JPH0829915B2 - Crane voice controller - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial application] The present invention relates to a crane voice control device for remotely operating a crane by a command word issued by an operator.

[Prior art]

In recent years, in crane equipment, various crane voice control devices using a voice recognition device have been developed in order to rationalize operation work and save labor. This is for example
As disclosed in Japanese Patent Publication No. 61-267689, an operator carries a mobile radio device with a microphone, and a command word uttered by the operator is transmitted to a voice recognition device via a voice radio device. . In the voice recognition device, standard patterns of voices of a plurality of command words are registered in advance by the operator, and these and the characteristic patterns of the voices of the command words issued by the operator during driving are collated and the corresponding standard patterns are determined. Be recognized. When the command word is recognized by the standard pattern, the signal according to the command content is output to the crane controller,
The crane controller controls the drive unit of the crane, and the operation is performed according to the command content.

[Problems to be Solved by the Invention]

By the way, in the above-mentioned device, the most important thing is the recognition ability of the command word by the voice recognition device. In many cases, since the usage environment is under high noise, it is necessary to surely recognize the command word issued by the commander without error. In the above device, a function of answering the command word once recognized to the operator is given, and the operator issues a command word to actually start driving after confirming this, but here If you set the recognition condition of the voice recognition device to be very strict in order to prevent false recognition due to noise etc.,
Due to a slight difference in the utterances of the operator, it is rejected if the corresponding command word is not registered, and the operation cannot be started. On the contrary, if the recognition condition is set relatively roughly, the voice recognition device is likely to erroneously recognize the command word, and the operator needs to repeat the command until the matching command word is answered back. There is also a problem that an unintended driving operation is performed when the operation is started by missing or answering the answerback by the user.

For this reason, in the conventional device, a large number of rejects and malfunctions are caused, safety and operability are low, and practical use is hindered.

Moreover, as a measure to prevent malfunction, it is considered effective to utilize the regularity due to the driving operation of the crane.
In this regard, although a little is described in JP-A-61-267689, the specific description is poor and it is not clarified.

The present invention has been made in view of such circumstances, and sets a criterion for the recognition condition of the voice recognition device according to the degree of difference from the standard pattern, and a command word that is erroneously recognized based on the command sequence according to the crane operation rule. An object of the present invention is to provide a crane voice control device which is excellent in practicality by reducing the rejection rate and the false recognition rate by detecting it.

[Means for solving the problem]

Crane voice control device according to the present invention, the standard pattern of the voice of a plurality of command words that are registered in advance by a specific speaker, and the feature pattern of the input voice is collated to recognize the matching standard pattern, In a crane voice control device provided with a voice recognition device for outputting a signal corresponding to this standard pattern to a crane control unit, registration means for multiple registration of standard patterns for each command word, characteristic pattern of input voice, and said registration Calculated by the difference degree calculating means from the difference degree calculating means for individually calculating the difference degree of the standard pattern of each command word registered by the means and the standard pattern registered in the registering means. The difference is 1
Selection means for selecting the lowest first standard pattern and second lowest standard pattern, and comparison means for comparing the command words corresponding to the first and second standard patterns selected by the selection means. And when the command words of the first and second standard patterns are equal, and when the degree of difference of the first standard pattern is smaller than the first reference value, the first recognition that recognizes the first standard pattern as a match When the means and the command words of the first and second standard patterns are different, the degree of difference of the first standard pattern is smaller than the second reference value lower than the first reference value,
And a second recognition means for recognizing the first standard pattern as a match when the difference between the degrees of difference between the first and second standard patterns is larger than a third reference value lower than the second reference value. It is a feature of the first invention to have, and
In addition to the features of the first aspect of the invention, storage means for storing a standard pattern that has been previously recognized and output a corresponding signal,
An estimation unit that estimates a standard pattern to be recognized next from the standard pattern stored in the storage unit based on a command order according to a predetermined driving rule, and a standard pattern other than the standard pattern estimated by the estimation unit. When it is recognized next, it comprises a means for rejecting the output of a signal to the corresponding crane control unit, which is a feature of the second invention.

[Action]

In the crane voice control device according to the first aspect of the present invention, the standard pattern is multiply registered in advance for each command word by the specific speaker. When the operator utters the command word, the dissimilarity between this characteristic pattern and the standard pattern that is multiply registered is calculated for each, and the lowest dissimilarity among the dissimilarities and the second calculated standard pattern are calculated. A standard pattern for which a low dissimilarity is calculated, in other words, a standard pattern having the highest similarity to the voice feature pattern and a standard pattern having the second highest similarity are selected, and the command words corresponding to them are compared. Here, when the command words are the same, the difference degree of the first standard pattern is the first
When it is smaller than the reference value, the first standard pattern is recognized as matching with the characteristic pattern of the voice of the command word uttered by the operator. That is, when the difference between the first and second standard patterns is larger than the first reference value, it is rejected as there is no matching standard pattern. On the other hand, the first
And the command words corresponding to the second standard pattern are different, the degree of difference of the first standard pattern is smaller than the second reference value that is lower than the first reference value, and the degree of difference of the first and second standard patterns is When the difference is larger than the third reference value which is lower than the second reference value, the first standard pattern is recognized as matching the characteristic pattern of the voice of the command word uttered by the operator.

A crane voice control device according to a second aspect of the present invention is the first voice control device.
In addition to the operation of the invention, the standard pattern previously recognized and the corresponding signal is output is stored, and the standard pattern to be recognized next is estimated from this standard pattern according to a predetermined driving rule based on the command sequence. . And if the next recognized standard pattern is other than the estimated standard pattern,
No signal corresponding to the recognized display pattern is output to the crane controller.

〔Example〕

Hereinafter, the present invention will be specifically described with reference to the drawings illustrating the embodiments. FIG. 1 is a block diagram showing the configuration of a crane voice control device according to the present invention. This embodiment shows an example of application to an overhead traveling crane, which is roughly classified and equipped with a headset 1, and is a portable mobile station carried by an operator.
2. It is composed of the mobile station 2 and a ground installation device A for transmitting and receiving signals by radio, and a crane onboard device B for transmitting and receiving signals by radio.

The headset 1 includes a voice input microphone 1a and voice response headphones 1b, and is worn on the head of the operator. The command word of the operator input to the voice input microphone 1a is transmitted as a radio wave by a voice signal from the portable mobile station 2 and is received by the voice input system antenna 3 of the ground-based apparatus A. The voice input system antenna 3 constitutes a voice system wireless device together with the antenna coupler 4 and the voice input fixed station 5, and the voice signal received by the voice input system antenna 3 is included in the antenna coupler 4 and the voice input fixed station 5. Are transmitted to the voice recognition device 6 via the respective.

For example, DP-3000 manufactured by NEC Corporation can be used as the voice recognition device 6, and multiple types (three or more types) of standard patterns of voices for each command word can be registered in multiple layers. Multiple register each command word in advance. This is to cope with the fact that the operator may not always speak in the same voice pattern as the standard pattern in which the command words are registered during crane operation. By registering the standard patterns of the types, it becomes easy to recognize the command word uttered by a specific operator.

The voice recognition device 6 is connected to the control device 7, and the control device 7 recognizes a recognition condition of a command word described later by the voice recognition device 6 and a command sequence based on the crane operation rule, and finally uses these to recognize the command sequence. Whether to perform the operation corresponding to the recognized command word while recognizing the command word,
It has a function to judge whether or not.

The control device 7 is connected to a wireless ground station 8, and the wireless ground station 8 is connected to a hollow wire coupler 9 having an operation output system antenna 10. An operation output signal corresponding to the command word that is finally recognized by the control device 7 and is permitted to operate is transmitted from the operation output system antenna 10 and is connected to the radio station 12 of the crane onboard device B. Received at 11. The radio station 12 is connected to the sequencer 13,
An operation output signal 13 is inputted as a digital signal, and a logic processing of an operation condition and an operation time is carried out, and a final operation output is outputted to a crane control panel 14.

The crane control panel 14 inputs the final operation output and outputs a drive signal to a corresponding drive unit (not shown).

Further, while the control device 7 outputs the operation output signal, the result of acceptance or rejection of the command word, that is, the command word when the driving operation is permitted, or the error word when rejected is voiced by a voice synthesizer (not shown). The signal is output and transmitted from the antenna coupler 4 and the voice input system antenna 3 to be received by the portable mobile station 2. The voice response headphone 1b of the headset 1 causes the operator to answer back.

Now, in the device of the present invention configured as described above,
First, the recognition conditions by the voice recognition device 6 and the control device 7 will be described. As described above, a plurality of standard patterns are registered in the voice recognition device 6 for each command word, and these standard patterns are compared with the characteristic pattern of the voice of the command word input during operation. By comparing. Here, when the standard pattern and the feature pattern completely match, the degree of difference (hereinafter referred to as score value) becomes zero, and the greater the degree of difference, the higher the score value. Of course, this matching can be performed based on the degree of similarity. The score values of all the standard patterns are calculated with respect to the characteristic pattern of the voice thus input. Then, the first standard pattern having the lowest score value and the second standard pattern having the second lowest score value are selected. Now, normally, when the command words (hereinafter, referred to as first candidate term and second candidate term) corresponding to these standard patterns are equal, the standard pattern of the first candidate term is recognized, and when they are different, it is not recognized. It may be rejected.

However, this increases the false positive rate and reject rate, making it difficult to secure a normal recognition rate (95%).
The application of the voice control device becomes impossible.

Therefore, in the device of the present invention, the recognition conditions are set as shown below.

(1) When the first candidate term = the second candidate term, the first candidate term is recognized if the score value of the first candidate term <S ₁ .

Rejected if the score value of the first candidate term is ≧ S ₁ .

(2) If the first candidate term is not equal to the second candidate term, the score value of the first candidate term is <S ₂ and the score value of the first candidate term-the score value of the second candidate term> S ₃ is the first. Recognize candidate terms.

Reject if the score value of the first candidate term ≧ S ₂ or the score value of the first candidate term−the score value of the second candidate term ≦ S ₃ .

However, the score values were S ₁ > S ₂ > S _3, and in this example, good determination results were obtained when S ₁ = 100, S ₂ = 80, S ₃ = 5.

In other words, even when the first candidate term and the second candidate term are equal, if the score value of the first candidate term is relatively high as S ₁ or more, there is a high possibility of misidentification, and therefore reject determination is performed. Even when the first candidate term and the second candidate term are different, the score value of the first candidate term is lower than S ₂ and the difference between the score values of the first and second candidate terms is higher than S _3. In this case, the first candidate term is highly likely to be recognized, and thus the first candidate term is recognized without rejecting.

When the command word is recognized as described above, the control device 7 determines whether the next recognized command word conforms to a predetermined command sequence based on the crane operation rule.

The control device 7 is adapted to store the standard pattern of the command word that has already been recognized and generated the corresponding operation output signal, and when the next standard pattern is recognized, the command word of the stored standard pattern is stored. Estimate the next command group from
Whether or not the command word of the next recognized standard pattern is included in this command word group is determined, that is, the rationality of the driving rule is checked. If the command word of the next recognized standard pattern is included here, the corresponding operation output signal is generated with the command word as the final recognition command word, and if not included, the recognized standard pattern is detected. It is determined that the command word is erroneous recognition and the generation of the operation output signal is prohibited, thereby suppressing the erroneous recognition of the command word and preventing the abnormal operation. The operation will be specifically described below with reference to the timing charts of the operation operations shown in FIGS.

Fig. 2 shows an example of a single operation. First, "Higasi" is uttered from the direction instruction word group of the command word group shown in Fig. 6, and when this is recognized, the crane control panel is recognized.
14 is set to the command waiting state for the eastbound operation. Then, the term "start" that commands execution start is uttered, and when this is recognized, the eastbound operation is started. After that, if the command words that are recognized follow the command sequence, the operation is not stopped, but command words that do not comply with the command sequence are erroneously recognized or unrecognizable. When the term is entered and rejected, the eastbound operation is stopped. Here, the rationality check of the driving rules is performed for "higashi" and "start". In the next command word, the fact that the order of commands according to the driving rules is violated is to command the opposite direction indicating term, that is, westward driving. It is "nishinishi", and the fact that it is recognized is considered to be a misunderstanding, so it is rejected. On the other hand, if a direction term other than “nishinish” and a stop term “tomare” are recognized, it is permitted. In other words, "Higashi", "Nishinishi"
Directional terms other than those are permitted because they may be operated in combination, and "legacy" in the same direction as the driving direction is permitted as a stop word in this case, as is the case with "Tomare."

Figures 3 and 4 show an example of compound operation. In Figure 3, after the eastbound operation was similarly started with "Higashi" and "Start", the next command word when "Makioroshi" was recognized The eastbound operation has been stopped due to misidentification or rejection. The command words that are judged as misidentification here are all directional terms other than "Start", "Tomare", "Choi", "Choichoi", "Fredome", and "Gakusun". When "Minamigawa" is recognized, the direction terminology continues with "Makioroshi", so it is judged as a misunderstanding and is not permitted.

In Fig. 4, after the eastbound operation is started, further downwind operation is started by "Makiorosi" and "start", and the eastbound and downwind operation is stopped due to a misunderstanding or a rejection decision by the next command word. Has been done. The command word that is judged as misidentification here is a term other than "Higasi", "Makioroshi", and "Tomare". For example, even if "Kitakita" that commands northbound driving is recognized, it is already in the eastern direction Since a combined operation with is being performed, it is judged as a misidentification, and inching motions such as "choi" will not be executed unless any operation under the eastbound winding is stopped by "tomare" etc. Therefore, it is also judged as a false positive.

FIG. 5 is an example of a command sequence system diagram in which the contents of the above-described combined operation are summarized. First, each power unit of the crane is set to a drivable state by "power on". Then, one of the turn-indicating terms is uttered and recognized, and driving is started in the direction recognized by "start". It is not the same directional term that is recognized and recognized in the following, and directional terms other than the directional terms of opposite directions. When this is recognized and then "start" is recognized, the combined operation is started. Then, only the turn-around term and the "tomare" uttered by are permitted to be recognized, and when the "tomare" is recognized, both of the combined operations of are stopped.
If either one of the turn signals is recognized and only one of the driving is stopped, the next one is recognized as the other turn signal which is not recognized in, and
"Tomare", "Brake", "Choi", "Choichoi",
They are "Fredome" and "Gaxon".

If "Tomare" is recognized here, the other operation is stopped. In addition, braking is applied to the driving operation stopped by the recognition that "brake" is recognized, and "choice",
When "choi choi", "fredome", and "gaksund" are recognized, inching motions corresponding to these are performed within 10 seconds.

When the other turning direction word currently driving is recognized in, the next recognition is permitted for the inching motion term and "brake", and when "brake" is recognized, it is stopped by the same recognition. Braking is applied to the driving operation performed, and in the inching operation term, the inching operation is performed within 10 seconds, and when it is 10 seconds or more, the initial power-on state is restored.

Table 1 below shows the test results of the voice recognition test by the rationality check of the above operating rules in the small diameter pipe factory and the cold pipe manufacturing factory, and the noise level is lower in the cold pipe manufacturing factory. large.

As can be seen, the number of false positives detected by the rationality check at each factory was 0.3%, which can be said to be effective in reducing false positives.

Table 2 shows the results of comparison of the voice recognition ability between the conventional device and the device of the present invention. The conventional device shows the test results in the cold pipe manufacturing plant, and the device of the present invention shows the cold pipe manufacturing plant and the small diameter pipe factory. The test results are shown respectively.

As is clear from the table, in the device of the present invention, since the rejection rate and the false recognition rate are significantly reduced, the normal recognition rate is increased to 98%, which is higher than 95%, which is considered to be practically acceptable. .

The reason that the normal recognition rate is lower in the cold pipe manufacturing factory than in the small diameter pipe factory by about 1% is mainly due to the difference in the frequency of the impact noises generated in the voice. For this reason, in order to prevent the rejection due to the noise input, it is necessary to set the voice input power parameter in the voice input device 6 according to the noise condition of the applied place. For example, in order to prevent the impulsive sound from being erroneously detected as the start point of the voice input, the fact that the impulsive sound has a large peak and the input time is relatively short can be used, and it can be eliminated by setting the integral value higher.

In addition, in the command word in this embodiment, “nishinishi”,
Because we use easy-to-identify terms such as "Minamigawa",
It is effective for improving the normal recognition rate. Furthermore, terms such as "higashi" and "start" that begin with unvoiced fricatives are difficult to input at the beginning and are easy to reject. For example, "higashi" is changed to "east".
It is also possible to deal with it, for example.

In this embodiment, the main devices such as the voice recognition device 6 and the control device 7 are installed on the ground side, but instead of this, they may be installed on the crane machine. However, it is desirable to install the crane on the ground side because it is in a bad environment due to vibration and temperature rise during the operation of the crane, which can avoid failure and malfunction, and is easy to maintain. Also, when the main equipment is installed on the ground side, the crane operation output signal is transmitted wirelessly to the crane machine.However, if an abnormality in the transmission data or an abnormality in the wireless equipment occurs, it will be automatically It is advisable to take safety measures for emergency stop.

〔effect〕

As described above, in the crane voice control device according to the first aspect of the present invention, first, based on the magnitude of the difference between the plurality of standard patterns in which the recognition conditions for voice recognition are multiply registered and the characteristic pattern of the input voice, When the command words of the first and second standard patterns are the same, and when the dissimilarity of the first standard pattern is smaller than the first reference value, the first recognition means for recognizing the first standard pattern as a match, When the commands of the first and second standard patterns are different, the difference between the first standard patterns is smaller than the second reference value which is lower than the first reference value, and the difference between the first and second standard patterns is different. Is larger than the third reference value, which is lower than the second reference value, the second recognizing means for recognizing the first standard pattern as a match is provided so that the command word which has been erroneously recognized can be recognized. And prevent malfunctions It is possible to improve the normal recognition rate.

In addition, in the second invention, in addition to the above effects, the recognized command word conforms to the command sequence based on the crane operation rule, and finally the corresponding operation is permitted, or Because it is rejected, it is possible to reliably prevent malfunction due to misidentification.

As described above, in the device of the present invention, unnecessary voice can be accurately rejected, and misidentification can also be detected, so that the rejection rate and the misidentification rate can be significantly reduced, and safety and operability are significantly improved. The present invention has excellent effects such as that it is sufficiently practical.

[Brief description of drawings]

FIG. 1 is a block diagram showing the configuration of a crane voice control device according to the present invention, FIGS. 2 to 4 are timing charts of operation operations, FIG. 5 is a command sequence system diagram in combined operation, and FIG. It is explanatory drawing of a driving instruction term. 1 ... Headset, 1a ... Voice input microphone,
1b …… Voice response headphones, 2 …… Portable mobile station,
6 ... Voice recognition device, 7 ... Control device, A ... Ground installation device, B ... Crane on-board installation device

Front page continued (72) Inventor Ken Mori 4-533 Kitahama, Chuo-ku, Osaka City, Osaka Prefecture Sumitomo Metal Industries, Ltd. (72) Inventor Takeki Nakata 4-53-3 Kitahama, Chuo-ku, Osaka City, Osaka Prefecture Sumitomo Metal Industries, Ltd. (72) Inventor Toshio Nakamura 4-7 28 Kitahama, Chuo-ku, Osaka City, Osaka Prefecture Sumitomo Metal Engineering Co., Ltd. (72) Toshio Kinoshita 4-7 Kitahama, Chuo-ku, Osaka City, Osaka Prefecture No. 28 Sumikin Control Engineering Co., Ltd. (56) Reference JP-A-63-295395 (JP, A) JP-A-61-114299 (JP, A) JP-A-63-295393 (JP, A)

Claims (2)

[Claims] 1. A standard pattern of voices of a plurality of command words registered in advance by a specific speaker and a characteristic pattern of an input voice are collated to recognize a standard pattern that matches, and the standard pattern is dealt with. In a crane voice control device equipped with a voice recognition device that outputs a signal to the crane control unit,
Registration means for multiple registration of standard patterns for each command word, feature pattern of input voice, and difference degree calculation for calculating the degree of difference between the standard pattern of each command word registered by the registration means separately Means and a standard pattern registered in the registration means, a first standard pattern having the lowest dissimilarity calculated by the dissimilarity calculating means and a second standard pattern having the second lowest dissimilarity are selected. Selecting means for comparing, a comparing means for comparing command words corresponding to the first and second standard patterns selected by the selecting means, and first and second
If the command words of the standard patterns are equal to each other, and the degree of difference from the first standard pattern is smaller than the first reference value, the first
If the command word of the first and second standard patterns is different from the first recognizing means for recognizing the standard pattern of No. 1, the difference between the first standard pattern and the first standard pattern is lower than the first standard value. When the difference between the first and second standard patterns is smaller than the second standard value and is larger than the third standard value which is lower than the second standard value, the first standard pattern is recognized as a match. A second voice recognition means for controlling the crane voice control device. 2. A standard pattern of voices of a plurality of command words registered in advance by a specific speaker and a characteristic pattern of an input voice are collated to recognize a standard pattern that matches, and correspond to the standard pattern. In a crane voice control device equipped with a voice recognition device for outputting to a crane control unit, registration means for multiple registration of standard patterns for each command word, characteristic pattern of input voice, and registration by the registration means The difference degree calculated by the difference degree calculation means is 1 out of the difference degree calculation means for individually calculating the degree of difference between a certain command word and the standard pattern and the standard pattern registered in the registration means. Selection means for selecting the lowest first standard pattern and the second lowest second standard pattern, and command words corresponding to the first and second standard patterns selected by the selection means Comparing means for comparing the first, second
If the command words of the standard patterns are equal to each other, and the degree of difference from the first standard pattern is smaller than the first reference value, the first
If the command word of the first and second standard patterns is different from the first recognizing means for recognizing the standard pattern of No. 1, the difference between the first standard pattern and the first standard pattern is lower than the first standard value. When the difference between the first and second standard patterns is smaller than the second standard value and is larger than the third standard value which is lower than the second standard value, the first standard pattern is recognized as a match. The second recognizing means, the storage means for storing the standard pattern previously recognized and the corresponding signal is output, and the standard pattern to be recognized next from the standard pattern stored in the storage means, in a predetermined operation. If a standard pattern other than the standard pattern estimated by the estimating means and the estimating means for estimating based on the order of commands according to the rules is recognized next, a means for rejecting the output of a signal to the corresponding crane control unit is rejected. Crane voice control apparatus characterized by comprising and.