To estimate correctly the number of signal sources in an actual environment.
A frequency domain conversion part 20 converts observation signals xj(t) (j={1, ..., M}) into time series data Xj(f, τ) of each frequency, and a signal separation part 31 generates separation signals Yi(f, τ) from the time series data Xj(f, τ) and stores them in a memory 10. Then, a power calculation part 32 determines a power value of each separation signal Yi(f, τ) and stores it in the memory, and an envelope correlation calculation part 33 calculates an envelope correlation value to the time difference Δτ between different separation signals Yi(f, τ) and stores it in the memory. A determination part 34 compares the power value and the envelope correlation value of each separation signal Yi(f, τ) with parameters thnoise, threv and thcor showing each threshold of a reverberation level and the envelope correlation value, and determines whether the separation signal Yi(f, τ) is a source signal component or not.
MUKAI MAKOTO
ARAKI AKIKO
MAKINO SHOJI
Minoru Inagaki
Next Patent: METHOD FOR CALCULATING SURGE PROPAGATION VELOCITY IN FAULT LOCATION SYSTEM