Speech-input control device

• US 5,826,233 A
• Filed: 05/15/1997
• Issued: 10/20/1998
• Est. Priority Date: 05/16/1996
• Status: Expired due to Fees

• Current Assignee: Honda Giken Kogyo Kabushiki Kaisha (Honda Motor Company)
• Original Assignee: Honda Giken Kogyo Kabushiki Kaisha (Honda Motor Company)
• Inventors: Matsumoto, Yoshiyuki
• Primary Examiner(s): Hudspeth, David R.
• Assistant Examiner(s): Opsasnick, Michael N.
• Application Number:
  US08/857,042
• Time in Patent Office: 523 Days
• Field of Search: 704/231, 704/246, 704/251, 704/270, 704/275
• US Class Current: 704/275
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 2015/225 Feedback of the input speech

Abstract

A speech input control device for recognizing an input speech and giving a specified control instruction to a target device, which device is provided with a portion for detecting that a word inputted by voice is unregistered in a voice word dictionary, a portion for extracting a feature of the detected unregistered words and storing the extracted feature, a portion for counting the input frequency of each unregistered word and renewably storing the counted input frequency, and a portion for changing a response message in response to an unregistered word depending to its input frequency and which is capable of adaptively responding by avoiding outputting the same response message such as "The input cannot be recognized" and "Correct input is required" or the like whenever the user'"'"'s meaningless word or habitual phrase is inputted thereafter.

Claims

1. A speech input control device having a speech recognition processing portion for extracting a feature of each word inputted by voice, comparing the extracted feature of each input word with stored features of previously registered words in a voice word dictionary, recognizing the input word by its similarity to the registered word and giving a control instruction to an objective device to be controlled, an improvement comprising providing the device with unregistered word detecting means for detecting a word that is included in an input speech and has not yet been registered in the voice word dictionary, storing means for storing the detected unregistered word inputted by voice together with its feature pattern, input-frequency counting means for counting an input frequency of each unregistered word and renewably storing the counted input-frequency of each unregistered word, and response message changing means for changing a responding program according to the counted input-frequency of an inputted unregistered word when said word is inputted.

2. A speech input control device as defined in claim 1, characterized in that the response message changing means gives a full warning message in response to a voice input of an unregistered word whose input-frequency is less than a preset minimal frequency threshold, neglects, not giving any response message, a voice input of an unregistered word whose frequency exceeds a preset maximal threshold, and gives a simple warning message in response to a voice input of an unregistered word whose input-frequency is more than the minimal threshold and less than the maximal threshold.