JPS60247697A - Voice recognition responder - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

DETAILED DESCRIPTION OF THE INVENTION [Technical Field of the Invention] The present invention relates to a voice recognition response device used in an information processing system using voice input.

[Technical background of the invention and its problems]

In recent years, the development of speech recognition and synthesis technology has been remarkable, and it has become possible, for example, to recognize continuous speech percentages and to recognize speech patterns for unspecified speakers, and on the other hand, to perform highly accurate speech synthesis.

Telephone voice response services that use such technology to provide various services over telephone public telephone lines, such as bank account balance inquiries, have been discussed, and their usefulness is attracting attention. By the way, the users of this system are an unspecified number of people; for example, there are people who are inexperienced with the system, such as elderly people and children, and there are also people who are experienced and use the system even if they use it every day. Despite this, in conventional systems, the content (format) of the voice response is fixed, and the time from when the user inputs voice until the voice response is output and the speed of the voice response are also constant. Therefore, it could not be said that it was easy for all users to use, and interaction between humans and machines was not smooth. For example, in a bank account balance inquiry service by telephone, when a user inputs an account number "123..." by voice over the phone, when an input request signal sound of "beep" is heard, the user first says "1". Then, after about 10 seconds, you will hear a voice response of "1" for N confirmation.

Then the user says "2". For users who are accustomed to this type of system, the response time is lengthy and irritating, and for users who are not accustomed to this type of system, it is difficult to understand the contents of the responses.

[Purpose of the invention]

An object of the present invention is to provide a voice recognition response device that enables smooth interaction between humans and machines.

[Summary of the invention]

The present invention includes an input means for inputting a voice signal, a voice detection means for analyzing the voice signal input from the input means as a voice pattern and detecting a voice response signal, and a voice detection means. A voice recognition means for recognizing matched voice patterns, and this voice i! a voice output means for outputting a voice response signal and a voice input request signal based on the g recognition result of the voice pattern output from the voice recognition means; and a voice input request signal output from the voice output means and the voice detection. A measuring means for measuring time data of the voice section signal outputted by the means, and the voice output means controls and outputs the voice response signal based on the time data outputted from the measuring means. It is something to do.

〔Effect of the invention〕

According to the present invention, since it becomes possible to give an appropriate response according to each user, interaction between a human and a machine can be carried out smoothly, and practicality for the user is improved.

[Embodiments of the invention]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a schematic configuration diagram of the first embodiment of the present invention, FIG. 2 is a schematic diagram illustrating the timing of voice input request P and voice input to voice response, and FIG. 3 is a schematic diagram of the first embodiment of the present invention. FIG. 3 is a processing flow diagram of the first embodiment. In this first embodiment, the time from when a voice input request is output to when voice is input by the user is measured, and the voice response is controlled and output according to the measured time. Each block within the dotted line in FIG. 1 constitutes a voice recognition response device, whose input and output are connected to a service terminal (not shown). For example, when a user enters a predetermined phone number on a telephone, which is a service terminal, a bank account inquiry service is performed through the handset. When the beep signal sounds, please answer the account numbers one by one.'' is sent to the voice response output unit 7. The present invention is applied to voice input and response from this point onwards.

First, a voice input request signal P is outputted to a terminal (not shown) from the voice input start request s8 in FIG. 1, but this is also sent to the timing measuring section 3 at the same time (step 11 in FIG. 3).
. The timing measuring section 3 measures the time point when this audio input request signal P is input. On the user side, when the user hears the signal tone ``beep'' which is the voice input request signal, the user inputs the voice ``1'' from the receiver (step 12 in FIG. 3). When this input audio signal is input to the analyzer 1, A/D conversion, spectrum analysis processing, etc. are performed, and the input audio signal is converted into a series (sound) of feature parameters (3rd Figure step 13). The speech section detection section 2 uses the energy information of the feature parameter series (speech pattern) output from the analyzer 1 to detect the beginning and end of the speech pattern and cut out the speech section. (3rd
Figure step 14). When the voice section detection section detects the start and end of the voice pattern, it sends the start and end signals to the timing measurement section 3, respectively. Timing measurement section 3
Now, measure the time point at which this start signal is input, and calculate the time T from the time point at which the voice input request signal P is inputted to the time point at which the start point signal is inputted (see TI+ in Figure 2).
Figure 3 step 15). On the other hand, the speech section detection section 2 sends the cut out speech pattern (characteristic parameter series) to the speech recognition section 4 . The speech recognition unit 4 recognizes the input speech pattern by using a speech dictionary registered in advance in the dictionary memory 5 (step 16 in FIG. 3).
'). This recognition is performed, for example, by a similarity calculation method. The voice pattern recognition result by the voice recognition unit 4 is sent to the voice response control unit 6 together with T1 calculated by the timing measurement unit 3. The voice response control section 6 controls and outputs the voice response bird based on the length of T (step 17 in FIG. 3), and there are three ways of controlling this:

(1) The voice response control section 6 is connected to the timing measurement section 3 by T1.
Also input the data at the time when the termination signal is input. Then, the voice response control unit 6 determines the time Tll from the time of inputting the terminal signal to the time of outputting the voice response @bird according to the length of T1.
The length ('rs in Figure 2) is variably controlled. Tsuma DTt
If the time length is shorter than the predetermined length of time, it means that the user has uttered the voice as soon as he/she hears the "bee" signal sound, and it is assumed that the user is busy with the system or is in a hurry. Therefore, it is necessary to output the response voice to the terminal as soon as possible (step 18 in FIG. 3), so the time length of T3 is made shorter than the predetermined length. Also, if T1 is longer than the predetermined time length, the user may not have uttered the voice after a short delay after hearing the "bee" signal sound, and the user may not be familiar with the system. Or maybe you have time. Therefore, it is necessary to output the response voice to the terminal later (Step 19 in Figure 3).
), the time length of T3 is made longer than the default length.

The voice response control unit 6 generates a voice response according to the length of T.
The output time (response speed) is variably controlled (Fig. 2 T
4). In other words, if TI is shorter than a predetermined time length, the response speed is increased and a voice response bird is output for the above-mentioned reason. If T is longer than the predetermined time length, the response speed is slowed down and the bird is output for the reason described above. At this time, when the voice response 4 is output by the rule synthesis method, various parameters (accent, etc.) for rule synthesis are output.

control the speed (pitch, etc.). Further, when R2 is output by the recording/editing method, the response speed is controlled by selecting pre-recorded words or speech segments having different speaking speeds.

(m) The voice response control unit 6 controls the content (expression format) of the voice response bird according to the length of T1 (bird in FIG. 2).

For example, if the user utters "1" after hearing the dial tone "bee", when outputting a voice response bird to confirm this, the length of T1 is longer than the predetermined time length. If it is short, a response of "1" is output for the reason mentioned above. If T1 is longer than the predetermined time, the answer will be ``It's 1 minute.''

” is output. The voice response control unit 6 accepts “1” as the recognition result of the input voice pattern from the voice recognition unit 4, but outputs it by changing the expression format of the voice response for confirmation of “1”. It is something.

ζ then (l l (II l (iii) by T
s I T41 Once the control method of R1 is determined (third
(Step 20 in FIG. 3), the voice response output section 7 outputs the voice response 4 according to the instruction from the voice response control section 6 (Step 21 in FIG. 3).

In this embodiment configured in this way, as shown in the schematic diagram of FIG.
Depending on the situation, the time T3 from the voice input to the voice response bird is changed, the response time T4 to the voice response is changed, and the form of expression for the voice response is changed, so it is easy to become familiar with how to use the system. For users who are in a hurry or who are in a hurry, you can shorten the response time, respond quickly, or make the content concise. You can take longer to respond, respond in a slower tone, and be more detailed. Furthermore, it is also possible to perform the controls in (i), (if), and (iii) in combination by the voice response control section, rather than individually.

By doing this, it is possible to facilitate dialogue between humans and machines.

Next, a second embodiment of the present invention will be described with reference to the drawings. FIG. 4 is a schematic configuration diagram of a second embodiment of the present invention, and FIG. 4 is a processing flow diagram of the second embodiment. In the second embodiment, as shown in FIG. 2, the time T1 from the input request signal to the start of voice input and the utterance time T2 of the human voice R are detected to control the output to the voice response. It is something. The configuration shown in FIG. 4 is compared with the configuration shown in FIG.
The analysis section 1, speech section detection section 2, timing measurement section 3, speech recognition section 4, dictionary memory 5, speech response control section 6, speech response output section 7, and speech input start request section 8 are the same;
A speech time measuring section 9 is added to these. The voice section detection unit 2 detects the beginning of the input voice pattern,
When the end is detected, these start end signals and end signals are sent to the timing measurement section 3 and also to the speech time measurement section 9.

The speech time measurement section 9 measures the time T from the time when the start signal is input to the time when the end signal is input (step 22 in FIG. 5). In the voice response control section 6, the timing measurement section 3
The T1 requested by the user and the T2 requested by the speech time measuring section 9 are input. The voice response control unit 6 compares T2 with a predetermined time length (step 23 in FIG. 5).
The output of the voice response is controlled according to the result and the comparison result of the time length of T mentioned above. In other words, if the utterance time T2 is shorter than the predetermined time length, it is assumed that the user is accustomed to the system or is in a hurry and is therefore uttering quickly, and the time T3 + T4 shown in FIG. 2 is shortened as described above. or keep the content of the voice response bird concise (5th
Figure step 24). If T is longer than the predetermined time length, it is assumed that the user is not familiar with the system or has time to spare, and that the user uttered Yutsu〈shi late in the day. T, , T.

or make the voice response more detailed (step 25 in Figure 5).

In this manner, according to the second embodiment, the time TI shown in FIG.
and T! Since the output of the voice response bird is controlled in accordance with the result, it is possible to provide a response that better reflects the user's personality and the situation at the time of vocalization, compared to the first embodiment. The naturalness of the interaction between the user and the machine can be further enhanced.

First mentioned above. In the second embodiment, the voice input start request signal P is output from the voice input start request section 8, but it is outputted from the voice response output section 1, and the response voice and the input request voice are continuously output. You can also output it. In other words, the user's voice and the machine's response are performed one after another (dotted line in the flowchart in Figure 5).
. FIG. 6 is a schematic diagram illustrating the timing of a response voice including an input request and an input voice. In this figure, R
o, R-i, R4, Re are response voices including each input request, ``+, '%, Rs are input voices from the user.For example, in the balance inquiry service mentioned above, RO
Please enter the numbers for your account number one by one." Th+
rx" - "It's 1. It's 9 minutes. Please give me the next number."
It is called Br2J. Even when the response output method is modified in this way, the time T3 from the response voice to the input voice is the same as in the second embodiment. By measuring Tll, T0, and the speaking time of the input voice T2 + T6 + T10,
Time from input voice to response voice T8 + T7 +
T11, response voice utterance time T2 + T6 + T
lG + TI2, response voice content RO, Rt, R
4 and 6 can be changed. By modifying the above-described embodiment in this way, voice input and response can be performed quickly, and the cost of using the line can be further reduced, resulting in tremendous economic value.

Similarly, the present invention is not limited to the above embodiments. For example, the timing measuring section may measure both times T1 and T. Alternatively, the response output may be changed after the user's personality can be clearly detected by measuring the history of the time from the input request signal to the input voice, that is, by measuring the time period between the input request signal and the input voice. Furthermore, the speech time may be measured by measuring the speech rate, or may be performed by using not only voice as a response output (CRT, printer, etc.).Input speech recognition processing and speech synthesis methods may be performed using various conventionally known methods. The following method may be adopted as appropriate.In short, the present invention can be implemented with various modifications without departing from the gist thereof.

[Brief explanation of the drawing]

FIG. 1 is a schematic configuration diagram of the first embodiment of the present invention, FIG. 2 is a schematic diagram of the timing of an input request, input voice, and response voice, FIG. 3 is a processing flow diagram of the first embodiment, and FIG. FIG. 4 is a schematic configuration diagram of a second embodiment of the present invention, FIG. 5 is a processing flow diagram of the second embodiment, and FIG. 6 is a schematic diagram of the timing of a conversational continuous input response format. 1.Analysis section 2.Speech section detection section 3..Timing measurement section 4..Speech recognition section 5 Dictionary memory 6..Speech response control section 7..Speech response output section 8
．． Voice input start request section 9...Speech time measurement section Agent Patent attorney Noriyuki Chika (and 1 other person) Fig. 1 Fig. 2 Fig. 3 Fig. 4 Fig. 6

Claims (1)

[Scope of Claims] voice recognition means for recognizing the voice pattern analyzed by the detection means; and voice output means for outputting a voice response device and a voice input request signal based on the recognition result of the voice pattern output from the voice recognition means. , the time from the time when the voice input request signal output from the voice output means is input to the time when the start signal of the voice section signal output from the voice detection means is input; measuring means for measuring either or both of the time from the time when the start signal of the voice section signal is input to the time when the end signal is input; A voice recognition response device that outputs a voice response signal by changing the manner in which the voice response signal is generated based on the following information. 2. The voice recognition response device according to claim 1, wherein the time period is one or more of the times from the time when the end signal of the voice section signal is input to the time when the voice output means outputs the voice response signal.