Speech dialogue system in which a recognition and understanding process, application process, and voice input response are performed simultaneously with voice input

A speech dialogues system comprising:

• voice analyzing means for converting a voice input signal into a time series feature vector, the converting being done in synchronism with voice input;
• voice detecting means for detecting a leading end and a trailing end of a voice period in said voice input signal on the basis of the time series feature vector, the detecting being done simultaneously with the voice input;
• recognition and understanding means for translating the time series feature vector into a meaning expression for each voice period, and for outputting the meaning expression in time series simultaneously with the voice input;
• input dividing means for dividing said meaning expression into partial strings respectively aggregated in meaning and for outputting each of said partial strings simultaneously with the voice input;
• and
• application executing means for performing a corresponding application process for each of said partial strings of said meaning expression simultaneously with the voice input, the corresponding application process being based on the corresponding meaning of said each of said partial strings,
• said voice detecting means outputting a leading end detection signal to said application executing means when the leading end of said voice period is detected;
• said application executing means having mutually distinct first and second processes for mutually different process content with respect to the partial strings of the meaning expression input from said input dividing means, said application executing means being responsive to said leading end detection signal input during execution of said first process for a currently processed partial string;
• and
• when the leading end detection signal is input during execution of said second process, said application executing means interrupts execution of said second process and executes said first process for a next one of the partial strings.

A speech dialogues system comprising:

• voice analyzing means for converting a voice input signal into a time series feature vector, the converting being done in synchronism with voice input;
• voice detecting means for detecting a leading end and a trailing end of a voice period in said voice input signal on the basis of the time series feature vector, the detecting being done simultaneously with the voice input;
• recognition and understanding means for translating the time series feature vector into a meaning expression for each voice period, and for outputting the meaning expression in time series simultaneously with the voice input;
• input dividing means for dividing said meaning expression into partial strings respectively aggregated in meaning and for outputting each of said partial strings simultaneously with the voice input;
• and
• application executing means for performing a corresponding application process for each of said partial strings of said meaning expression simultaneously with the voice input, the corresponding application process being based on the corresponding meaning of said each of said partial strings,
• said voice detection signal outputting a trailing end detection signal to said input dividing means when the trailing end of said voice period is detected;
• and
• said input dividing means outputting said meaning expression for said each voice period accumulated before input of said trailing end detection signal upon reception of said trailing end detection signal as the partial strings aggregated in meaning.

A speech dialogeustic processing method comprising the steps of:

• converting a voice input signal into a time series feature vector, the converting being done in synchronism with voice input;
• detecting a leading end and a trailing end of a voice period in said voice input signal on the basis of the time series feature vector, the detecting being done simultaneously with the voice input;
• translating the time series feature vector into a meaning expression for each voice period and outputting the meaning expression in time series simultaneously with the voice input;
• dividing said meaning expression into partial strings respectively aggregated in meaning and outputting each of said partial strings simultaneously with the voice input;
• and
• performing a corresponding application process for each of said partial strings of said meaning expression simultaneously with the voice input, the corresponding application process being based on the corresponding meaning of said each of said partial strings.

A speech dialogeustic processing method comprising the steps of:

• converting a voice input signal into a time series feature vector, the converting being done in synchronism with the voice
• detecting a leading end and a trailing end of a voice period in said voice input signal on the basis of the time series feature vector, the detecting being done simultaneous with the voice input;
• translating the time series feature vector into a meaning expression for each voice period and outputting the meaning expression in the series simultaneously with the voice input;
• dividing said meaning expression into partial strings respectively aggregated in meaning and outputting each of said partial strings simultaneously with the voice input;
• performing a corresponding application process for each of said partial strings of said meaning expression simultaneously with the voice input the corresponding application process being based on the corresponding means of said each of said partial strings;
• outputting a leading end detection signal to said application executing means when the leading end of each of said voice periods is detected;
• and
• said application executing step having mutually distinct first and second processes for mutually different process content with respect to the partial strings of the meaning expression input from said input dividing means, said application executing step is responsive to said leading end detection signal input during execution of said first process for a currently processed partial string;
• and
• when the leading end detection signal is input during execution of said second process, said application executing means interrupts execution of said second process and executes said first process for a next partial string.