract feature parameters from the signals inputted into the microphones 1 and 2. A voice detector 5 compares the parameters extracted by the circuit 3 with a prescribed threshold level to detect the beginning of voices. A noise analyzer 6 stores the parameters extracted by the circuit 3 as a compensation value if it is judged that no voice is inputted. If the detector 5 detects a sound interval and noise components are changed, the analyzer 6 updates the stored parameters with the feature parameters extracted by the circuit 4. A compensating circuit 7 compensates for the feature parameters extracted by the circuit 3 using the compensation value from the analyzer 6. |