Speech recognition apparatus and speech recognition method

A speech recognition apparatus for recognizing an input speech, comprising:

• distance calculation means for determining the distance to a sound source of said speech;
• storage means for storing plural sets of acoustic models produced from speech uttered by sound sources located at different distances;
• acquisition means for selecting tap coefficients for realizing an inverse filter having a frequency characteristic corresponding to the distance determined by the distance calculation means, from the plural sets of acoustic models for various distances stored in the storage means;
• filter means for filtering said speech using the tap coefficients acquired by the acquisition means;
• and
• speech recognition means for recognizing said speech filtered by the filter means, on the basis of a particular set of acoustic models.

A speech recognition method of recognizing an input speech, comprising the steps of:

• determining the distance to a sound source of said speech;
• storing plural sets of acoustic models produced from speech uttered by sound sources located at different distances;
• selecting tap coefficients for realizing an inverse filter having a frequency characteristic corresponding to the distance determined in the distance determination step, from the plural sets of acoustic models for various distances stored in the storing step;
• filtering said speech using the tap coefficients selected in the selection step;
• and
• recognizing said speech filtered in the filtering step, on the basis of a particular set of acoustic models.

A program for causing a computer to recognize an input speech, said program comprising the steps of:

• determining the distance to a sound source of said speech;
• storing plural sets of acoustic models produced from speech uttered by sound sources located at different distances;
• selecting tap coefficients for realizing an inverse filter having a frequency characteristic corresponding to the distance determined in the distance determination step, from the plural sets of acoustic models for various distances stored in the storing step;
• filtering said speech using the tap coefficients selected in the selection step;
• and
• recognizing said speech filtered in the filtering step, on the basis of a particular set of acoustic models.

A storage medium having a program stored thereon for causing a computer to recognize an input speech, said program comprising the steps of:

• determining the distance to a sound source of said speech;
• storing plural sets of acoustic models produced from speech uttered by sound sources located at different distances;
• selecting tap coefficients for realizing an inverse filter having a frequency characteristic corresponding to the distance determined in the distance determination step, from the plural sets of acoustic models for various distances stored in the storing step;
• filtering said speech using the tap coefficients selected in the selections step;
• and
• recognizing said speech filtered in the filtering step, on the basis of a particular set of acoustic models