JPH0373775A - Information input device for elevator - Google Patents

Abstract

PURPOSE:To easily correct the result of an operation instruction by constituting the title device so that a user for elevator approaches a voice input microphone and generates the operation instruction and confirms the result of recognition by eyes or ears and generates the operation instruction again if the result is erroneous, while leaves from the voice input microphone when the result is correct. CONSTITUTION:When a user for elevator approaches a voice input microphone 4 in a certain distance and an approach sensor 6 detects the approach state, a voice recognizing means 10 starts voice recognition treatment, and the result the recognition of the voice inputted at that time point is informed to the user for elevator by an information device 18, but the result is not processed as the operation signal for the control of the elevator and kept in reservation state. At the time point when the approach sensor 6 detects the departure of the user for elevator from the voice input microphone 4, an operation output means 19 instructs an elevator controller 20 of the actual operation processing.

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Industrial Field of Application) The present invention relates to an information input device for an elevator that recognizes the voices of elevator users and performs elevator input operations.

(Conventional technology) Lifts and elevators are installed in various buildings, but usually
Actions that are directly controlled by elevator users include: ■ Pressing the call button in the elevator hall.

■Press the destination call button inside the elevator car.

■Press the button to open and close the elevator door.

There are three types of items that can be considered.

These are in the order in which the user presses the corresponding button and the elevator responds as a result.

However, the buttons attached to elevators are not necessarily easy to use, and often leave users dissatisfied. For example, if you are carrying luggage in both hands, you must unload your luggage before pressing the button. Also, when used by a blind person, the user must manually search for the small buttons. Furthermore, if a person is standing in front of the button, it may be difficult to operate it, which may be inconvenient.

As a measure to eliminate these inconveniences, an information input device that recognizes and processes the user's voice has been proposed. In this system, a microphone for voice input is placed in the place where the call button was conventionally installed, and as a result of analyzing the voice, if it is determined to be a command to operate the elevator, the corresponding action is performed. . For example, if the user pronounces "gokai", the call response lamp for the 5th floor is turned on and the elevator is moved to the 5th floor. This is exactly the same as pressing the call button for the 5th floor. In speech recognition processing, several words are registered and stored in advance as a dictionary, and the results of frequency analysis of the input speech are compared with the word data stored in the dictionary, and the results are compared to the closest speech and a certain percentage. If the word is similar to the above, the operation process indicated by the corresponding word is performed.

As a speech recognition method used in such elevators, a method called speaker-independent word recognition is generally used, and this method attempts to recognize the speech of any speaker. Also, the audio content is "Hirake (Rke)"
, "Tojiro (rrJjiru)", "Gokai (5th floor)", etc. are recognized in word units.The number of registered words is about 16, and several types of words are used in the dictionary for speech that is difficult to recognize and distinguish. For example, recognition is improved by having multiple dictionaries such as ``Ikarai'' and ``Ichikaij'' on the first floor.

As described above, voice recognition enables general elevator operation input.

However, when registering voices into a dictionary, the environment is a quiet place, and if an elevator is installed in a similar place to that environment, the recognition rate will be over 90%, but in the case of an actual elevator. The recognition rate is low because the environment in which they are placed has a lot of background noise. In order to prevent this, attempts are being made to construct a practical information input device by performing processing such as not performing a recognition operation unless the voice exceeds a certain level.

FIG. 6 shows an external view of an elevator hall using a conventional example of such an elevator information input device. A hall indicator 2 and a hall call button 3 are attached near the elevator door 1, and a voice recognition microphone 4 is also attached. And this microphone 4
If the elevator user speaks in the direction of the button, the elevator user can perform input operations for the elevator. For example, if "sea urchin" occurs, the upward direction hall call button 3 lights up and the elevator can be operated to arrive at the corresponding hall.

(Problem to be Solved by the Invention) However, with such conventional devices, if an elevator user consciously speaks to the microphone, voice processing is performed with a recognition rate of about 90%. Compared to the method of operating buttons, there is a greater possibility that incorrect processing will be performed, and the user may be left dissatisfied with unexpected results.

For example, even though the user utters "Nikai" to instruct the call registration for the second floor, the call registration for the fifth floor may end up being done. In this case, the user needs to say "Nikai" again, but since the already registered call for the 5th floor will not be deleted, the user will end up creating calls for the 2nd and 5th floors, and the elevator will stop. You end up making unnecessary movements. In addition, when a door opening/closing device makes sounds such as "Tojiro" or "Hirake", the door may not close and the elevator may not be able to run.

In addition, group control elevators that operate multiple elevators collectively have a reservation display that instantly determines which elevator will respond to a call, but if a call is registered based on erroneous recognition, Not only the users who used it,
There was also the problem that it caused confusion for other users at the same hall.

The present invention has been made in view of such conventional problems, and an object of the present invention is to provide an information input device for an elevator that facilitates non-contact correction of operation command results based on voice recognition.

[Structure of the Invention] (Means for Solving the Problems) The elevator information input device of the present invention inputs audio by WgII.
a voice recognition means for recognizing elevator operation voice from the microphone input; a proximity sensor for detecting that a passenger approaches the microphone; a notification means for notifying the state of the voice recognition result; and the proximity sensor. means for giving priority to a later input when there are multiple voice inputs while the proximity sensor is working; and operation output means for outputting the operation recognized by the voice recognition means when the proximity sensor is no longer working. It is something.

(Function) In the elevator information input device of the present invention, the proximity sensor detects when the elevator user approaches the voice input microphone to speak, and when the user approaches within a certain distance, Voice recognition processing by the voice recognition means is started, and the recognition result of the voice input at that time is notified to the elevator user by the notification means, but it is kept in a pending state without being processed as an operation signal for elevator control.

Thereafter, when the proximity sensor detects that the elevator user has moved away from the voice input microphone, the operation output means instructs the elevator control device to perform actual operation processing.

Furthermore, if there are multiple voices input at the time when the proximity sensor detects the presence of an elevator user, the latter voice is given priority as a correction input, and this voice recognition result is processed as an operation signal, Misrecognition errors are corrected using audio.

In this way, an elevator user issues an operation command close to the voice input microphone, confirms the recognition result with his or her eyes or ears, and if the result is incorrect, the command is issued again, and if the result is correct, voice input is performed. The operation process can be executed by moving away from the microphone.

(Example> Embodiments of the present invention will be described in detail below with reference to the drawings.

FIG. 1 shows an elevator hall used in an embodiment of the present invention, and shows a case where a voice Wgx microphone 4 is installed to register a user's destination floor by voice. In this embodiment, since the destination floor is input before getting on the elevator, there is no need to register a call in the elevator car, and the destination call is automatically registered using a hall call.

Reference numeral 5 denotes a hall destination call button with a destination floor display lamp that lights up the registered destination floor lamp when the destination floor is registered. In addition, in the hall, there is a position sensor 6 for detecting elevator users, a sensor lamp 7 indicating that the voice recognition process is possible, an OK lamp 8 that indicates the voice recognition process is normal, and an indicator lamp 9 that indicates that the voice recognition process is not possible as shown in FIG. This configuration is added to the conventional example.

FIG. 2 is a circuit block diagram of an embodiment of Kinoe and Akira.
It includes a CPU 10, an A/D conversion section 11, a bandpass filter section 12, a voice section detection section 13, a sampling section 14, a dictionary storage section 15, and a program storage section 16. Furthermore, the position sensor processing unit 1 for the signal from the position sensor 6
7. A recognition result notification unit 18 that provides a lighting signal to each lamp 7°8.9, an operation output unit 19 that outputs the voice recognition result by the CPLI 10 as an operation signal, and an elevator control unit 20 that controls the operation of the elevator. We are prepared.

A diffuse reflection type dark infrared sensor is used as the position sensor 6, thereby making it possible to detect distance without making the user aware of the presence of the sensor. Further, the output of the position sensor 6 is normally set at a current output of 4 to 20 mA, and the distance signal from the sensor 6 is processed by the position sensor processing unit 17 at 8 mA.
The data is converted into a digital amount on the order of bits, and converted into a form that can be determined by the CPU 10.

As shown in FIG. 1, the sensor lamp 7, OK lamp 8, and reject lamp 9 are arranged in an integrated manner so that they come into view when the elevator user is speaking into the voice recognition microphone 4. ing.

The sensor lamp 7 is a lamp that indicates the start of voice II, and when the CPU 10 determines that the value measured by the position sensor 4 is within a certain range, the recognition result notification unit 18 consisting of a digital signal is activated. Lights up depending on the output. As a result of a series of voice recognition processes, the OK lamp 8 and the reject lamp 9 are recognized by the CPU 10 as an OK clamp when the similarity of the input voice exceeds a certain standard, and as a reject lamp 9 when it is below a certain standard. The light is turned on for 1 to 2 seconds via the result notification unit 18 to notify elevator users of the result. At this time, if the similarity exceeds a certain standard, the CPU 10 flashes the result not only on the OK clamp but also on the registration lamp of the destination button 5,
An operation is performed in which the elevator user is asked to determine whether the recognition result is normal.

Conventionally, this destination button 5 is controlled by the elevator control section 20, but by lighting the registration lamp processing by logical OR with the output result from the voice recognition device, the destination button 5 is controlled by the elevator control section 20. The embodiments of the present invention can be configured without any changes.

The blinking display of the registration lamp of the destination call button 5 is the same as the data format conventionally output from one operation output unit to the elevator control unit 20, and is turned on and off at a constant cycle by CPIJlo. , -0 in the shell
．． A 5 second period is used.

Further, this output and the conventional call registration lamp are configured by simply performing a wired OR, and therefore, if the button is pressed while the lamp is blinking, priority will be given to the constantly lit state.

Since the audio input from the microphone 4 is an analog value, the A/D converter 11 converts it into a digital signal according to the amplitude of the audio for processing by the CPU 10. Then, the bandpass filter section 12 performs audio band limitation on the digital signal from the A/D conversion section 11 to obtain an array of digital values at a sampling frequency of 12 KH 2112 pits.

However, this value is a continuous number for the audio, so
In order to extract only the voice features, the amount of information is compressed by converting it into a spectral sequence of 8 m5ec units.

The speech section detection unit 13 distinguishes between unvoiced parts and effective speech in order to extract effective speech, and as a result, compiles data of speech that should actually be recognized.

The sampling unit 14 normalizes the length of the input speech, which differs for each user, according to the length of the speech section of the dictionary pattern used for recognition. In other words, the audio data is converted into 256-point data, and compared with the 256-point dictionary pattern stored in the dictionary storage unit 15, the CPU 10 calculates the degree of similarity.
The dictionary pattern with the highest degree of similarity is output to the operation output unit 19 as a recognition result. This operation output section 19 is often a general digital output circuit.

The program storage unit 16 stores a program for the processing procedure performed by the CPLllo described above.

Data from the operation output unit 19 is input to the elevator control unit 20 to perform various elevator operation controls.

Next, the operation of the elevator information input device having the above configuration will be explained.

If the operation process is not performed using voice recognition, the elevator user selects the floor he or she wants to go to by pressing the hall destination call button 5.
Select and enter, and the lamp of the button for the floor that was pressed will light up. After a while, the elevator arrived,
The hall destination call is transferred to the call in the elevator car,
Users can be moved to the desired floor.

Next, when performing voice recognition processing, when the elevator user approaches the voice recognition microphone 4 and the position sensor 6 determines that the distance is such that voice recognition processing can be performed, the sensor lamp 7 lights up. Notify the user to speak.

This distance is usually set to about 30 cm.

At this time, the elevator user vocalizes the floor he or she wants to go to, and if the voice recognition process is performed and the voice is determined to be normal, the
The K clamp lights up, and the detector lamp 9 lights up when the sound is determined to be abnormal.

However, in this case, when the voice recognition device compares the input voice and the contents of the dictionary storage unit 15 and the similarity exceeds a certain standard, the OK clamp is lit.
Therefore, even if "Nikai" occurs, the voice recognition by the CPU 10 may determine that it is the third floor, but the OK clamp lights up even in such a case. Therefore, when the OK clamp lights up, the user is notified by blinking the destination floor registration lamp of the destination call button 5 as a recognition result display.

When the user visually sees this display and determines that the user's intended destination floor has been registered, the user moves away from the voice recognition microphone 4, but the position sensor 6 detects each of them and the operation output unit 19 The recognition result is sent to the elevator control unit 20, and the lamp, which had been flashing until now, is switched to constant lighting, indicating that the call has been registered.

FIG. 3 illustrates the above-mentioned voice recognition processing process as a voice waveform.

Figure (a) shows the audio waveform input from the microphone 4,
Since this waveform contains various noises in addition to the words to be recognized, the necessary word sections are extracted by the speech section detecting section 13 as shown in FIG. 3(b). This process has to be performed by detecting that the most characteristic amplitude of audio has been input, so if there is loud noise, there is a possibility of false detection.
Therefore, as in this embodiment, if a new input is received from the microphone 4 while the position sensor 6 is on, the new input is corrected to be used for voice recognition.

FIG. 2C shows the output M result of the bandpass filter unit 12, which is frequency band data of 16 channels.

The above operation will be explained in detail with reference to the processing flowcharts of FIGS. 4 and 5. In step S1, it is determined whether the distance between the elevator user and the voice recognition microphone 4 measured by the position sensor 6 is within 30 cm. , the voice recognition processing from step S2 onwards is performed, and if it is not, the sensor processing from step S9 onwards is performed.

In step S2, the sensor lamp 7 is turned on to encourage the elevator user to make a voice.

In step S3, the voice section detection unit 13 determines whether proper voice is input, and if it is proper, step S3
Perform the audio n1ffi processing in step 4, and if there is no audio input,
If it is determined, this routine ends and the next cycle is executed.

In step S4, compared to the conventional speech recognition device, the fifth
As shown in the flowchart in the figure, correction by re-input is possible. In the flowchart in Figure 5, step 8
41 detects the voice input after the Wg recognition process is completed,
If there is an input, the previous recognition result is invalidated and the result is prioritized over Ii2 based on the new voice (steps 842 and 843). Therefore, when the sensor lamp 7 is on, the last voice command is valid, and the misrecognition can be corrected by simply uttering the destination floor into the microphone 4 again later.

In FIG. 4, as a result of the voice recognition process in step S4,
In step S5, the input voice and the data in the dictionary storage unit 15 are compared, and if a degree of similarity exceeding a certain standard is recognized, an OK clamp is lit in step S6, and the voice is valid for elevator users. to inform that it was. If the voice is below the standard, the reject lamp 9 is turned on in step S7 to indicate that the voice is ambiguous, and re-input is requested.

In step S8, the registration lamp of the destination call button 5 is blinked together with the OK clamp to indicate the voice recognition result to the elevator user.

If a misrecognition is confirmed here, the result can be changed by speaking again. Normally, an elevator user approaches the voice recognition microphone 4 and speaks, and then looks at the destination call button 5 without leaving the position until the user confirms the result, so the sensor 6 remains on and the input is repeated. is possible.

If the sensor 6 is turned off, the sensor lamp 7 is turned off in step S9, the OK clamp and reject lamps 9 are turned off in step 810, and the registration lamp of the destination call button 5 is blinking in step 811, that is, voice recognition is activated. If the processing result is being displayed, it is actually registered as a call input in step 812.

As described above, in this embodiment, the elevator user stands in front of the microphone 4 for voice recognition, speaks the floor he or she wants to go to, confirms the result by checking the illuminated state of the destination call button 5, and selects the desired result. In this case, if the user moves a little away from the voice recognition microphone 4, the call will be registered as a call, and the voice input will be performed in an easy-to-understand manner to match the normal actions of the elevator user. be able to. Also, if the result is incorrect, by saying it again at that position, you can erase the previously input content, that is, the currently blinking data, and give priority to the current result, which will give the correct result. It is possible to correct up to

Moreover, since elevator users can register the hall destination floor using only their voice and their own movements, this process can be performed as an original voice recognition process that does not require manual operation of buttons or the like.

In the above embodiment, the elevator user was detected using a diffuse reflection type position sensor, but this position sensor is used to detect that the elevator user is standing between the voice recognition microphones. Therefore, it is also possible to use floor-standing mat sensors, photoelectric sensors, ultrasonic sensors, etc. In addition, we decided to notify the voice recognition result by flashing the destination floor display lamp on the destination call button, but if the display displays something like ``Registering the second floor'' or if the speaker outputs a voice message. It is also possible to broadcast similar content, and the notification means is not particularly limited.

[Effects of the Invention] As described above, according to the present invention, in order to correct malfunctions of operation commands by voice recognition in a non-contact state, changes in the position of elevator users are detected to display and execute operation results. Because it is a processing method, elevator users can easily correct misrecognition results by simply inputting their voice again, making it easy to obtain the desired result regardless of voice failure or surrounding sounds. This solves the conventional problem of difficulty in registering destinations by voice.

Further, for the elevator itself, it is possible to eliminate unnecessary movements due to a destination floor that should not actually be registered due to an erroneous ml, and the service can be improved. In addition, even if the user inputs voice input multiple times, the actual operation of the elevator will not be performed unless the user moves away from the sensor, so the elevator service will not be degraded.

[Brief explanation of drawings]

Fig. 1 is an external view of an elevator hall according to an embodiment of the present invention, Fig. 2 is a circuit block diagram of the above embodiment, Fig. 3 is an audio waveform diagram explaining the speech recognition processing process of the above embodiment, and Fig. 4
The figure is a flowchart explaining the operation of the above embodiment.
The figure is a flowchart explaining the voice recognition correction operation of the above embodiment, and FIG. 6 is an external view of a conventional elevator ball. 1... Door 2... Indicator 4... Microphone for voice recognition 5... Destination call button 6
...Position sensor 7...Sensor lamp 8...OK
Lamp 9...Reject lamp 10...CPU
15... Dictionary storage section 16... Program storage section 17... Position sensor processing section 18... Recognition result notification section 19... Operation output section 20
...Elevator control section 1ilA Patent attorney Hidekazu Miyoshi Figure 2 Figure 4 Figure 55! J Figure 6

Claims (1)

[Claims] a microphone for recognizing voice, a voice recognition means for recognizing elevator operation voice from microphone input, a proximity sensor for detecting when a passenger approaches the microphone, and a notification means for notifying the state of the voice recognition result; means for giving priority to a later input when there are multiple voice inputs when the proximity sensor is working; and operation output means for outputting the operation recognized by the voice recognition means when the proximity sensor is no longer working; An elevator information input device comprising: