PURPOSE: A method and apparatus for tracing vocal track resonance using a nonlinear estimator and target guide constraint are provided to allow accurate formant trace even in a non-resonant speech region and easily calculate a simulated characteristic vector during run time. CONSTITUTION: Functions of respective vocal track resonant frequencies are calculated and the calculated results are summed up to form a simulated characteristic vector, to thereby map a vocal track resonant frequency set to the simulated characteristic vector. The simulated characteristic vector is applied to a model together with an input characteristic vector to determine probability that the vocal track resonant frequency set exists in a speech signal.
展开▼
