JPS6291047A - Voice recognition telephone set - Google Patents

Abstract

PURPOSE:To distinguish an unregistered state from a recognition disable state by using a switch means to switch the changeover at voice registration and talking based on a signal from a control section, using a count means at voice registration to add number of registrations, giving an output of the count means to the control section and displaying the unregistered state on a display section. CONSTITUTION:A hook switch 4 detects the off-hook state at the voice recognition, and when a counter 12 is zero, that is, no standard voice pattern is registered in a standard voice pattern memory, a control section 5 does not close an SW2 7 and displays the unregistration of the standard voice pattern on the display section 16. Thus, useless voice input regardless of unregistration is avoided and the recognition disabling state caused due to noise is prevented.

Description

DETAILED DESCRIPTION OF THE INVENTION (A) Field of Industrial Application The present invention relates to a voice recognition telephone that automatically sends out a dialing signal using a voice+m recognition means.

(o) Conventional technology In recent years, so-called multi-function phones with many functions have become popular, and voice recognition is one such function that has attracted attention. ) is also known (Japanese Unexamined Patent Publication No. 58-10
(See Publication No. 8849).

According to this type of voice recognition telephone, if the phone number of the other party (called party) and the corresponding voice such as the name of the called party are registered in advance, the registered called party can be recognized by speaking to the transmitter. Just say the name and a dial signal will be sent out. Therefore,
You can easily call the called party without having to press a button or turn a dial.

However, in voice recognition telephones, because the standard voice pattern is registered, a state may occur where the standard voice pattern is not registered (unregistered), or even if it is registered, it may be registered due to a power cut or an incorrect operation. The standard speech patterns that were used may disappear.

On the other hand, even if standard voice patterns are registered, an unrecognizable state occurs when there is no pattern whose distance between patterns obtained as a result of pattern matching between the input voice pattern and the standard voice pattern is equal to or less than a predetermined threshold. This occurs, for example, due to the presence of noise in the voice.

However, in the conventional voice recognition telephone,
Since it is not possible to distinguish between the unregistered state and the unrecognized state, you may input voice without knowing that it is in the unregistered state and realize that it is stuck on the display, or if it is not displayed, the oscillation occurs after repeatedly inputting voice. There is a problem that people only notice it when they hear a sound or the like.

f→ Problems to be Solved by the Invention The present invention solves the above-mentioned problems, including the need to perform meaningless voice input due to the inability to distinguish between an unregistered state and an unrecognized state. It is something.

B) Means for Solving the Problems The present invention provides a voice recognition device with a feature extraction unit and an input sound plt4.
A turn creation section, a standard voice pattern memory, an identification section, a control section, a counting means connected to the standard voice pattern memory, and a switch means for switching the output of the hexagonal voice pattern creation section between the counting means and the identification section, Switching between voice registration and talking is performed by a switch means based on a signal from the control section, and when registering a voice, a counting means adds up the number of registrations and adds the output of the counting means to the control section. This indicates that the application is unregistered.

(E) The number of registrations is added by the action counting means when registering the voice, and the output of the counting means is applied to the control section, so that the control section determines whether the voice is in an unregistered state or a tongue.

(f) Example FIG. 1 is a block diagram of one embodiment of the present invention.

(1) is the transmitter, (2) is the receiver, (3) is the transmitter (1
), a handset with a built-in handset (2). (4) is a hook switch that detects whether the handset (3) is on-hook or off-hook. (5) is a control unit, and two switches SW1 (61 and SW2 (7)) are provided inside the control unit (5). (8) is an external line connected to a handset (3 ) in the transmitter (1
) and a handset (2). Also, 5WI
Reference numeral 61 also serves as a pulse signal sending contact.

(9) is a feature extraction unit, which is connected to the transmitter in the handset (3) by 5W2 (71).
9) is an input voice pattern creation unit that creates an input voice pattern from the characteristic doctor extracted in step 9); 01) is a standard voice pattern memo IJ for storing the input voice pattern as a standard voice pattern; (121 is a standard voice pattern memory); A counter for counting the number of standard voice patterns stored in the memory is connected to the control section (5) and the standard voice pattern memory αD.

Also, (13) is an identification unit that compares the input voice pattern with the standard voice pattern in the standard voice pattern memo (January 11) and outputs the result, and (14) is an input voice pattern creation unit (10) that outputs the output from the counter ( This is the switch SW3 for switching to 12 or identification section (1
0 is a display section, each of which is connected to a control section (5).

Next, its operation will be explained. When registering standard voice patterns (
When registering voice), press the push button (
(not shown) to enter the registration mode, and then pick up the handset (3) from the phone body, the hook switch (4) detects the off-hook state and sends a signal to the control unit (5).
Output to. The control unit (5) closes the internal SW2 (7) and displays an instruction to prompt the first voice input on the display unit 06). Next, when voice is input to the transmitter (1), the voice signal from the transmitter (1) is input to the feature extraction section (9).

Next, an input speech pattern is created in the input speech pattern creation section α■ using the feature quantities extracted by the feature extraction section (9). Since 5W3Q4) is switched to the counter decoy side by the signal from the control unit (5), the input voice pattern is stored in the standard voice pattern memory αυ via 5WsQ41. At this time, the telephone number is stored in the telephone number storage section (to), 1 is added to the contents of the counter, the change is detected in the control section (5), and the next voice input instruction is displayed on the display section 06). Is displayed. below,
Standard voice patterns are stored using the same procedure.

On the other hand, during a call (voice recognition), the push button (not shown) provided on the control unit (5) is pressed to set the call mode, and the input voice pattern created by the same procedure as when registering the voice is 5W3 (141 is switched to the identification unit 0th floor side by the signal from the control unit (5), so 5WsalBi
Therefore, it is input to the identification section αJ. Then, the identification unit αJ compares the input voice pattern with the standard voice pattern in the standard voice pattern memory aD, and outputs the result to the control unit (5). The control unit (5) reads the telephone number corresponding to the output from the identification unit α3) from the telephone number storage unit,
At the same time, open SW2 (7) and close 5W1 (51) to connect to the external line (8). By opening and closing 5W1 (6) and sending out a pulse signal corresponding to the read telephone number, you can contact the desired party. When the hook switch (4) detects an off-hook state and the counter 02+ is zero, that is, no standard voice pattern is registered in the standard voice pattern memory, when the hook switch (4) detects an off-hook state during voice recognition that can call a telephone, the controller (5) ) displays on the display that the standard voice pattern is not registered without closing 5W2 (7). Figure 2 is an example of displaying that the standard voice pattern is not registered. An example using a 7-segment display, and (b) an example using a dot matrix display.

In this way, it is recognized that it is meaningless even if the recipient's name is uttered into the handset (3), so meaningless voice input is prevented.

Note that in the above explanation, the pre-registered telephone number is output as is according to the recognition result.

For example, if it is recognized as "Sanyo", it will immediately be r06.
-991-XXXXJ and a pulse signal are output, but this is not the only option.When a "Sanyo" is recognized, the display (le) will display "06-q q 1-XXX".
It may be configured such that XJ is displayed, then "dial" is input by voice, and when this is recognized, the telephone number is sent out as a pulse signal. In addition, numbers themselves, such as "zero" or "bee", can be included in the recognition target.

In addition to the pulse signal, it is also possible to provide a tone signal generator and to send out a tone signal.

Furthermore, even when a speaker designation switch is provided and used by multiple speakers, the counter (2) can be configured to count on the speaker side. Shipa ν is also good.

Needless to say, this does not limit the present invention in any way.

(G) Effects of the Invention As is clear from the above explanation, according to the present invention, in a voice recognition telephone, when the handset is picked up and placed in an off-hook state, even though the standard voice pattern is not registered, It will not be confused with the unrecognizable state that can occur due to the influence of voices, noise, etc.

[Brief explanation of drawings]

FIG. 1 is an overview diagram showing an embodiment of the present invention, and FIG. 2 is a diagram showing display contents on a display section. (1)...Telephone, (2)...Telephone receiver, (3)...
・Hook switch, (4)...Hook switch, (5)...
- Control unit, (11)...Standard voice pattern memory, Mouth 2...Counter, (IJ...Identification unit, 04J...
switch. Figure 2 5 Procedures subject to amendment Written amendment (voluntary) Date of June 1986

Claims (1)

[Claims] (1) a handset and a control unit connected to the handset;
A voice recognition telephone having a telephone number storage section connected to the control section, a display section connected to the control section, and a voice recognition device connected to the control section, wherein the voice recognition device is connected to the voice recognition section. a feature extractor that extracts features of the audio signal from the audio signal input to the device; an input audio pattern creator that creates an input audio pattern from the extracted features; and a standard audio generator for pre-registering standard audio patterns. a pattern memory; an identification unit that compares the input audio pattern and the standard audio pattern and outputs the result; a counting unit connected to the control unit and the standard audio pattern memory; and an input audio pattern creation unit. switch means for switching the output of the standard voice pattern to the counting means or the identification section, the switch means switches between the time of registering the standard voice pattern and the time of talking based on a signal from the control section; At the time of pattern registration, the number of registered standard voice patterns is added by the counting means, and the output of the counting means is added to the control section, and when the standard voice pattern is not registered, the display section indicates that the standard voice pattern is not registered. A voice recognition telephone characterized by: