The present invention relates to a strong voice recognition method using independent vector analysis and model-based feature enhancement. The voice recognition method comprises the steps of: (a) estimating an independent vector analysis (IVA) target voice signal and an IVA noise signal by analyzing an independent vector for the signals input through a microphone; (b) setting up the estimated IVA target voice signal and IVA noise signal respectively as an input signal and a noise signal, respectively, and estimating a target voice signal by applying a method of model-based feature enhancement, thereby improving the accuracy of voice recognition.
展开▼