JPH02250097A - Speech recognition system - Google Patents

Abstract

PURPOSE:To prevent the recognition rate from decreasing by inhibiting a speech analyzing means from performing speech analytic processing unless a speaking person inputs a voicing start signal to the speech analyzing means by a voicing indication means. CONSTITUTION:When the speaking person voices a recognizable word, the start of voicing can be transmitted to a speech recognition system by the voicing indication means 9. Namely, when the switch 11 of the voicing indication means 9 is turned on, an 'H' signal (voicing start signal) is inputted to the speech analyzing means B10 and the speech analytic processing is performed. Further, when the switch 11 of the voicing indication means 9 is turned off, an 'L' signal is inputted to a speech analyzing means B10 and the speech analyzing processing is not performed. Thus, a period wherein the speaking person voices the recognizable word is determined to suppress the voicing of words other than the recognizable word and increase the recognition rate.

Description

[Detailed Description of the Invention] Industrial Application Field The present invention relates to a speech recognition system for recognizing word speech input by specific speakers and unspecified speakers, and performing various processes based on the speech content. In particular, the present invention relates to a voice recognition system for vending machines that supports unspecified speakers.

2. Description of the Related Art Conventionally, in a voice recognition system for vending machines such as vending machines for cup beverages (hereinafter simply referred to as cup vending machines), the user first presses the microphone 1, as shown in FIG.
The input voice is analyzed by the voice analysis means A2 and a voice pattern is extracted. BPF (Band Path Filter) using a group of band pass filters for analysis
A method is used to sample the analysis results on the time and frequency axes and digitally process the intensity. The standard pattern storage means 3 stores in advance, as standard patterns, audio patterns of a plurality of discrete words uttered by a large number of unspecified speakers, extracted using a similar method. However, the words stored as standard patterns here are the names of flavors sold in cup vending machines (product names of coffee, June, etc.) and some response words (``Yes'', ``6 No'', ``Hot'', ``Ainu'', etc.). ”, etc.). Then, the standard pattern selection means 4 selects the standard pattern closest to the input audio pattern from among the standard patterns as D P (Dynamic Prog.
ramming) is used to select and recognize speech using the 7-timing method.

The DP matching method, which is translated as dynamic programming, is a method of mathematical programming proposed by Bellman in the United States in 1967, and is applied to the optimization of multi-stage decision processes. The method involves making a certain decision (control) at each stage to transform the state, and then evaluating the quality of the control in the process until the desired state is reached.
The function that evaluates the badness is maximized/minimized. In addition, when the speech recognition system is compatible with a specific speaker, the speech pattern of the recognized word uttered by the specific speaker is registered in the standard pattern storage means 3, and on the other hand, when the speech recognition system is compatible with an unspecified speaker, Some representative speech patterns of recognized words uttered by many speakers are registered.

The utterance guiding means 5 is composed of a voice synthesizing means, and performs voice guidance to encourage the user to utter according to the processing of the control means 6, which will be described later. For example, when choosing a flavor, say, ``Welcome.

What would you like. ” to prompt the user to say flavored salt. However, the flavored salts are clearly marked on the panel on the front of the cup vending machine, and the user must select one flavored salt of their choice and say it out loud.

The control means θ instructs the utterance guidance means 5 to utter the guidance voice according to the process, recognizes the words uttered by the user from the standard pattern selected by the standard pattern selection means 4, and uses the recognition result to This controls the subsequent operation of the cup vending machine. Further, 8 is a coin receiving means for receiving coins and refunding 4 sen, and 8 is a beverage discharging means for pouring the selected flavor into a cup and carrying it out.

Problems to be Solved by the Invention However, with the above method, the period during which the speaker utters recognizable words cannot be determined.
When the speaker is unsure about flavor selection, there is a possibility that the speaker may utter a word other than the recognizable word, such as ``Um, (flavor salt)''. For this reason,
There was a problem in that the recognition rate was reduced due to incorrect recognition. The present invention solves the above-mentioned conventional problems, and provides a speech recognition system that suppresses the utterance of words other than recognizable words and increases the recognition rate by determining the period during which the speaker utters recognizable words. The purpose is to do something.

Means for Solving the Problems In order to solve the above-mentioned problems, the speech recognition system of the present invention includes a standard pattern storage means that stores a group of standard patterns of a plurality of discrete word sounds, and a standard pattern storage means that analyzes the speech of a speaker and generates speech patterns. A voice analysis means for extracting a voice pattern, a voice analysis means for extracting a voice pattern by making a sound based on the voice start signal indicating that the speaker has started speaking, and a standard pattern closest to the voice pattern extracted by the voice analysis means. The present invention includes standard pattern selection means for selecting a word from the standard pattern group, and utterance guidance means for guiding a speaker to utter the word.

Operation The present invention, with the above-described configuration, enables the speech analysis means to notify the speech recognition system of the start of speech when the speaker pronounces a recognizable word.

Embodiment Hereinafter, a speech recognition system according to an embodiment of the present invention will be described with reference to the drawings. In this embodiment, a voice recognition system for unspecified speakers is applied to a cup vending machine. However, among the configuration requirements, those with the same configuration as the conventional example are given the same numbers and explanations are omitted.

FIG. 1 shows a functional block diagram of a speech recognition system in an embodiment of the present invention. Reference numeral 9 denotes a voice analysis means, which outputs a generation start signal when a speaker utters a recognizable word. Reference numeral 10 denotes a voice analysis means B, which analyzes the voice input to the microphone 1 and extracts a voice pattern when the voice generation start signal is input to the voice analysis means S. FIG. 2 shows a circuit diagram of the voice analysis means 90 in an embodiment of the present invention. 11 is a switch, and 12 is a resistor. When the speaker turns on the switch 11, an "H" signal is input to the voice analysis means B10.

The table shows an operation relationship table between the voice analysis means 9 and the voice analysis means B10. As shown in the table, when the switch 11 of the voice analysis means 9 is turned on, an H" signal (voice start signal) is input to the voice analysis means B10,
Voice analysis processing is performed. Further, the voice analysis means 9
When the switch 11 is turned off, the voice analysis means B
An "L" signal is input to 10, and no voice analysis processing is performed.

The vending operation of the voice recognition system for a cup vending machine configured as described above will be explained using the flowchart shown in FIG.

In FIG. 3, first in step 201, the coin receiver determines whether a coin has been inserted into the taking means 7, and if a coin has been inserted, the process proceeds to step 202. Step 20
In step 2, the voice guiding means 5 is used to guide the user by saying, ``Welcome, what would you like to say?''.

Then, in step 203, the process continues to wait until there is an audible notification signal from the occurrence notification means 9. When there is a voice announcement signal, in step 204, the voice pattern selection means 4 selects a standard pattern closest to the input voice pattern from among the standard patterns stored in the standard pattern storage means 3 and selects the name of the voice pattern. Recognize. In step 205, it is determined whether the recognition result in step 204 is appropriate or not, and in the case of rejection, the process proceeds to step 206.
The voice guidance means 6 is guided to "Please answer again" and the process returns to step 203. On the other hand, if the request is not rejected, the process advances to step 207.

At the completion of step 20, the operation after the flavor recognized at step 204 is branched, but in this embodiment, it is assumed that coffee has been recognized, and the operation when another flavor is recognized is that of coffee. Since this is the same as in the case, the explanation will be omitted. Next, step 208
Then, the voice recognition means 6 confirms, ``It's white coffee,'' and prompts the customer to respond.Then, in step 209, the wait is continued until there is a voice notification signal from the occurrence notification means 9.There is a voice notification signal. Then, in step 210, the answer of yes or no is recognized in the same way as for flavored salt.In step 211, it is determined whether the recognition result in step 210 is appropriate, and if it is rejected, step 2 is performed.
Return to step 08; otherwise, proceed to step 212.

In step 212, if the answer recognized in step 210 is yes, the process advances to step 213; if the answer is no, the process returns to step 208. In step 213, the coffee is poured into a cup and carried out using the beverage carrying means 8. Then, in step 214, if there is change, the coin receiver refunds the change to the taking means 7, and finally, in step 215, it sends a message to the voice guiding means 5 saying "Thank you very much." ” and complete the series of actions (sales).

As described above, according to the present embodiment, the present invention is configured such that the voice analysis means does not perform voice analysis processing unless the speaker inputs a voice start signal to the voice analysis means using the voice analysis means. No, when the speaker utters a recognizable word, the voice analysis means is used to signal the start of the utterance, so that, for example, when he or she is unsure about choosing a flavor, he or she can say, "Um, (flavor salt)." The number of utterances that lead to erroneous recognition is drastically reduced, making it possible to prevent a drop in the recognition rate.

In addition, since the speaker can decide the period during which he/she will utter recognizable words, the speaker can proceed with the dialogue at his/her own pace with peace of mind. Moreover, when the speaker is in a hurry, if the order of vocalizations is correct, the product can be quickly purchased without having to wait for the vocalization guidance, so the effect is great.

Effects of the Invention As described above, the speech recognition system of the present invention includes: a standard pattern storage means that stores a group of standard patterns of a plurality of discrete word sounds; a speech analysis means that analyzes a speaker's speech and extracts a speech pattern; A voice analysis means for sounding the start of voice by a voice start signal and extracting a voice pattern, and selecting a standard pattern closest to the voice pattern extracted by the voice analysis means from the group of standard patterns. By providing a standard pattern selection means and a utterance guidance means for guiding the speaker to utter a word, the speaker can decide when to start uttering a word, so that words other than those that the speaker can recognize can be used. The number of words subjected to phonetic analysis processing is drastically reduced, making it possible to prevent a drop in the recognition rate.

[Brief explanation of drawings]

FIG. 1 is a functional block diagram of a speech recognition system according to an embodiment of the present invention, FIG. 2 is a circuit diagram of the speech analysis means of FIG. 1, and FIG. 3 is an example of the operation of the speech recognition system according to an embodiment of the present invention. The flowchart shown in FIG. 4 is a functional block diagram of a conventional speech recognition system. 3...Standard putter memory means, 4...Standard pattern selection means, 6...Voice guidance means, 9
...Group/Voice analysis means, 10...Voice analysis means B0 Name of agent Patent attorney Shigetaka Awano and 1 other person

Claims (1)

[Claims] a standard pattern storage means that stores a group of standard patterns of a plurality of discrete word sounds; a voice analysis means that analyzes the voice of a vocalization and extracts a voice pattern; a standard pattern selection means for selecting a standard pattern closest to the voice pattern extracted by the voice analysis means from the standard pattern group; and a voice guidance for guiding the speaker to pronounce the word. A voice recognition system consisting of means.