JP2001265364A - Sound processor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a sound processor which gives unexpectedness and makes a user more feel keeping company with a living body. SOLUTION: The sound inputted from a sound input means 4 is automatically taken in and stored in a storage means 3. Relation is set to the sound data stored in the storage means 3. A control means 1 selects at least one sound out of sounds stored in the storage means 3 and makes a sound output means 7 to output it, so that a continuous pronunciation sound of the automatically taken-in sound and related sound is enjoyed.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice processing device for storing voice and outputting the stored voice.

[0002]

2. Description of the Related Art The present inventors have already applied for a speech processing apparatus (speech recognition and speech generation apparatus) which can be applied to, for example, a character on an electronic toy or a personal computer, and registers input speech while returning it as speech. Japanese Patent Application No. 09-262557). The voice processing device can arbitrarily store the input voice, and can arbitrarily utter the stored voice. Therefore, it is possible to provide the user with unexpectedness and fun.

[0003]

However, in this conventional voice processing apparatus, the types of voices uttered are still poor.
For example, in a case where the surroundings are in a quiet state, there are cases where it is rather uninteresting.

The present invention has been made in view of such circumstances, and an object of the present invention is to provide a voice processing apparatus which is surprising and makes it easier to obtain a sense of being opposed to a living thing. .

[0005]

In order to solve the above-mentioned problems of the prior art and to achieve the above object, an audio processing apparatus according to the present invention comprises: an audio output unit; a storage unit for storing audio; A control means for selecting at least one of the sounds stored in the storage means based on a randomly generated signal in the first operation state, and outputting the selected sound to the sound output means. .

In the present invention, a random number generating means is included, and in the first operating state, at least one of the voices stored in the storage means is selected based on a signal generated by the random number generating means. Then, the sound is output to the audio output means. In the present invention, the random number generation means and the first
In the operation state, at least one of a predetermined number of voices among the voices stored in the storage means is randomly selected based on the random number generated by the random number generation means, and is output to the voice output means in a predetermined order. And one voice selecting means is randomly selected based on the random number in the first operation state.
In the present invention, an association between two voices stored in the storage unit is set, and the association is stored in the storage unit.
The voice selecting means randomly selects a predetermined number of the associations stored in the storage means based on the random number, and outputs the voices set in the selected associations in a predetermined order. And means for selecting sound to be output by the means.

In the present invention, the voice selecting means determines one of the voices stored as names in the storage means and a predetermined number of voices among the other voices stored in the storage means based on the random number. And a voice selecting means for outputting the voice stored in the name as the first voice to the voice output means in a predetermined order.

In the present invention, in the second operation state, the control means stores the sound input from the sound input means in the storage means for a predetermined time, and the sound selection means
Voice selecting means for randomly selecting a predetermined number of sounds among the sounds stored in the second operation state based on the random numbers and outputting the selected sounds to the voice output means in a predetermined order.

In the present invention, in the second operating state, the control means stores the sound input from the sound input means in the storage means for a predetermined time at a predetermined time interval. Based on a random number of runs.

In the present invention, the control means randomly sets the number of times the sound selected by the sound selection means is output to the sound output means based on the random number.

In the present invention, in the third operation state, the control means recognizes a voice based on the sound input from the sound input means, and compares the voice with the voice stored in the storage means. Based on the result of the collation, the recognized voice is stored in the storage unit, and the voice stored in the storage unit is output to the voice output unit.
In the present invention, the control means randomly selects a plurality of operation states based on the random numbers.

According to the above configuration of the present invention, the voice processing apparatus of the present invention has a first operating state, a second operating state, and a third operating state, and includes a plurality of operating states including the operating state. Is randomly selected based on the random numbers. In the first operation state, one of the voice selecting means is randomly selected based on the random number. Then, a predetermined number of voices among the voices stored in the storage means are randomly selected by the selected voice selection means based on a random number generated by the random number generation means, and the voice output is performed in a predetermined order. Output to the means. Preferably, the voice selecting means preferably causes a predetermined number of associations stored in the storage means to be randomly selected based on the random number, and outputs the voices set in the selected association in a predetermined order. Is output to the audio output means.
Further, the voice selecting means randomly selects one of the voices stored as names in the storage means and a predetermined number of voices among other voices stored in the storage means based on the random number. Then, the sound stored as the name is output to the sound output means in a predetermined order of the first sound. In the second operation state, the sound input from the sound input means has been stored in the storage means for a predetermined time. The sound is in the first operating state,
A predetermined number is randomly selected by the voice selection means based on the random numbers, and is output to the voice output means in a predetermined order. In the third operation state, a voice is recognized based on the sound input from the sound input unit, and the voice is compared with the voice stored in the storage unit. Based on the result of the collation, the recognized voice is stored in the storage unit. Then, the sound stored in the storage means is output to the sound output means.

According to the present invention, as a voice processing device for verifying a predetermined feature of the recognized voice and the stored voice, a storage means for storing the voice; When the stored voice shares the predetermined feature and the stored voice is different before and after, the stored voice before and after the voice and the voice sharing the predetermined feature are further alternately predetermined. And a control unit for storing the association between the two stored voices in the storage unit on condition that the recognition is performed m times (m: a predetermined natural number) in time.

According to the present invention, there is provided a voice output means for outputting a voice, wherein the control means includes one of the associations of the stored voice sharing the predetermined feature with the voice recognized in the collation. One of the parameters, the parameter indicating the degree of association between the stored voice and the voice indicated by the association, when the voice indicated by the association is output to the output unit, The logarithmic relevance parameters attached to the generated voice and the voice indicated in the association are respectively increased and stored in the storage means.

[0015] In the present invention, the control means decreases the word relevance parameter by a predetermined amount every time a predetermined time elapses.

According to the above configuration of the present invention, in the collation performed before and after, the recognized voice and the stored voice share the predetermined characteristic, and the stored voice is In the case where the sound is different between before and after, the stored sound before and after and the sound sharing the above-mentioned predetermined feature are further alternately recognized m times (m: a predetermined natural number) within a predetermined time, so that the memory before and after is stored. The associated voice is associated. In the collation, one of the associations of the stored speech that shares a predetermined characteristic with the recognized speech is selected, and the speech indicated by the association is output to the output unit. It will be increased if it is done. At this time, the word relevance parameters of the stored voice and the output voice are respectively increased. Preferably, the word relevance parameter is decreased by a predetermined amount each time a predetermined time elapses.

According to the present invention, as a voice processing device for collating a predetermined feature between the recognized voice and the stored voice, a storage means for storing voice, and the recognized voice and the stored voice are stored in the memory. If you share certain features,
Control means for increasing a storage degree parameter relating to storage and holding of the stored voice, and replacing the stored voice and the recognized voice in the storage means in accordance with the magnitude of the storage degree parameter And

In the present invention, when the number of the stored voices reaches a predetermined number, the control means compares the voice stored in the storage means and the recognized voice with the storage degree parameter. The information is replaced in accordance with the size and stored in the storage means.

In the present invention, when the number of voices stored in the storage unit has reached the predetermined number, the control unit determines that the memory parameter of the voice stored in the storage unit is the minimum. The voice and the recognized voice are replaced and stored in the storage unit.

In the present invention, the control means includes a random number generating means for generating a random number, and the amount of increase of the memory parameter is changed randomly based on the random number generated by the random number generating means.

In the present invention, the control means reduces the storage parameter by a predetermined amount every time a predetermined time elapses.

According to the above configuration of the present invention, when a voice is recognized, the stored voice is collated with the recognized voice, and the recognized voice and the stored voice are compared with each other with a predetermined characteristic. Is shared, the memory parameter is increased. Then, when the number of stored voices reaches a predetermined number, the stored voice is replaced with a voice having the smallest storage parameter and the recognized voice, and stored in the storage unit. Preferably, the increase amount of the memory parameter changes randomly based on a random number generated by the random number generation unit. The storage parameter is reduced by a predetermined amount every time a predetermined time elapses.

In the present invention, the control means collates a voice recognized based on the sound input from the sound input means with a predetermined feature of the voice stored in the storage means, and determines a result of the collation. Based on the stored voice, the voice is stored or registered in the storage unit, and the voice registered among the stored voices is output to the voice output unit.

According to the configuration of the present invention, in the third operation state, a voice is recognized based on the sound input from the sound input means, and the recognized voice is stored in the storage means. A predetermined feature of the sound is compared. Then, based on the matching result, the voice is stored or registered in the storage unit. Input from the sound input means,
The registered voice of the voices stored in the storage means is output from the voice output means.

[0025] In the present invention, the control means may control the recognized voice on condition that the recognized voice and the registered voice stored in the storage means do not share a predetermined characteristic. Is stored in the storage means.

In the present invention, when the recognized voice and the registered voice stored in the storage unit do not share predetermined characteristics, the control unit performs the predetermined voice recognition. The recognized voice is stored in the storage means on condition that the number of times is performed.

[0027] In the present invention, when the recognized voice and the registered voice stored in the storage unit do not share a predetermined feature, the control unit controls the recognized voice and the recognized voice. A predetermined feature of the registered voice is collated a predetermined number of times, and the recognized voice and the registered voice in the collation do not share a predetermined feature. The voice is stored in the storage means.

In the present invention, when the recognized voice and the registered voice stored in the storage unit do not share a predetermined characteristic, the control unit continuously performs the control within a predetermined time. A predetermined feature of the two recognized voices is collated a predetermined number of times, and the recognized voice is stored in the memory under the condition that the voices recognized in the collation share a predetermined feature with each other. It is stored in the means.

According to the above configuration of the present invention, the recognized voice is stored in the storage means under a predetermined condition. One condition is that the recognized voice and the registered voice do not share predetermined characteristics. As another condition, the above-mentioned recognition of voice may be performed a predetermined number of times. Further, as another condition, a predetermined feature of the recognized voice and the registered voice is collated a predetermined number of times, and the recognized voice and the registered voice in the collation are a predetermined feature. May not share. Further, as another condition, a predetermined feature of two voices continuously recognized within a predetermined time is collated a predetermined number of times, and the voices recognized in the collation share a predetermined feature with each other. May be.

In the present invention, when the recognized voice and the unregistered voice stored in the storage unit share a predetermined feature, the control unit determines that the unregistered voice is unregistered. Is registered in the storage means.

In the present invention, the control means includes a random number generation means for generating a natural number random number n, wherein the recognized voice and the voice stored in the storage means and not registered have predetermined characteristics. In the case of sharing, the unregistered voice is registered in the storage means on condition that the recognition of the voice is performed n times.

In the present invention, the control means includes a random number generation means for generating a natural number random number n, wherein the recognized voice and the voice stored in the storage means and not registered have predetermined characteristics. When sharing, the predetermined feature of the recognized voice and the non-registered voice are collated n times, and the recognized voice and the unregistered voice in the verification are compared with a predetermined feature. The unregistered voice is registered in the storage means on condition that the feature is shared.

In the present invention, the control means includes a random number generating means for generating a natural number random number n, and the recognized voice and the voice stored in the storage means and not registered have predetermined characteristics. When sharing, a predetermined feature of voices continuously recognized within a predetermined time is collated a predetermined number of times, and the voices recognized in the collation share predetermined characteristics with each other. , The unregistered voice is registered in the storage means.

According to the above configuration of the present invention, the unregistered voice is registered in the storage means under a predetermined condition. One condition is that the recognized voice and the unregistered voice share predetermined characteristics. As another condition, the above-mentioned recognition of voice may be performed a predetermined number of times. Further, as another condition, predetermined characteristics of the recognized voice and the unregistered voice are collated a predetermined number of times, and the recognized voice and the unregistered voice in the collation are determined by a predetermined number. Features may be shared. Also, as other conditions,
There is a case where predetermined characteristics of two voices continuously recognized within a predetermined time are collated a predetermined number of times, and the voices recognized in the collation share predetermined characteristics with each other.

[0035]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, an embodiment of a voice processing apparatus according to the present invention will be described with reference to an example in which the present invention is applied to an electronic pet toy.

FIG. 1 is a diagram showing the appearance of a toy according to an embodiment of the voice processing device of the present invention. In FIG.
M8 indicates each part of the body of the toy in the form of a simulated creature, M1 indicates a head button, M2 and M3 indicate ears, and M4 indicates an ear.
Indicates a right hand, M5 indicates a left hand, M6 indicates a right foot, M7 indicates a left foot, and M8 indicates a main body.

The head button M1 is a button for controlling activation and sleeping in the present invention, and incorporates a head switch described later inside. When the user depresses the head button M1, a head switch described later can be turned on. The ears M2 and M3 have a built-in microphone (sound input means) described later, and the voice recognized by the toy is taken into the toy from here. Right hand M
Reference numeral 4 has a magnet built therein, which will be described later.
Has a built-in reed switch described below.
When the magnet approaches the left hand and the reed switch conducts, the communication switch described later is turned on. Further, the right hand M4 and the left hand M5 can be moved up and down in the directions of arrows a and b.
N. The right foot M6 and the left foot M7 are disposed on the side of the main body M8 so that the toy can stand upright on a horizontal plane.
8 is supported. The main body M8 includes a battery for operating the toy, and a board including a clock operation mechanism, a speaker, a control circuit, and the like to be described later.

FIG. 2 is a diagram showing a configuration of a toy according to an embodiment of the voice processing device of the present invention. In FIG. 2, 1 is a control circuit (registering means), 2 is a ROM (notification sound storage means), 3 is a FLASH ROM (storage means and temporary storage means), 4 is a microphone (sound input means), and 5 is a microphone. Is a pitch conversion circuit, 6 is an audio amplifier circuit, 7 is a speaker (audio output means), 8 is a head switch, 9 is a communication switch, 10 is an alarm switch, 11 is a reset switch, and 12 is a reset switch. Each shows a clock operation mechanism.

The control circuit 1 converts the analog audio signal output from the microphone 4 into digital audio data and
The audio data recorded on the ROM 2 and the FLASH ROM 3 are converted into an analog audio signal S 3 and output to the pitch conversion circuit 5. Further, the pitch of the audio signal output from the pitch conversion circuit is controlled by the control signal S4. Clock operation mechanism 1
The alarm operation in the alarm mode, which will be described later, is started by the control signal S1 by the control signal 2 and the time is measured by the pulse signal S2. In addition, voice recording and voice output are performed in accordance with control signals from the head switch 8, the communication switch 9, the alarm switch 10, and the reset switch 11. The operation of the control circuit 1 is executed based on a program read from the ROM 2.

The ROM 2 stores voice data to be uttered by the toy and a program of the control circuit 1 as data, and outputs data at an address designated by the control circuit 1 to the control circuit 1. FLASH ROM3
Records various data including audio data output from the control circuit 1 at a predetermined address specified by the control circuit, and outputs data at a predetermined address specified by the control circuit 1 to the control circuit 1. The microphone 4 outputs an audio signal to the control circuit 1. The pitch conversion circuit 5 outputs a signal obtained by converting the pitch of the audio signal S3 output from the control circuit 1 based on the control signal S3 to the audio amplification circuit 6. The audio amplification circuit 6 amplifies the signal output from the pitch conversion circuit 5 and outputs the amplified signal to the speaker 7. The speaker 7 converts an audio signal output from the audio amplification circuit 6 into audio and outputs the audio. The head switch 8, the communication switch 9, and the alarm switch 10 all output a signal for controlling an operation mode of a toy described later to the control circuit 1. Reset switch 11
Resets all the voices stored in the toy and the values of each parameter for controlling the operation of the toy, and sets the operation mode of the toy to an initial state. The clock operation mechanism outputs a signal S1 for controlling start and stop of the alarm operation to the control circuit 1, and outputs a pulse signal S2 having a constant cycle to the control circuit 1 as a pulse signal for time measurement.

Next, the operation of the toy having the above configuration will be described.

FIG. 3 is a diagram showing an operation mode related to the operation at the time of the first activation of the toy according to the embodiment of the present invention. FIG.
In the table, M0 indicates an initial startup mode, and M indicates a normal mode. Also, I0 to I2 indicate the mode transition in each mode.

The initial start mode M0 is an operation mode in a state where a battery is inserted for the first time or a state where the reset switch 11 is turned on. In this first boot mode,
The operation of registering the name of the toy itself (hereinafter referred to as a pet name) and the name of the user of the toy (hereinafter referred to as an owner name) to the toy is performed. The normal mode M is a mode in which a normal operation of the toy described later is performed.

When the battery is inserted into the toy for the first time, the operating state of the toy is the initial activation mode M0. At the beginning of this mode, all parameters described later are in the initial state, and all registered words described later are also deleted.
In this mode, the process of registering the pet name and the owner name in the toy is performed. When the registration is completed, the operation mode is shifted to the normal mode M by the transition I2. When the mode shifts to the normal mode, an operation of a normal mode M described later is executed. In the normal mode M, the state of the reset switch 11 is constantly monitored, and when the reset switch 11 is turned on, the operation mode is immediately shifted to the initial startup mode M0 by the transition I1.
At this time, all parameters and registered words described later are lost. The transition I0 indicates a mode transition due to a temporary stop of power supply to the internal circuit due to replacement of a battery or the like. Even if the power supply is stopped in the normal mode M due to battery replacement or the like, if the power supply is started again, the operation mode remains unchanged in the normal mode M. That is, parameters and registration words described later are FLASH
It is stored in the ROM 3. Even in the first activation mode M0, the mode transition by the transition I0 does not occur.

FIG. 4 is a diagram showing details of the operation mode in the normal mode M. In FIG. 4, M11 indicates a wake-up mode, M12 indicates a sleeping mode, M13 indicates an alarm mode, and M14 indicates a communication mode.

The wake-up mode M11 includes a voice recognition process for learning conversation and words, a random utterance process for randomly uttering the stored words, an automatic voice registration process for automatically storing the sound input to the microphone 4, and the like. Is a mode for performing the processing described later.

The sleeping mode M12 is a mode in which all operation processes such as conversation in the wake-up mode M11 are stopped.

The alarm mode M13 is a voice alarm for notifying the user that the set time has come by uttering registered words and the like as an alarm when a predetermined time set in advance by the user is reached. This is the mode in which the operation is performed. After the toy emits an alarm sound, the user inputs the words stored in the toy into the microphone 4 of the toy.
By talking to the user and recognizing the words, the alarm of the time signal can be temporarily stopped without touching the toy.

The communication mode M14 is a mode in which the words stored in the toy are uttered together while the toys are brought close to each other, and the sound being uttered is stored. Such an operation gives the user an impression as if the toys are autonomously communicating. FIG. 5 is an overhead view showing the arrangement of the toys in the communication mode M14. At the tips of the left hand M5 and the right hand M4 of the toy, a reed switch 91 and a magnet 92 are respectively built. When the magnet 92 of the other toy approaches the reed switch 91 of one toy and the reed switch 91 conducts and the communication switch 9 is turned on, the operation mode shifts to the communication mode M14.

Next, the transition between the respective modes in the normal mode M will be described.

When the above-described operation in the first activation mode M0 is completed, the operation mode is changed to the wake-up mode M11 in transition I2.
Move to Timing starts with the transition to the wake-up mode M0, and the elapsed time when the operation mode is in the wake-up mode is measured. If the elapsed time exceeds, for example, 10 minutes, the operation mode shifts to the sleep mode M12 in transition I3.

In the sleep mode M12, when the head button M1 of the toy is pressed and the head switch 8 incorporated in the head button M1 is turned on, the operation mode shifts to the wake-up mode M11 at transition I4. When the head switch 8 is turned on in the wake-up mode M11, the operation mode does not change if the toy does not emit any sound. In this case, the elapsed time in the wake-up mode M11 described above is reset, and the wake-up mode M11 is changed to the sleep mode M11.
The time to shift to 12 is extended by 10 minutes. On the other hand, in the wake-up mode M11, if the head switch 8 is turned on in a state where the toy is making some sound, the operation mode is forcibly shifted to the sleep mode M12.

When the left hand M5 is rotated in the direction of arrow a in FIG. 1 and the alarm switch 10 is turned on, the wake-up mode M
11. A transition from the sleep mode M12 and the communication mode M14 to the alarm mode M13 is made in transition I5. The alarm mode M13 is a mode having the highest priority among the other three modes in the normal mode M. When the right hand M4 or the left hand M5 is rotated in the direction opposite to the arrow a and the alarm switch 10 is turned off, the operation mode shifts to the sleep mode M12 at transition I6.

When the reed switch 91 is turned on and the communication switch 9 is turned on, the wake-up mode M11 and the sleeping mode M12 shift to the communication mode M14 at the transition I7. When the alarm switch 10 and the communication switch 9 are simultaneously turned on, the alarm mode M13 by the alarm switch 10 is set.
Does not shift to the communication mode M14.

There are three operation modes relating to the reliability parameters described later, independently of the operation modes relating to the initial start-up mode M0 and the normal mode M described above. FIG. 6 is a diagram illustrating an operation mode according to the reliability parameter. In FIG. 6, M21 is a health mode, M22 is a disease mode,
M23 indicates the asphyxia mode, respectively.

After the first battery insertion and after the reset switch 11 is turned on, the mode relating to the reliability parameter is the health mode M21. When the value of the confidence parameter fluctuates due to a condition described later and exceeds a threshold value that defines the boundary between the health mode M21 and the disease mode M22, the transition I
At 9, the mode is shifted from the health mode M21 to the disease mode M22. In the sick mode M21, a part of the operation processed in the health mode M21 is stopped, or the toy makes a sick voice as if the toy became sick to the user. Make the impression that you are. If the user of the toy recovers the confidence parameter within a predetermined threshold value by continuing the action such as making the toy recognize words, the illness mode M22 is made in transition I11.
To the health mode M21. Conversely, when the confidence parameter fluctuates and exceeds a threshold value that defines the boundary between the illness mode M22 and the asphyxia mode M23, the state shifts from the sickness mode M22 to the asphyxia mode M23 in transition I10. In the asphyxia mode M23, most of the processing performed in the health mode M22 is not performed, all stored words are deleted, and parameters described later are not updated. Turn on reset switch 11
Then, the mode relating to the confidence parameter is sick mode M
22 and a transition to the health mode M21 from the asphyxia mode M23 in transition I1.

Next, a specific operation in each of the above operation modes will be described.

FIGS. 7 and 8 are flow charts for explaining the processing in the initial startup mode M0. 7 and 8, Pa1 to Pa54, Pb and Pc indicate steps. FIG. 7 and FIG. 8 show the connection relation of the steps.

In the process in the first activation mode M0, first, voice data input from the microphone 4 is registered as a pet name. In that case, to prevent the wrong voice from being registered, the user is asked to input the same voice twice to confirm the voice to be registered. If the characteristics of the voice data match twice, the pet is Registered as first names. Following the registration of the pet name, the owner's name is also registered. This registration procedure is similar to the registration of the pet name. The user is required to input the same voice twice, and when the characteristics of the voice data match twice, the user is registered as the owner name.

In step Pa1, the initial start mode M
0 processing is started. In step Pa2, a process in which the toy emits sound is executed at the beginning of the initial startup mode. In this step, since the audio data has not been recorded yet, the audio is generated from the audio data recorded in the ROM 2 in advance. Specifically, ROM2
The digitized data of the voice recorded in the ROM 2, for example, "Hiyahaha" is output to the control circuit 1 from a predetermined address of the ROM 2 in response to a request from the control circuit 1. The audio data output to the control circuit 1 is subjected to D / A conversion, converted into an analog audio signal, and output to the pitch conversion circuit 5 as a signal S3. The sound signal output to the pitch conversion circuit 5 is converted in pitch according to the signal S4 output from the control circuit 1, amplified by the voice amplification circuit 6, and converted into sound waves by the speaker 7. By emitting a predetermined sound from the toy in step Pa2, the user can recognize that the toy is operating.

In step Pa3, a waiting time parameter described below is reset. In step Pa4, a predetermined time interval, for example, 2
In response to the pulse signal S2 input every second, an interrupt process for incrementing the waiting time parameter is started. Since the value of the waiting time parameter has been reset in Step Pa3, the elapsed time from Step Pa4 can be measured by referring to this value.

In step Pa5, the value of the integer k used in step Pa10 described later is reset to zero.

In step Pa6, the control circuit 1 determines whether or not there is a voice input from the microphone 4.
Specifically, when the level of the audio signal input from the microphone 4 is smaller than a predetermined threshold, it is determined that the audio signal is not input, and when the level is higher than the predetermined threshold, the audio signal is input. Is determined. If it is determined in the control circuit 1 that no audio signal has been input, the process proceeds to step Pa7. In step Pa7,
The elapsed time from step Pa4 indicated by the value of the above-mentioned waiting time parameter is compared with a predetermined time, for example, 10 seconds, and when the elapsed time is shorter than 10 seconds,
The process returns to step Pa6 again, and the presence or absence of the voice input from the microphone 4 is detected. If a voice is detected in Step Pa6 within 10 seconds, the interrupt processing for incrementing the waiting time parameter is stopped, and the processing shifts from Step Pa6 to Step Pa8. If no voice is detected in Step Pa6 within 10 seconds, the interrupt process for incrementing the waiting time parameter is similarly stopped, and the process shifts from Step Pa7 to Step Pa25 described later.

At step Pa8, the analog audio signal detected at step Pa6 is subjected to A / D conversion by the control circuit 1 and converted into digitized audio data. At this time, for example, when the volume of the voice is larger than the upper threshold set in the control circuit 1, or smaller than the lower threshold, the voice length is set to the upper limit. If it is longer than the threshold, step P
The process shifts to a9, and a sound is issued to notify that the level of the input sound is inappropriate. More specifically, for example, data of the sound “B” is read from the ROM 2, D / A-converted, and transmitted as a sound from the speaker 7 through the pitch conversion circuit 5 and the sound amplification circuit 6.

Next, at step Pa10, the value of the integer k is incremented by one, and
At a11, the magnitude relationship between the integer k and the integer 1 is determined.
If the integer k is equal to or smaller than a predetermined integer, for example, 1, the process returns to step Pa6 to wait for the input of a voice again. If the integer k is larger than 1, the process returns to step Pa25 described later. Processing is shifted. As described above, when the level of the sound converted in step Pa8 is inappropriate, the opportunity to re-input the sound a predetermined number of times is provided, so that when the user inputs an incorrect sound, The inconvenience of repeating the procedure is reduced.

At step Pa12, the voice data obtained at step Pa8 is substituted for the variable Vd. This audio data is referred to in later processing.

At step Pa13, a voice for notifying the user that the voice data has been recognized, for example, "un?"
Is read from the ROM 2 and D / A-converted. The sound data is transmitted as a sound from the speaker 7 through the pitch conversion circuit 5 and the sound amplification circuit 6. As described above, different voices (in the above example, “bu-” and “un?”) Are output according to the result of recognition of the input voice by the toy, so that the user interacts with the toy. Sound can be registered in the toy with a sense.

At step Pa14, the above-mentioned waiting time parameter is reset again, and at step Pa15
In, an interrupt process for incrementing the waiting time parameter is started according to the pulse signal S2 input from the clock operating mechanism 12 at predetermined time intervals. In step Pa16, the value of the integer k used in step Pa21 is reset to zero.

In step Pa17, the control circuit 1 determines again whether or not there is a voice input from the microphone 4, and if it is determined that no voice signal has been input, the process proceeds to step Pa18. Step Pa18
Then, the elapsed time from step Pa15 indicated by the value of the waiting time parameter described above is compared with a predetermined time, for example, 10 seconds. If the elapsed time is shorter than 10 seconds, the process is returned to step Pa17 again. , The presence or absence of voice input from the microphone 4 is detected. If a voice is detected in Step Pa17 within 10 seconds, the interrupt process for incrementing the waiting time parameter is stopped, and the process proceeds from Step Pa17 to Step Pa19. If no voice is detected in step Pa17 within 10 seconds, the interrupt process for incrementing the waiting time parameter is similarly stopped, and the process proceeds from step Pa18 to step Pa25 described later.

In step Pa19, step Pa17
The analog audio signal detected by the control circuit 1
It is subjected to D conversion and converted to digitized audio data. If it is determined that the voice input under the same conditions as in step Pa8 described above is inappropriate, step Pa20
The processing is shifted to the step S3, and the voice that informs that the level of the input voice is inappropriate, for example, the voice data “B” is read out from the ROM 2 and D / A-converted. The sound is emitted from the speaker 7 through the amplifier circuit 6. Next, in step Pa21, the value of the integer k is incremented by one, and in step Pa22, the magnitude relationship between the integer k and a predetermined integer is determined. If the integer k is equal to or smaller than a predetermined integer, for example, 1, the process returns to step Pa17 and waits for the input of a voice again.
If it is larger, the process proceeds to Step Pa25 described later. Similarly to steps Pa9 to Pa11,
Since an opportunity is provided to re-enter the sound a predetermined number of times when the sound level is inappropriate, the inconvenience of the user is reduced.

In step Pa23, step Pa12
And the voice data substituted for the variable Vd at step Pa19
Are compared in the control circuit 1. The features of the audio data are extracted using, for example, the amplitude of a frequency spectrum obtained by performing Fourier transform on the audio data, and are compared by a predetermined comparison procedure. Variable V
If it is determined that the voice data of step d and the characteristics of the voice data acquired in step Pa19 match, the process proceeds to step Pa24, and the voice data acquired in step Pa19 is used as the pet name in the FLASH ROM.
3 is written to a predetermined address. Conversely, if it is determined that the characteristics of the audio data do not match each other, the process proceeds to step Pa20 described above and proceeds to step Pa17.
Is given to the user again.

When the pet name is registered in step Pa24, the process proceeds to step Pa28, and a sound for notifying the user that the pet name has been registered, for example, a sound registered as the pet name is issued. As a result, the user can know that the pet name has been successfully registered, and can confirm the registered voice.

In this case, the voice data written in the FLASH ROM 3 is data of a voice uttered by the user himself, but the voice signal output from the pitch conversion circuit 5 is
Since the pitch can be controlled by the signal S4, the sound can be converted to an adorable sound such as a parakeet, and the sound can be output from the speaker 7. As a result, it is possible to avoid the unpleasantness of listening to the user's own voice, and to give the user the impression that the toy is emitting his own voice. Increase.

Steps Pa25 to Pa27 are used to randomly shift to another processing when the audio input waiting time exceeds 10 seconds or when the audio data is inappropriate due to the sound being too loud. Step. The procedure for shifting at random is based on steps Pa2 to P
a24: a pet name registration procedure, a random utterance process Pb described later, and an automatic voice registration process Pc.

At step Pa25, an integer from 0 to 2 is randomly generated and assigned to a variable m. In step Pa26, it is determined whether the numerical value assigned to the variable m matches the integer 0, and if they match, the process returns to step Pa2, and the process of registering the pet name again is executed from the beginning. If the variable m and the integer 0 do not match, the process proceeds to step Pa27, where it is determined that the variable m and the integer 1 match, and if they do match, the process proceeds to a random utterance process Pb to be described later. If not, the process proceeds to an automatic voice registration process Pc described later. As described above, even in a state where the pet name is left unregistered in the first activation mode, processes such as a random utterance process Pb and an automatic voice registration process Pc to be described later are randomly executed. According to such a process, the mechanical impression of the user on the toy can be reduced as compared with a process in which the pet name is not registered, and the registration is urged forever.

Here, the random speech processing Pb and the automatic voice registration processing Pc will be described.

In the random speech processing Pb, the control circuit 1 randomly reads out voice data prepared in advance stored in the ROM 2 of the toy or voice data registered in the FLASH ROM 3 by the user. Then, the sound is emitted as sound from the speaker 7 through the pitch conversion circuit 5 and the sound amplification circuit 6. When the operation mode is the initial startup mode, since voice data has not yet been registered in the FLASH ROM 3 by a user's registration operation, the voice data automatically registered by the automatic voice registration process Pc to be described later, or the ROM 2 The recorded audio data prepared in advance is exclusively output.
The type of the voice data selected at random by the random utterance process Pb changes depending on a mood parameter and a reliability parameter described later.

FIG. 17 is a flowchart for explaining the processing in the random speech processing Pb. In FIG. 17, Pb1 to Pb22 indicate each step of the processing. The random utterance process Pb starts in step Pb1, substitutes an integer randomly generated in a range of 1 to 10 in step Pb2 into a variable m, and executes steps Pb3 to Pb
At 11, it is determined whether the variable m matches the integers 1 to 9, and the process proceeds to steps Pb12 to Pb21 according to the matched integer. Then, the stored sound is output from the speaker 7 according to the processing determined in each step (speech processing 1 to 10). In each utterance processing, for example, various synthetic sounds stored in the ROM 2 in advance,
The FLASH ROM 3 can be operated by the user to
Randomly select various registered words stored in
You can make them speak at random times. Further, it is possible to randomly generate a pair of each voice described later. Furthermore, the voice registered in the automatic voice registration mode described below can be selected at random and made to speak.
Alternatively, the owner name and pet name stored in the toy in the initial startup mode can be spoken first, and then other sounds can be uttered continuously.The toy is something with the owner name and pet name as subjects. You can also make them speak as if they were talking. Thus, the toy is ROM2 or F
Various voice processing procedures are provided for the voice stored in the LASH ROM 3 and the voice is output by randomly selecting them, so that the user can be surprised and the toy will not get tired. it can.

In the automatic voice registration process Pc, voices within a predetermined length inputted to the microphone 4 are automatically converted into voice data by the control circuit 1 and FLASH
Recorded in the ROM 3. The number of audio data to be registered is set to a predetermined maximum number, for example, three or less. If the number of audio data to be registered exceeds this, any one of the recording areas where audio data is recorded in the automatic audio registration process Pc. Is overwritten and recorded. The voice data registered in the automatic voice registration process Pc is uttered in the above-described random utterance process Pb and in a conversation process described below, so that the content of the voice uttered by the toy increases, and as a result, The mechanical impression is reduced, and the taste as a simulated creature can be enhanced.

In the course of the pet name registration process, the process shifts to the above-mentioned random utterance process Pb or the automatic voice registration process Pc. In addition, a number of other different processes can be provided to shift the process randomly. This can reduce the disadvantage that the toy is bored by forcing the user to perform a single process.

Next, the processing after the completion of the pet name registration in the first activation mode M0 will be described.

After the pet name is registered, the owner name is registered by the same steps. Specifically, steps Pa29 to Pa49 and steps Pa3 to Pa23
Then, the same processing is performed respectively.

At step Pa52, an integer from 0 to 2 is randomly generated and assigned to a variable m. In step Pa53, it is determined whether the numerical value assigned to the variable m is equal to the integer 0, and if they match, the process returns to step Pa28, in which the pet name is uttered and then the owner name is registered again. Executed from the beginning. If the variable m and the integer 0 do not match, the process proceeds to step Pa54, where it is determined that the variable m matches the integer 1, and if they do match, the process proceeds to the above-described random utterance process Pb. And if not, the process proceeds to the automatic voice registration process Pc described above. Through the same procedure as steps Pa25 to Pa27 in the registration of the pet name, processes such as the random utterance process Pb and the automatic voice registration process Pc are executed at random, so that the mechanical impression of the user on the toy is reduced. be able to.

In step Pa49, step Pa
If it is determined that the characteristics of the voice data acquired in step 45 and the voice data of the variable Vd match, step Pa50
The audio data obtained in step Pa45 is written to a predetermined address of the flash ROM 3 as the owner name. Thereafter, a sound indicating that the owner name has been registered, for example, a registered owner name is output. As with the process for registering pet names, FLASH ROM
When the voice data registered as the owner's name in 3 is converted to voice, the pitch is changed by the pitch conversion circuit 5, so that the user is given the impression that the toy is emitting his own voice. Can be.

When the registration of the pet name and the owner's name is completed, the operation of the initial activation mode M0 is completed, and the operation mode is shifted to the wake-up mode M11 by the transition I2.

Next, the operation in the wake-up mode M11 will be described. FIG. 9 is a flowchart illustrating processing in the wake-up mode M11. 4 and 9 indicate the same contents. In addition, Pd1 to Pd17 indicate steps.

In the wake-up mode M11, in addition to the above-described random utterance processing Pb and the automatic voice registration processing Pc, a voice recognition processing Pe, which will be described later, is randomly selected and executed by the generated random numbers. At this time, when the mode based on the trust parameter is the sick mode M21 described above, the random utterance process Pb and the automatic voice registration process Pc are not executed, and a voice that sometimes indicates the sickness is generated. In the temporary death mode M23, the random utterance process Pb, the automatic voice registration process Pc, and the voice recognition process Pe, which will be described later, are all not executed, and a voice notifying the temporary death mode is issued. Also, the status of the head switch 8, the communication switch 9, and the alarm switch 10 are monitored,
The processing is shifted to each operation mode according to the state of the switch. Further, the time from when the process is shifted to the wake-up mode M11 is measured, and when a predetermined time has elapsed, the process is shifted to the sleeping mode M12.

When the process is started in step Pd1 of the wake-up mode M11, in step Pd2, the wake-up time parameter described below is reset to zero. In step Pd3, an interrupt process for incrementing the wake-up time parameter is started in response to a pulse signal S2 input from the clock operation mechanism 12 at predetermined time intervals, for example, every two seconds. Since the value of the wake-up time parameter has been reset in step Pd2, the elapsed time from step Pd3 can be measured by referring to this value.

At step Pd4, it is confirmed whether or not the mode relating to the reliability parameter is the temporary death mode M23. In the case of the temporary death mode M23, a voice notifying the user of the temporary death mode is issued in step Pd5. More specifically, audio data representing a suspended state recorded at a predetermined address in the ROM 2 is read out, D / A-converted, pitch-converted by a pitch conversion circuit 5, passed through a voice amplification circuit 6 and passed through a speaker. 7 is output as audio. Thereafter, a step Pd1 described later.
The process is shifted to 1.

In step Pd4, the suspended-matter mode M23
If not, it is next checked in step Pd6 whether the mode is the sick mode M22.
If it is in the sick mode M22, step Pd7
At step S22, a voice is issued to notify the user that the user is in the sick mode M22. More specifically, voice data expressing a disease state recorded at a predetermined address in the ROM 2 is read out, D / A converted, the pitch is converted by a pitch conversion circuit 5, passed through a voice amplification circuit 6, and passed through a speaker. 7 is output as audio. Thereafter, the process proceeds to a speech recognition process Pe described later.

In steps Pd4 and Pd6, the modes related to the reliability parameters are not the asphyxia mode M23 and the sick mode M22, ie, the health mode M2.
After it is confirmed that the value is 1, Step Pd8-
In Pd10, a speech recognition process Pe, a random utterance process Pb, and an automatic speech registration process Pc are randomly selected and executed. Specifically, at step Pd8, 0
An integer of ~ 2 is randomly generated and assigned to a variable m. Next, in step Pd9, a match between the integer 0 and the variable m is determined.
e is transferred to step e.
The process is shifted to 0. In step Pd10, it is determined that the integer 1 matches the variable m. If the two match, the process proceeds to the above-described random utterance process Pb, and if they do not match, the process proceeds to the above-described automatic voice registration process Pc. Therefore, any one of the speech recognition processing Pe, the random speech processing Pb, and the automatic speech registration processing Pc is selected and executed.

Speech recognition processing Pe, random speech processing Pb
After the completion of the automatic voice registration process Pc, the process proceeds to Step Pd11. In step Pd11, the state of the alarm switch 10 is confirmed, and if the alarm switch 10 is ON, the process is shifted to the alarm mode M13 by the above-described transition I5. If it is confirmed that the alarm switch 10 is turned off, the process proceeds to Step Pd12.

At step Pd12, the state of the communication switch 9 is confirmed. If the communication switch 9 is turned on, step Pd1
3, the state of the mode relating to the reliability parameter is confirmed, and if the mode is not in the suspended
7, the processing shifts to the communication mode M14. When the mode related to the reliability parameter is the temporary death mode, the process does not shift to the communication mode M14, and the process shifts to the next step Pd14. Also, when it is confirmed in step Pd12 that the communication switch 9 has not been turned ON, the process proceeds to step Pd14.

In step Pd14, it is confirmed whether or not the sleep mode has been selected. When the sleep mode is selected, the process is shifted to the sleep mode M12 by the above-described transition I3, and when not selected, the process is shifted to Step Pd15 described later.

Whether the sleep mode has been selected or not is determined by
This is done by checking the state of the sleep mode selection flag. When the head switch 8 is turned on, an interruption process occurs, and in that process, a sleep mode selection flag is set. Specifically, when the head switch 8 is turned on while the voice signal is being output from the control circuit 1 in the random speech processing Pb, the voice recognition processing Pe, and the communication mode M14, and the above-described interrupt processing occurs, this interrupt is generated. In the process, a sleep mode selection flag is set. Then, in step Pd14, the state of the sleep mode selection flag is determined, and if it is confirmed that the sleep mode selection flag is set, the process proceeds to the sleep mode M12 in transition I3, and the sleep mode selection flag is set. If it is confirmed that the process has not been performed, the process proceeds to Step Pd15. That is,
In the random speech processing Pb, the voice recognition processing Pe, and the communication mode M14, when the toy is hit and the head switch 8 is turned on while the toy is emitting sound, the operation mode of the toy is shifted to the sleeping mode. You. When the state of the sleep mode selection flag is confirmed in step Pd14 and the process proceeds to the next process, the sleep mode selection flag is reset. By the above processing, the user can stop the sound of the toy by hitting the head of the toy when the sound generated by the toy becomes harsh.

At step Pd15, it is confirmed whether or not the extension of the wake-up time is selected. If the extension of the wake-up time is selected, the process proceeds to step Pd16, and after the wake-up time parameter is reset, the process proceeds to step Pd17. If the extension of the wake-up time has not been selected, the process proceeds to Step Pd17. Step Pd17 to be described later
From the wake-up mode M11 to the sleep mode M1 when the time for measuring the wake-up time parameter exceeds a predetermined time.
The processing is shifted to 2, but when the wake-up time parameter is reset in step Pd16, the measured time is returned to zero. Is extended.

Whether the extension of the wake-up time is selected or not is determined by checking the state of the wake-up time extension flag. When the head switch 8 is turned on, an interrupt process occurs as in the case of the sleep mode selection flag,
In the process, a wake-up time extension flag is set.
Specifically, in the wake-up mode M11 in a period other than the above-described conditions for setting the sleeping mode selection flag (when a voice signal is output from the control circuit 1 in the random speech processing Pb or the voice recognition processing Pe), When the switch 8 is turned on and the above interrupt processing occurs,
In this interrupt processing, the wake-up time extension flag is set. Then, in step Pd15, the state of the sleep mode selection flag is determined, and if it is confirmed that the wake-up time extension flag is set, the process proceeds to step Pd
If the wake-up time parameter is reset at 16 and it is confirmed that the wake-up time extension flag has not been set, the process proceeds to step Pd17. That is, when the toy head is hit in the wake-up mode M11 in a period other than when the toy is emitting a voice in the random speech processing Pb, the voice recognition processing Pe, and the communication mode M14 to turn on the head switch 8,
The time until the operation mode is shifted to the sleep mode M12 can be extended. When the state of the wake-up time extension flag is confirmed in step Pd15 and the process proceeds to the next process, the wake-up time extension flag is reset. By the above processing, the user can wake up mode M11 as much as desired.
Can be extended to enjoy the operation of the toy.

When the head switch 8 is turned on, a process for causing the toy to utter predetermined voice data can be provided. Thereby, the reaction of the toy caused by the user touching the head of the toy can be enjoyed. Also, different sounds can be generated according to the time during which the head switch is ON. For example, a pulse having an appropriate period is provided in the control circuit 1 and the number of pulses is counted by the control circuit 1 during the period from when the head switch is turned on to when the head switch is turned off. The sound to be uttered by the toy can be changed according to the number of pulses counted by such a method. For example, turning on the head switch
When the time is between 0.03 seconds and less than 0.4 seconds, it is considered that the head has been hit, and the voice expressing anger is uttered by the toy, and the head switch is turned ON for more than 0.4 seconds. In this case, the process of causing the toy to utter a voice expressing pleasure by regarding the head as stroking is executed in an interrupt process or a branch process that occurs when the head switch 8 is turned on. By incorporating a sensor for detecting pressure in the head switch, the sound can be changed according to the magnitude of the pressure. Further, the type of voice data can be changed by a mood parameter to be described later to express pleasant or unpleasant mood by voice. By the above processing,
Since the toy can respond to the operation of the user by emitting various sounds, the mechanical impression of the toy is weakened,
The taste as a simulated creature can be enhanced.

Next, the operation of the speech recognition processing Pe will be described. FIG. 10 is a flowchart illustrating processing in the voice recognition processing Pe. In FIG.
1 to Pe16 and Pf, Pg, Ph indicate steps.

First, the outline of the speech recognition processing Pe will be described. In the voice recognition processing Pe, the voice data input from the microphone 4 and the features of the voice data recorded in the FLASH ROM 3 are compared. If the voice data having the matching features is recorded in the FLASH ROM 3, the speech processing Ph is performed. Let it run. In the conversation processing Ph, the voice input from the microphone 4 is read from the ROM 2 and the FLASH RO.
The voice data recorded in M3 is converted into voice in various combinations and uttered. Therefore, the user can enjoy an unexpected voice response from the toy to the voice input to the microphone. On the other hand, audio data whose characteristics match the audio data input from the microphone 4 is FLA
If not recorded in the SH ROM 3, a temporary registration process Pf and a registration extension process Pg, which will be described later, are executed. As a result, the sound input from the microphone 4 is converted into sound data by the control circuit 1 and recorded on the FLASH ROM 3. The temporary registration process Pf and the registration extension process Pg include a process of randomly changing a process until voice data is recorded in the FLASH ROM 3, and provide a user with an interest in an action of causing a toy to remember a voice. I have it.

In the speech recognition processing Pe, the microphone 4
The mood parameter is incremented each time the control circuit 1 converts the sound input into the mood parameter. The mood parameter is a parameter that changes the type of voice to be uttered in the above-described random utterance processing Pb or the type of voice to be uttered when the head switch 8 is turned on.

FIG. 18 is a diagram showing an example of a statement changing according to the value of the mood parameter. In this figure, the contents of remarks of the toy in the range of the value of each mood parameter are shown. If the head switch 8 is turned on when the pet is not speaking, it is determined whether the user has hit or struck according to the ON time, and the speech content is changed accordingly, and the mood parameter is further changed. The content of the remark is changed according to. Also, the above-mentioned random comment processing Pb
In, two types of sounds can be output according to the probability even with the same mood parameter. In this way, the mood of a constantly moving creature can be expressed by changing the content of a statement.

The mood parameter is related to the above-mentioned confidence parameter, and when the mood parameter is raised to a predetermined maximum value, the confidence parameter is incremented. That is, each time a voice is recognized in the voice recognition processing Pe, the mood parameter increases, and the reliability parameter increases in accordance with the mood parameter. On the other hand, the mood parameter and the reliability parameter are decremented every predetermined time in the interrupt processing generated every predetermined time by counting the pulse signal S2 of the predetermined cycle output from the clock operation mechanism 12, so that the toy in the voice recognition processing Pe is performed. If the user neglects the operation of recognizing the voice, the type of voice generated by the toy changes due to the decrease in the mood parameter, and the operation mode of the toy automatically becomes ill after a certain period due to the reduction in the reliability parameter. The mode shifts to the mode M22 or the temporary death mode M23, and a part of the normal processing becomes inoperable. As described above, since the process of causing the toy to recognize the voice and the other behavior of the toy are associated with each other, the user is given an interest in the act of continuously giving a voice to the toy and taking care of the toy. I have.

Hereinafter, a detailed operation of the speech recognition processing Pe will be described.

The speech recognition processing Pe is started from step Pe1. First, the voice recognition time parameter is reset in step Pe2, and thereafter, in step Pe3, an interrupt for incrementing the voice recognition time parameter in response to a pulse signal S2 input from the clock operating mechanism 12 at predetermined time intervals, for example, every two seconds. Processing is started. The value of the speech recognition time parameter is determined by the step Pe.
2, the elapsed time from step Pe3 can be measured by referring to this value.

After the start of the measurement of the voice recognition time, the control circuit 1 determines whether or not there is a voice input from the microphone 4 in step Pe4. Specifically, when the level of the audio signal input from the microphone 4 is smaller than a predetermined threshold, it is determined that the audio signal is not input,
When it is larger than the predetermined threshold value, it is determined that the audio signal is being input. If the control circuit 1 determines that no audio signal has been input, the process proceeds to step Pe5. In step Pe5, the elapsed time from step Pe3 indicated by the value of the above-described voice recognition time parameter is compared with a predetermined time, for example, 30 seconds.
The process is returned again to detect the presence or absence of the voice input from the microphone 4. If a voice is detected in step Pe4 within 30 seconds, the interrupt process for incrementing the voice recognition time parameter is stopped, and the process shifts from step Pe4 to step Pe6. If no voice is detected in step Pe4 within 30 seconds, the interrupt processing for incrementing the voice recognition time parameter is similarly stopped.
The process proceeds from step Pe5 to step Pe16, and the voice recognition process ends.

In step Pe6, the analog audio signal detected in step Pe4 is subjected to A / D conversion by the control circuit 1 and converted into digitized audio data. At this time, for example, when the volume of the voice is larger than the upper threshold set in the control circuit 1, or smaller than the lower threshold, the voice length is set to the upper limit. If it is longer than the threshold, step P
The process proceeds to e7, and a sound is generated to notify that the level of the input sound is inappropriate. More specifically, for example, data of the sound “B” is read from the ROM 2, D / A-converted, and transmitted as a sound from the speaker 7 through the pitch conversion circuit 5 and the sound amplification circuit 6. Thereafter, the process returns to step Pe7, and the process of waiting for the input of the voice is performed again. In step Pe6, when the audio signal from the microphone 4 is normally converted into audio data, the acquired audio data is substituted for the variable Vd.

After the voice data is substituted for the variable Vd, the above-mentioned mood parameter is incremented, for example, by 1 in step Pe9, and the process proceeds to step Pe10. Thereafter, in step Pe10, it is confirmed whether or not the operation mode of the toy is in the sick mode M22. When the toy mode is in the sick mode M22, voice data indicating that the toy is in the sick mode is read from the ROM 2 in step Pe11. The audio signal is converted into an audio signal, and is output as a sound from a speaker 7 through a pitch conversion circuit 5 and an audio amplification circuit 6. Then, after incrementing the number of times of voice recognition in step Pe12, the process is returned to step Pe4, and the process of waiting for the input of voice again is executed. As described above, when the toy is in the sick mode M22, a voice registration process described later is not executed.

If it is determined in step Pe10 that the toy is not in the sick mode M22, the process proceeds to step Pe13. In step Pe13, the characteristics of the audio data substituted for the variable Vd in step Pe8 and the characteristics of the audio data already registered in the FLASH ROM 3 (hereinafter, referred to as registered words) are compared. Then, when a registered word that is determined to have the same voice data characteristics by a predetermined determination method is found, the number indicating the storage location of the registered word is substituted into the variable J as the registered word number (step Pe1
4). Next, it is determined whether or not the acquired registration word is a temporary registration word described later (step Pe15),
If the word is not a temporary registration word, after the above-described conversation processing Ph is executed, the process returns to step Pe2 again, and a process of waiting for a speech input is executed.

In the search in step Pe13,
If no voice data whose characteristics match the voice data assigned to the variable Vd is found in the FLASH ROM 3, the input word is input again by the user.
Temporary registration process P for storing in a toy as a registration word
f and registration extension processing Pg are executed. Details of these processes will be described later. When the registration extension process Pg ends, the process returns to step Pe2, and the process of waiting for the input of a voice is executed again, similarly to the conversation process Ph.
When it is confirmed in step Pe15 that the registered word searched in step Pe14 is the temporarily registered word, the registration extension process Pg is executed instead of the conversation process Ph.

Next, the operation of the above-described temporary registration process Pf will be described. FIG. 11 and FIG. 12 are flowcharts for explaining processing in the temporary registration processing Pf. 11 and 12, Pf1 to Pf33 indicate steps. FIG. 11 and FIG. 12 show the connection relation of the steps.

In the temporary registration process Pf, the voice input from the microphone 4 is used as a registration word in FLASH RO
The process of recording in M3 is performed. At this time, in order to confirm whether or not the user has the intention to store the voice input in step Pf5 as a registration word in the toy, a process of inputting the same voice a plurality of times is performed. When it is determined that the characteristics of the input audio data match, the audio data is recorded in the FLASH ROM 3 as a registered word. By performing the above-described processing, it is possible to prevent inconvenience in which a voice that has no intention to be registered is registered in the toy, such as when an erroneous voice is input to the toy, so that a new voice is registered. In such a case, a troublesome operation such as pressing a button becomes unnecessary.

In step Pf1, a temporary registration process Pf
1 is started. In step Pf2, a process in which the toy emits sound is executed at the beginning of the temporary registration process Pf1. More specifically, audio data recorded in the ROM 2, for example, audio data such as "un?" The sound is output from the speaker 7 as sound. As described above, the predetermined sound is emitted from the toy in step Pf2, so that the user can confirm that the input sound is sound not registered in the toy.

At step Pf3, the waiting time parameter is reset, and at step Pa4, an interrupt for incrementing the waiting time parameter in response to a pulse signal S2 inputted from the clock operating mechanism 12 at predetermined time intervals, for example, every two seconds. Processing is started. Since the value of the waiting time parameter has been reset in step Pf3, the elapsed time from step Pf4 can be measured by referring to this value.

In step Pf5, the control circuit 1 determines whether or not there is a voice input from the microphone 4.
Specifically, when the level of the audio signal input from the microphone 4 is smaller than a predetermined threshold, it is determined that the audio signal is not input, and when the level is higher than the predetermined threshold, the audio signal is input. Is determined. If the control circuit 1 determines that no audio signal has been input, the processing shifts to step Pf6. In Step Pf6,
The elapsed time from step Pf4 indicated by the value of the waiting time parameter described above is compared with a predetermined time, for example, 10 seconds, and when the elapsed time is shorter than 10 seconds,
The process returns to step Pf5, and the presence or absence of the voice input from the microphone 4 is detected. If a voice is detected in Step Pf5 within 10 seconds, the interrupt process for incrementing the waiting time parameter is stopped, and the process shifts from Step Pf5 to Step Pf7. If no voice is detected in step Pf5 within 10 seconds, the interrupt process for incrementing the waiting time parameter is similarly stopped, and the process proceeds from step Pf6 to the first step of the above-described voice recognition process Pe. Will be migrated.

In step Pf7, the analog audio signal detected in step Pf5 is subjected to A / D conversion by the control circuit 1 and converted into digitized audio data. At this time, for example, when the volume of the voice is larger than the upper threshold set in the control circuit 1, or smaller than the lower threshold, the voice length is set to the upper limit. If it is longer than the threshold, step P
The process shifts to f8, and a sound is generated to notify that the level of the input sound is inappropriate. More specifically, for example, data of the sound “B” is read from the ROM 2, D / A-converted, and transmitted as a sound from the speaker 7 through the pitch conversion circuit 5 and the sound amplification circuit 6. Thereafter, the process returns to the first step of the above-described speech recognition processing Pe. If the audio data is normally acquired in Step Pf7, in the next Step Pf9,
Voice to notify the user that the voice data has been recognized,
For example, audio data "un?" Is read from the ROM 2 and D / A converted, and is emitted as sound from the speaker 7 via the pitch conversion circuit 5 and the audio amplification circuit 6.

In step Pf10, the above-mentioned waiting time parameter is reset again, and step Pf11
In, an interrupt process for incrementing the waiting time parameter is started according to the pulse signal S2 input from the clock operating mechanism 12 at predetermined time intervals. In step Pf12, step Pf17 described later
Is reset to zero.

In step Pf13, the control circuit 1 determines again whether or not there is a sound input from the microphone 4, and if it is determined that no sound signal has been input, the process proceeds to step Pf14. Step Pf14
Then, the elapsed time from step Pf11 indicated by the value of the waiting time parameter is compared with a predetermined time, for example, 10 seconds. If the elapsed time is shorter than 10 seconds, the process returns to step Pf13 again. , The presence or absence of a voice input from the microphone 4 is detected. If a voice is detected in Step Pf13 within 10 seconds, the interrupt process for incrementing the waiting time parameter is stopped, and the process shifts from Step Pf13 to Step Pf15. If no voice is detected in Step Pf13 within 10 seconds, the interrupt process for incrementing the waiting time parameter is similarly stopped, and the process returns to the first step of the voice recognition process Pe.

At Step Pf15, Step Pf13
The analog audio signal detected by the control circuit 1
It is subjected to D conversion and converted to digitized audio data. If it is determined that the voice input under the same conditions as in step Pf7 described above is inappropriate, the process proceeds to step Pf16.
The sound is transmitted from the ROM 2 and is subjected to D / A conversion, for example, sound data indicating that the level of the inputted sound is inappropriate, for example, sound data such as "B".
The sound is emitted from a speaker 7 through a pitch conversion circuit 5 and a sound amplification circuit 6. Next, in step Pf17, the value of the above-described integer k is incremented by one,
In step Pf18, the magnitude relationship between the integer k and the integer 1 is determined. Whether the integer k is equal to a predetermined integer, for example, 1;
If it is smaller than 1, the process returns to step Pf13 to wait for the input of a voice again, and if the integer k is larger than 1, the process returns to the first step of the voice recognition process Pe. As described above, since an opportunity is provided to re-input a predetermined number of times (in the above example, once) when the level of the audio is inappropriate, when an erroneous audio is input, It is not necessary to return to the initial step of the voice recognition processing Pe and repeat the input from the beginning, and the inconvenience of repeating the same operation is reduced.

If the voice data is normally obtained in step Pf15, the voice data obtained in step Pf19 is substituted for the variable Vd. Variable Vd
Is used to confirm a match with audio data input in a process described later.

Next, in step Pf20, a voice for notifying that the voice data has been recognized, for example, voice data such as "Woonmo" is read from the ROM 2 and D / A converted, and the pitch conversion circuit 5, the voice amplification circuit 6 , And is emitted as sound from the speaker 7. Since the sound emitted by the toy in response to the recognition of the sound changes (in the above example, “un?” Changes to “woonmo”), the user must confirm that the toy has recognized the sound and that the sound is being registered. You can know the stage.

In step Pf21, the above-mentioned waiting time parameter is reset again, and in step Pf22
In, an interrupt process for incrementing the waiting time parameter is started according to the pulse signal S2 input from the clock operating mechanism 12 at predetermined time intervals. In Step Pf23, Step Pf28 described later
Is reset to zero.

In step Pf24, the control circuit 1 determines again whether or not there is a voice input from the microphone 4, and if it is determined that no voice signal has been input, the process proceeds to step Pf25. Step Pf
In 25, the elapsed time from step Pf22 indicated by the value of the waiting time parameter and a predetermined time,
For example, the elapsed time is compared with 10 seconds, and if the elapsed time is shorter than 10 seconds, the process returns to Step Pf24 again.
The presence or absence of a voice input from the microphone 4 is detected. 10
If a voice is detected in step Pf24 within seconds, the interrupt processing for incrementing the waiting time parameter is stopped, and the processing shifts from step Pf24 to step Pf26. If no voice is detected in step Pf24 within 10 seconds, the interrupt process for incrementing the waiting time parameter is similarly stopped, and the process returns to the first step of the voice recognition process Pe.

In step Pf26, step Pf24
The analog audio signal detected by the control circuit 1
It is subjected to D conversion and converted to digitized audio data. If it is determined that the voice input under the same conditions as in step Pf7 described above is inappropriate, the process proceeds to step Pf27.
The sound is transmitted from the ROM 2 and is subjected to D / A conversion, for example, sound data indicating that the level of the inputted sound is inappropriate, for example, sound data such as "B".
The sound is emitted from a speaker 7 through a pitch conversion circuit 5 and a sound amplification circuit 6. Next, in step Pf28, the value of the above integer k is incremented by one,
In step Pf29, the magnitude relationship between the integer k and the integer 1 is determined. Whether the integer k is equal to a predetermined integer, for example, 1;
If it is smaller than 1, the process returns to step Pf24 and waits for the input of a voice again. If the integer k is larger than 1, the process returns to the first step of the voice recognition process Pe. Like steps Pf12 to Pf18,
Since the opportunity to re-enter the voice a predetermined number of times when the voice level is inappropriate is provided, the inconvenience of repeating the same operation is reduced.

In step Pf30, step Pf19
And the voice data substituted for the variable Vd at step Pf26
Are compared in the control circuit 1. The features of the audio data are extracted using, for example, the amplitude of a frequency spectrum obtained by performing Fourier transform on the audio data, and are compared by a predetermined comparison procedure. If it is determined that the characteristics of the audio data do not match each other, the process proceeds to step Pf27 described above, a voice is issued to notify that the level of the input voice is inappropriate, and the process goes through step Pf27. Depending on the number of times, the re-input of the voice is permitted, or the process is returned to the first step of the voice recognition process Pe.

If it is determined that the voice data of the variable Vd and the characteristics of the voice data obtained in step Pf26 match, the voice data obtained in step Pf26 is
Processing for recording in the LASH ROM 3 is performed. First, in step Pf31, the FLASH ROM 3
The registered word number related to the location where data is to be stored in the variable J is obtained and the number is substituted for the variable J.
FLASH whose registered word number is J in f32.
The sound data acquired in step Pf26 is stored in the storage area of the ROM 3. At this time, the temporary registration flag set in the temporary registration state is also recorded together with the information of the audio data. The temporary registration flag is a flag referred to in step Pe15 of FIG. The registration word for which the temporary registration flag is set is determined to be a temporary registration word in this step, and the execution of the conversation process Ph is not permitted, and the registration extension process Pg is executed. In the registration extension process, only the registered word that is temporarily registered by resetting the temporary registration flag is permitted to execute the conversation process Ph.

Subsequently, the operation of the registration extension process Pg will be described. FIGS. 13 and 14 are flowcharts for explaining the processing in the registration extension processing Pg. 13 and 14, Pg1 to Pg23 indicate steps. FIG. 13 and FIG. 14 show the connection relation of the steps.

As described above, when the registered word in which the temporary registration flag is set is recognized, the conversation processing Ph
Is not executed, the registration extension process Pg is executed. In the registration extension process Pg, the process of recognizing the voice is repeated a plurality of times until the temporary registration flag is reset and the registration word is formally registered, and the process of extending the formal registration is performed. The number of repetitions of the process of recognizing the voice is randomly determined within a range in which the above-described reliability parameter is added.

The conversation process Ph is a process in which a registration word and the above-described automatic voice registration process Pc are variously combined in accordance with the voice input to the microphone 4 by the user to produce a random voice. This is a process that is indispensable for enjoying the act of exchanging exchanges. Therefore, by prolonging the enjoyment of the conversation process Ph by the registration extension process Pg, it is possible to produce an effect of inspiring the user's interest and expectation for talking with the toy. In addition, if the registration of words is completed in the specified number of times, the user will have the impression that the toy mechanically records the sound, and the taste of the toy as a simulated creature will be impaired. By providing a process to extend the registration of words at random like this, it is possible to express the uncertainty when the creature memorizes the words, so that the user is stimulated as if the creature is remembering the words Can be.

The registration extension process Pg is started from step Pg1. First, in step Pg2, the magnitude of the above-mentioned reliability parameter is a predetermined integer value,
If the confidence parameter is smaller than the integer 45, the process proceeds to step Pg3. In step Pg3, for example, integers of 2 to 4 are randomly generated and assigned to the variable m. If the confidence parameter is larger than the integer 45, the process proceeds to Step Pg4. Also at step Pg4, the magnitude of the reliability parameter is compared with a predetermined integer value, for example, an integer 55, in magnitude relation, as in step Pg2. If the reliability parameter is smaller than the integer 55, the process proceeds to step Pg5. In Step Pg5, for example, integers of 1 to 3 are randomly generated and assigned to the variable m. The trust parameter is an integer 5
If it is larger than 5, the process proceeds to Step Pg6,
For example, an integer of 0 to 2 is randomly generated and assigned to a variable m. Through the above processing, an integer from 0 to 4 according to the magnitude of the reliability parameter is substituted for the variable m.

After an integer has been substituted for the variable m, step P
In g7, a match between the integer m and the integer 0 is confirmed. If the integer m is equal to the integer 0, the process proceeds to Step Pg21, where the provisional registration flag of the registered word having the registered word number J is reset and officially registered. On the other hand, when the integer m is not equal to the integer 0, steps Pg8 to Pg18
Is performed. In this step, in order for the operator to complete the formal registration, the operation of causing the toy to recognize the voice to be registered must be repeated a number of times according to the variable m.

In step Pg8, integer variables k and i to be referred to in the subsequent processing are reset to zero. Next, in step Pg9, the waiting time parameter is reset, and in step Pg10, the pulse signal S input from the clock operating mechanism 12 at predetermined time intervals is set.
In response to 2, the interrupt processing for incrementing the waiting time parameter is started. Since the value of the waiting time parameter has been reset in step Pg9, the elapsed time from step Pg10 can be measured by referring to this value.

In step Pg11, the control circuit 1 determines whether or not there is a voice input from the microphone 4. Specifically, when the level of the audio signal input from the microphone 4 is smaller than a predetermined threshold, it is determined that the audio signal is not input, and when the level is higher than the predetermined threshold, the audio signal is input. Is determined. If the control circuit 1 determines that no audio signal is input,
The process is shifted to Step Pg12. Step Pg1
In 2, the elapsed time from step Pg10 indicated by the value of the waiting time parameter is compared with a predetermined time, for example, 10 seconds. If the elapsed time is shorter than 10 seconds, the process returns to step Pg11 again. Thus, the presence or absence of the voice input from the microphone 4 is detected. If a voice is detected in step Pg11 within 10 seconds, the interrupt processing for incrementing the waiting time parameter is stopped, and the processing shifts from step Pg11 to step Pg13. If no sound is detected in step Pg11 within 10 seconds, the interruption process for incrementing the waiting time parameter is similarly stopped, and the process proceeds from step Pg12 to step Pg19 described later.

At step Pg13, at step Pg11
The analog audio signal detected by the control circuit 1
It is subjected to D conversion and converted to digitized audio data. At this time, for example, when the sound volume is larger than the upper limit threshold value determined in the control circuit 1,
If it is smaller than the lower threshold, or if the length of the voice is longer than the predetermined upper threshold, the process proceeds to step Pg14, and the level of the input voice is inappropriate. A voice is emitted to inform the user. More specifically, for example, data of the sound “B” is read from the ROM 2, D / A-converted, and transmitted as a sound from the speaker 7 through the pitch conversion circuit 5 and the sound amplification circuit 6. Thereafter, the process proceeds to Step Pg19 described later.

If the voice data is normally obtained in step Pg13, in the next step Pg15, a voice for notifying the user that the voice data has been recognized, for example, voice data "Wunnya" is stored in the ROM 2.
, And D / A-converted, and is emitted as sound from a speaker 7 via a pitch conversion circuit 5 and a sound amplification circuit 6. According to the above-described example, the sound uttered by the toy at various stages of registering the sound to the toy, such as “Un?” In step Pf2 and “Un?” In step Pf9 to “Woonmo” in step Pf20, and further, “Wunnya” in step Pg15, Since it changes, the user can roughly grasp the registration stage of the sound for the toy.

Next, at step Pg16, the integer 1 is added to the above-mentioned variable i, and at step Pg17, it is confirmed that the variables m and i match. If the variable m and the variable i do not match, the process returns to step Pg11, and the process of waiting for the input of a voice again and performing conversion to voice data is executed. Therefore, the number of times of repeating the processing of steps Pg11 to Pg17 is determined by the variable m randomly determined in consideration of the reliability parameters in steps Pg2 to Pg6.

In Step Pg17, the variable m and the variable i
If it is confirmed that they match, the step Pg
At 18, it is determined whether or not the features of the voice data whose registered word number is J and the voice data acquired in step Pg13 match. If they match, the process moves to step Pg21, where the temporary registration flag of the registered word whose registered word number is J is reset to FLASH.
It is stored at a predetermined address in the ROM 3. This allows
The audio data of the registration word J is officially registered. If the voice data of the registered word number J does not match the voice data acquired in step Pg13, the process proceeds to step Pg19 described later.

In steps Pg19 and Pg20, an error (step Pg12) in which the waiting time for the voice input exceeds a predetermined time or an error (step Pg1) in which the level of the input voice is too high.
3) or an error in which the characteristics of the audio data to be registered and the temporarily registered audio data do not match (step P
g18) occurs, the value of the variable k is set to the integer 1
Is added. Then, in step Pg20, the variable k
And the magnitude relationship of a predetermined integer, for example, the integer 3, is determined.
If the variable k is equal to or smaller than the integer 3, step Pg11
And the process of acquiring audio data is repeated. If the variable k is larger than the integer 3, the process returns to the beginning of the speech recognition process Pe. As described above, when the above-described error occurs, the process is not immediately returned to the beginning of the speech recognition process Pe. The process is returned to the process of acquiring audio. Therefore, for example, even when the voice is input with the wrong intonation, there is provided an opportunity to input the voice again, so that it is not necessary to re-input the voice from the beginning, and the same operation can be repeated. Inconvenience is reduced.

After the temporary registration flag is reset in step Pg21 and recorded in the FLASH ROM 3, in step Pg22, a sound expressing that the sound has been stored in the toy is emitted. For example, the audio data that has been officially registered in step P21 is D / A converted, and is output as a sound from the speaker 7 via the pitch conversion circuit 5 and the audio amplification circuit 6. This allows the user to be notified that the toy has memorized the sound.

According to the speech recognition processing Pe described above,
Until the temporary registration process is completed, voice input must be performed at least four times. Of these, in the voice input operation in step Pf5, the voice is tentatively registered by mistake and FL
This is a processing procedure for reducing the probability that the storage area of the ASH ROM 3 will be squeezed. Even if this processing is omitted, the operation of temporary registration is possible. The input of the voice in step Pf13 is a processing procedure for inputting a voice for confirming the match of the voice data in step Pf30. For example, as described above, when the process of step Pf5 is omitted, , If in step Pe8 the variable Vd
If the voice data substituted into the data can be referred to in step Pf30, the voice input operation in step Pf13 can also be omitted. That is, the processing from step Pf1 to step Pf20 in FIGS. 11 and 12 can be omitted.

As a modification of the temporary registration process, for example, after the process of Step Pf9, the processes of Step Pf3 to Step Pf8 are repeated, and finally, Steps Pf31 and Pf31 are executed.
It is also possible to perform f32. That is, the user repeatedly inputs the same voice, confirms the intention of the user to register the voice, and registers the voice. According to this, since matching of voices in each voice input is not confirmed, when different voices are continuously input, there is a possibility that the last input voice is registered, but the processing can be simplified. Therefore, the memory capacity can be saved.

Further, from step Pf19 to step Pf
Step 30 is repeated a plurality of times, and finally, Step Pf31,
It is also possible to perform Pf32. That is, before performing voice input, the previous voice data is stored in a variable Vd, and then the next voice data is obtained. After that, a match between the variable Vd and the obtained voice data is detected. This is the process of returning to the step. Since it is possible to check the case where the voice input last time and the voice input this time are different, the user can confirm the intention to register the voice in the toy also by this processing.

Further, from step Pf19 to step Pf
30 is repeated a plurality of times in the above-described modified example, the step Pf30 is performed by using the acquired audio data and the FLASH
The step may be changed to a step of searching for a match of the registered word in the ROM 3. In this case, if data matching the acquired audio data is stored in the FLASH ROM 3, the process proceeds to Step Pf27. According to this processing, since it is known that the user has input a voice that has already been registered, the user's intention can also be confirmed.

The processing in the registration extension processing is the same as the temporary registration processing as the processing for the purpose of allowing the user to input the same words and causing the toy to remember the words. It is also applicable to processing. Further, processing in which these are arbitrarily combined is also possible.

Next, the operation of the conversation processing Ph will be described.

In the conversation process Ph, a process of emitting various patterns of voice to the voice input by the user is performed. Further, a process of storing registered voices, which are alternately and continuously inputted, in association with each other, is also performed.

Although the speech processed in the conversation processing Ph is limited to the speech officially registered through the temporary registration, each speech registered officially is referred to as a memory parameter and a word relevance parameter. Two parameters are added. The memory parameter is officially registered and FLA
The parameter relating to the degree to which the storage of the voice stored in the SH ROM 3 is stored, and the word relevance parameter is a parameter relating to the degree to which the storage of the above-described information on the word relation is held.

The sound that can be stored in the FLASH ROM 3 is limited by the capacity of the memory. Therefore, when the number of stored voices increases and reaches a predetermined upper limit, new voices cannot be stored unless one of the stored voices is deleted. The storage parameter is a parameter for determining the voice to be deleted in such a case, and the parameter is deleted in order from the smallest data. Each time the voice is input and recognized, a value corresponding to the reliability value at the time of calling is added to the memory degree parameter. In addition, for example, it is slightly decreased every certain period such as one day. As a result, audio data that is not frequently input automatically decreases with time, and is replaced with new audio data and erased. Also, a predetermined lower limit value may be set in the memory parameter, and if the lower limit is exceeded, the voice data may be automatically deleted without being replaced with a new word. As described above, since a toy that forgets words reminds a creature, it is possible to enhance the interest of the user as a pseudo-creature held by the toy.

The number of words that can be stored in one voice data is also limited to a predetermined number, for example, three. For example, four words are added to voice data that currently has three words.
If you want to recognize the second different voice data as the opposite word,
In order to newly set this word, it is necessary to erase the memory of any of the words. The word relevance parameter is a parameter for determining a word to be deleted in such a case, and this parameter is deleted in order from the smallest data. The logarithmic relevance parameter is also referred to in a logarithmic continuous utterance function described later, and a parallax having a large logarithmic relevance parameter is selected and uttered.

The value corresponding to the confidence value is added to this word relevance parameter each time a word is uttered. For example, audio data: "Hello" for-word 1: "Good evening" for-word relevance parameter = 20 that the audio data and audio data: "Good evening" for-word 1: voice saying "Hello" for-word relevance parameter = 15 data is registered state, recognized this by entering the voice saying "Hello", the-word is when the voice of "evening" is uttered from toys, random in accordance with the confidence parameter to each-word relevance parameters 20 and 15 The numerical value is added.

The word relevance parameter is also slightly reduced at regular intervals, for example, one day, similarly to the memory parameter. As a result, the less-recognized word relevance parameter automatically decreases with time, and when a new word is learned, it is replaced and deleted.

FIG. 15 is a flowchart for explaining the processing in the conversation processing Ph. In FIG. 15, Ph1
To Ph13 indicate each step of the process.

The conversation process starts from step Ph1. In Step Ph2, a predetermined value corresponding to the reliability parameter is added to the storage parameter of the audio data of the registered word number J, and in Step Ph3, it is determined whether or not the registered word number J has an opposite word. First, the case where there is no opposite word will be described. In this case, the process shifts to step Ph12 to output the voice of the registered word number J. That is, when there is no voice to be used as a word, a process is performed in which a registered word that matches the voice input by the user is directly output as a parrot return. Next, the processing shifts to the later-described paring process Pi and is executed, and then the conversation process ends.

If it is determined in step Ph3 that the registered word number J has a pair word,
At h4, one of the plurality of steps is randomly selected with a predetermined probability, and the process proceeds to the selected step. For example, 60% to Step Ph5, Step Ph5
9% to 10%, Step Ph10 to 20%, Step Ph
The process is shifted to 11 with a probability of 10%.

When the process has proceeded to step Ph5, one of the three steps Ph6 to Ph8 for uttering the opposite word of the registered word J is randomly selected with a predetermined probability, and the process proceeds to the selected step. The process of shifting is performed. In this case, the probability of the transition is, for example, a magnitude corresponding to the above-described word relevance parameter of each word. That is, assuming that the logarithmic relevance parameters are r1, r2, and r3 for the log1, log2, and log3, respectively, and the probabilities of shifting the processing to the step of uttering each log are Q1, Q2, and Q3, respectively. The ratio has the following relationship: Q1: Q2: Q3 = r1: r2: r3 Therefore, for example, when the word relevance parameters are 25, 10, and 15 for the words 1, 2, and 3, respectively, steps Ph4 to Ph6, Ph7, The probabilities of shifting to Step Ph8 are 30%, 12%, and 18%, respectively. Note that a random numerical value according to the confidence parameter at the time of the utterance is added to the utterance degree parameter of the selected utterance. In addition, a random numerical value according to the confidence parameter is added to each of the paired word relevance parameter and the memory degree parameter of each selected and uttered word. As described above, not only can a plurality of utterances be uttered for one input voice, but also the probability of uttering the utterance is changed by giving a weight according to the recognition frequency of the utterance. The meaning is given to the act of recognizing the toy by repeating the predetermined word, and the toy can be made tired.

In step Ph4, when step Ph9 is selected and the process proceeds, one of all registered voice data is selected at random and uttered. Since the voice that is not related to the input voice is emitted, the user is surprised and is interested in talking with the toy.

In step Ph4, step Ph10
Is selected, and the process shifts, the voice data of the registered word number J is uttered, and the voice data having the highest degree of parallax relevance of the pair words of the registered word number J is uttered. Further, a word with the largest word relevance parameter of the voice data is uttered, and by repeating this, words having a maximum of, for example, 10 words are continuously uttered (continuous utterance function of word). However, processing is performed so that the same sound is not repeated. In addition, a random numerical value according to the reliability parameter is added to the logarithmic relevance parameter and the memory parameter of each of the consecutively uttered logistics. By continuously uttering the words in this way, it gives the impression that the toy is singing a song and gives the user an interest. After the continuous utterance processing of the paired words is completed, the processing shifts to the processing Pi described later and is executed, and then the conversation processing is ended.

In step Ph4, step Ph10
Is selected and the process shifts, the automatic voice registration process Pc
Is selected and uttered. Since the voice registered in the automatic voice registration process Pc is a voice that is not intended by the user at all, it gives the user an unexpected feeling that is not present in other voice generation processes, and makes the toy tireless.

Subsequently, the operation of the word processing Pi will be described. The purpose of the word processing Pi is to detect the number of times two voices are successively and alternately input, and when the number reaches a number randomly selected according to the reliability parameter, the number of times is calculated.
To associate two voices as a pair. FIG.
FIG. 7 is a flowchart for explaining processing in the word processing Pi. In FIG. 16, Pi1 to Pi22 indicate each step of the processing.

The word processing Pi starts in step Pi1. Next, in steps Pi2 and Pi3, it is determined whether the registered word number J input and recognized from the microphone 4 matches the registered word numbers stored in the variables BOX1 and BOX2. Here, it is to be noted that each variable (BO) used in the word processing Pi is described.
X1, BOX2, CNT1, CNT2, i, and p) are all reset at the beginning of the voice recognition processing Pe, and are held until the voice recognition processing Pe is completed or reset in steps Pi8 and Pi18 described later. You.

Here, the variable BOX1 and the variable BOX2
Holds registered word numbers of two voices when they are alternately and continuously input. Variable BOX1 and variable B
The registered word number held in OX2 is the registered word number J
If it is determined that they do not match, the step Pi4
In step Pi5, a match between the variables BOX1 and BOX2 and zero is detected. In this description, it is assumed that there is no voice storage area having a registered word number of zero. In step Pi4 and step Pi5, it is detected whether or not the voice that has been input alternately and continuously is in the middle of being recognized as the opposite word.
If it is in the process of being recognized as a word, the variable BOX1
And one of the variables BOX2 becomes zero.

If neither the variable BOX1 nor the variable BOX2 is zero, the flow shifts to step Pi6. In this case, it is considered that the user did not recognize the opposite word because the voices input two times before, the previous time, and the current time are all different. Steps Pi6 and Pi7 avoid the resetting of the variables in step Pi8 in order to give a delay. Since the variable i is incremented when passing through the step Pi6, the variable i is incremented to be greater than 1 again after passing through the step Pi6 again, and the variable is reset at the step Pi8 via the step Pi7. In step Pi8, each of the above variables is reset to zero.

In step Pi4 and step Pi5, if either the variable BOX1 or the variable BOX2 is zero, the registered word number J of the currently input voice is substituted for the zero variable, The variable CNT1 or the variable CNT2 that counts the number of input voices is incremented. Then step Pi
At 11, it is determined whether each of the variables BOX1 and BOX2 has been input once. That is, it is confirmed that two voices are alternately input as a pair each time. If either the variable CNT1 or the variable CNT2 is zero, the word processing ends.

When the first input of the opposite word is confirmed in step Pi11, it is next determined in step Pi12 whether or not the two input voices have already been registered as opposite words. If it is registered as a log word, a random numerical value according to the reliability parameter is added to each log word relevance parameter of the two voice data whose registration word numbers are indicated by the variables BOX1 and BOX2 in step Pi14. When the two voice data are not the opposite words, a random numerical value according to the reliability parameter is given to the variable p. The variable p indicates the number of times the user has to enter the word before the word is registered. In other words, it indicates the number of times that the operation of alternately inputting two voices once is repeated.

In Steps Pi2 and Pi3, when it is confirmed that the voice input from the microphone 4 matches the voice input to either the variable BOX1 or the variable BOX2, BO
Variable CNT1 or variable CNT2 corresponding to X1 or variable BOX2 is incremented (step Pi1).
5 or step Pi16). Then, step Pi1
At 7, the difference between the variable CNT1 and the variable CNT2 is detected, and if the difference between the values is a predetermined number, for example, 3 or more, the variable is reset at Pi18 and the conversation processing is terminated. That is, a state in which only one of the voices is continuously input is detected. If the state is three or more times, the user recognizes that the user has not learned the language, and resets the variables related to the language learning. I have.

If the difference between the variables CNT1 and CNT2 is less than 3, it is determined in step Pi19 whether the above-described variable p has been set. If it is not set, it is determined that the first recognition of the opposite word has not yet been performed, and the opposite word processing is terminated. If the variable p is set, it is determined in step Pi20 whether one of the variable CNT1 and the variable CNT2 is larger than the variable p. If smaller, the word processing is performed as it is to repeat the word recognition. To end,
If it is larger, it is determined that the number of times that the word has been recognized has reached the predetermined number, and in step Pi21, two voice data whose registration word numbers are indicated by the variables BOX1 and BOX2 are registered as words.

As described above, in the conversation processing Ph, the voices stored in various variations are output, so that unexpectedness is given each time and the enjoyment of the user is maintained. In addition, since it is possible to recognize the opposite word by repeatedly speaking the same voice as when letting the parakeet remember the words, the mechanical impression of the toy is weakened, and the feeling as if you are dealing with a real creature Let the user wake up. Furthermore, since the number of times for recognizing the opposite word is randomly generated, the authenticity is not spoiled in such aspects. In this way, a highly interesting toy as a pseudo-creature can be provided.

The embodiment of the present invention is not limited to the electronic pet toy as shown in the external view of FIG. 1, but can be widely applied to, for example, a computer or a game machine having an audio output device. Can be. Further, by expressing the flowchart in each drawing used in the description of the present embodiment of the present invention as a program, the present invention can be applied to a recording medium on which the program is recorded.

[0169]

According to the present invention, the stored voice can be randomly uttered in various patterns, such as the voice automatically stored and the voice registered by the user. And can make the device less tiring. In addition, various operation states such as an operation state of automatically capturing sound, an operation state of randomly uttering sound, and an operation state of recognition and generation of sound are randomly selected and executed, so that a mechanical impression is obtained. Is reduced, and it becomes easier to obtain a sense of being opposed to a creature.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating an appearance of a toy according to an embodiment of a voice processing device of the present invention.

FIG. 2 is a diagram illustrating a configuration of a toy according to an embodiment of the voice processing device of the present invention.

FIG. 3 is a diagram showing an operation mode related to an operation at the time of initial activation of the toy according to the embodiment of the present invention.

FIG. 4 is a diagram showing details of an operation mode in a normal mode.

FIG. 5 is an overhead view showing an arrangement of toys in a communication mode.

FIG. 6 is a diagram illustrating an operation mode according to a reliability parameter;

FIG. 7 is a first flowchart illustrating a process in an initial startup mode;

FIG. 8 is a second flowchart illustrating the processing in the first activation mode.

FIG. 9 is a flowchart illustrating a process in a wake-up mode;

FIG. 10 is a flowchart illustrating processing in voice recognition processing.

FIG. 11 is a first flowchart illustrating a process in a temporary registration process;

FIG. 12 is a second flowchart illustrating a process in a temporary registration process;

FIG. 13 is a first flowchart illustrating a process in a registration extension process;

FIG. 14 is a second flowchart illustrating the processing in the registration extension processing;

FIG. 15 is a flowchart illustrating a process in a conversation process Ph.

FIG. 16 is a flowchart illustrating a process in the word processing Pi.

FIG. 17 is a flowchart illustrating a process in a random utterance process Pb.

FIG. 18 is a diagram illustrating an example of an utterance that changes according to a value of a mood parameter.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Control circuit (control means), 2 ... ROM, 3 ... FLAS
H ROM (storage means), 4 ... microphone (sound input means),
5: pitch conversion circuit, 6: audio amplification circuit, 7: speaker (audio output means), 8: head switch, 9: communication switch, 10: alarm switch, 11: reset switch, 12: clock operation mechanism.

──────────────────────────────────────────────────続 き Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat ゛ (Reference) G10L 13/04 G10L 3/00 521M 551H 5/02 J (72) Inventor Manabu Hoshino 1 Kinshi, Sumida-ku, Tokyo Chome No.2-1 Rhythm Watch Industry Co., Ltd. F-term (reference) 2C150 BA06 BA11 CA01 CA02 DF02 DF04 DF06 DF33 ED42 ED52 EF16 EF23 EF29 FA42 5D015 AA03 BB01 GG04 HH16 KK01 KK04 LL03 LL10 5D045 BB11 CA5CA GG03 HH15 HH17 JJ76 KK09 KK45

Claims (40)

[Claims] 1. A sound output means, a storage means for storing a sound, and in a first operating state, at least one of the sounds stored in the storage means is based on a signal generated at random. Control means for selecting and outputting to the sound output means. 2. The control means includes a random number generation means,
2. The method according to claim 1, wherein in the first operation state, at least one of the sounds stored in the storage means is selected based on a signal generated by the random number generation means, and is output to the sound output means. An audio processing device according to claim 1. 3. The random number generating means, in the first operating state, a predetermined number of voices stored in the storage means based on a random number generated by the random number generating means. At least one voice selecting means for randomly selecting and outputting to the voice output means in a predetermined order. In the first operating state, one of the voice selecting means is randomly selected based on the random number. The audio processing device according to claim 1. 4. The control means sets an association between the two sounds stored in the storage means, stores the association in the storage means, and the audio selection means stores the association stored in the storage means. 4. A voice selecting unit according to claim 3, further comprising a voice selecting unit that randomly selects a predetermined number of the voices based on the random number, and causes the voice output unit to output voices set in the selected association in a predetermined order. Voice processing device. 5. The voice selecting means determines one of the voices stored as a name in the storage means and a predetermined number of voices among other voices stored in the storage means based on the random number. 5. The voice processing device according to claim 3, further comprising a voice selection unit that randomly selects and outputs the voice stored in the name as the first voice to the voice output unit in a predetermined order. 6. A sound input means, wherein the control means stores a sound input from the sound input means in the storage means for a predetermined time in a second operation state; 4. A voice selecting means for randomly selecting a predetermined number of sounds among the sounds stored in the second operating state based on the random numbers and outputting the selected sounds to the voice output means in a predetermined order. 6. The voice processing device according to any one of 5. 7. In the second operating state, the control means stores the sound input from the sound input means in the storage means for a predetermined time at a predetermined time interval based on the random number. The voice processing device according to claim 6, wherein the voice processing is executed at random times. 8. The sound according to claim 3, wherein the control means randomly sets the number of times the sound selected by the sound selection means is output to the sound output means based on the random number. Processing equipment. 9. A sound input unit, wherein the control unit recognizes a sound based on a sound input from the sound input unit in the third operation state, and stores the sound and the sound stored in the storage unit. 9. The method according to claim 1, further comprising: comparing the recognized voice, storing the recognized voice in the storage unit based on a result of the verification, and outputting the voice stored in the storage unit to the voice output unit. An audio processing device according to any one of the above. 10. The audio processing apparatus according to claim 6, wherein said control means randomly selects a plurality of operation states based on said random numbers. 11. A speech processing apparatus for collating a predetermined feature between a recognized speech and a stored speech, wherein the control means performs the collation performed before and after in the third operation state. If the recognized voice and the stored voice share the predetermined feature, and the stored voice is different before and after, the voice sharing the predetermined feature with the stored voice before and after 11. The audio processing apparatus according to claim 10, further comprising the step of storing the association between the stored audio data before and after the audio data stored in the storage unit on condition that the audio data is recognized m times (m: a predetermined natural number) alternately within a predetermined time. 12. A voice output means for outputting a voice, wherein the control means associates one of the stored voices sharing the predetermined feature with the voice recognized in the collation. When selecting and causing the output means to output the voice indicated by the association, the parameter is a parameter indicating the degree of association between the stored voice and the voice indicated by the association, and 12. The speech processing apparatus according to claim 11, wherein the speech relevance parameters attached to the speech and the speech indicated in the association are respectively increased and stored in the storage unit. 13. The controller according to claim 12, wherein the controller includes a random number generator for generating a random number, and randomly changes the increment of the parameter of the opposite word relevance parameter and the natural number m based on the random number generated by the random number generator. An audio processing device according to claim 1. 14. The audio processing apparatus according to claim 11, wherein said control means decreases said storage parameter by a predetermined amount every time a predetermined time elapses. 15. The control means, wherein the voice recognized in the collation and the stored voice share the predetermined feature, and the association between the stored voice and another stored voice is determined by the association. If stored in the storage means,
15. The audio processing device according to claim 11, wherein said other stored audio is output to said audio output means. 16. The control means, wherein the voice recognized in the collation and the stored voice share the predetermined feature, and the association between the stored voice and another stored voice is determined by the control unit. If stored in the storage means,
Randomly selecting at least one of the stored voice and the other stored voice with a predetermined probability;
15. The audio processing device according to claim 11, wherein the selected audio is output to the audio output unit in a predetermined order. 17. The speech processing apparatus according to claim 16, wherein said control means sets said predetermined probability in accordance with the magnitude of said word relevance parameter. 18. The control means, when the association of the sound output by the sound output means is stored in the storage means, selects one of the sounds associated with the sound, and 17. Outputting the sound to an audio output means.
Or the speech processing device according to 17. 19. The control means, if the association of the sound output by the sound output means is stored in the storage means, the sound having the largest logarithmic relevance parameter among the sounds associated with the sound. 19. The method according to claim 16, further comprising the step of selecting one of the following to output to the audio output means.
The audio processing device according to any one of the above. 20. The apparatus according to claim 16, wherein the control means stops the sound output from the sound output means when the number of sounds continuously output to the sound output means reaches a predetermined number. An audio processing device according to any one of the above. 21. The control device according to claim 1, wherein the predetermined probability includes a case where at least one voice is randomly selected from all voices stored in the storage device.
21. The audio processing device according to any one of 6 to 20. 22. The storage means is capable of storing a predetermined number of voices, and the control means is configured such that the voice recognized in the collation and the stored voice share the predetermined feature. In this case, the storage degree parameter related to the storage holding of the stored voice is increased, and the stored voice and the recognized voice are replaced in accordance with the magnitude of the storage degree parameter and stored in the storage unit. Claim 11
22. The audio processing device according to any one of claims 21 to 21. 23. The audio processing apparatus according to claim 22, wherein said control means decreases said storage degree parameter by a predetermined amount each time a predetermined time elapses. 24. When the number of voices stored in the storage unit has reached the predetermined number, the control unit determines that the voice with the smallest storage degree parameter among the voices stored in the storage unit. 24. The voice processing device according to claim 23, wherein the recognized voice is replaced and stored in the storage unit. 25. The audio processing apparatus according to claim 23, wherein the control means randomly changes the increment of the storage degree parameter based on a random number generated by the random number generation means. 26. In the third state, the control unit collates a voice recognized based on the sound input from the sound input unit with a predetermined characteristic of the voice stored in the storage unit, 26. The sound processing apparatus according to claim 9, wherein the sound is stored or registered in the storage unit based on a result of the comparison, and a registered sound of the stored sound is output to the sound output unit. Voice processing device. 27. The control means, if the recognized voice is stored in the storage means and registered voice does not share the predetermined feature, the recognized voice 27. The sound processing device according to claim 26, wherein the sound processing device is stored in the storage unit. 28. The control unit performs the recognition of the voice a predetermined number of times when the recognized voice and the voice stored in the storage unit do not share a predetermined feature. 28. The speech processing device according to claim 26, wherein the recognized speech is stored in the storage means on a condition that the speech is recognized. 29. The control means, if the recognized voice and the registered voice stored in the storage means do not share a predetermined characteristic, and the recognized voice and the registered voice are not shared. A predetermined feature of the voice is compared a predetermined number of times, and the recognized voice is not recognized on the condition that the recognized voice and the registered voice do not share a predetermined feature. 29. The sound processing device according to claim 26, wherein the sound processing device is stored in the storage unit. 30. The control means, if the recognized voice and the registered voice stored in the storage means do not share a predetermined characteristic, are continuously recognized within a predetermined time. The predetermined characteristics of the two voices are compared a predetermined number of times, and the recognized voices are stored in the storage unit on condition that the voices recognized in the verification share the predetermined characteristics with each other. Claims 26 to 2 for storing
The audio processing device according to any one of claims 9 to 10. 31. When the recognized voice and the unregistered voice stored in the storage unit share a predetermined feature, the control unit converts the unregistered voice into the unregistered voice. 31. The voice processing device according to claim 26, wherein the voice processing device is registered in a storage unit. 32. The control means includes a random number generation means for generating a natural number random number n, wherein the recognized voice and the voice stored in the storage means and not registered share a predetermined characteristic. 32. The speech processing device according to claim 26, wherein, when the recognition is performed, the unregistered speech is registered in the storage unit on condition that the recognition of the speech is performed n times. 33. The control means includes a random number generation means for generating a natural number random number n, wherein the recognized voice and the voice stored in the storage means and not registered share a predetermined characteristic. In this case, predetermined characteristics of the recognized voice and the unregistered voice are compared n times (n: a predetermined natural number), and the recognized voice and the unregistered voice are not registered in the verification. 33. The audio processing device according to claim 26, wherein the unregistered audio is registered in the storage unit on condition that the audio and the audio share a predetermined feature. 34. The control means includes random number generation means for generating a natural number random number n, wherein the recognized voice and the voice stored in the storage means and not registered share a predetermined characteristic. , A predetermined feature of voices continuously recognized within a predetermined time is collated a predetermined number of times, and in the collation, the recognized voices share a predetermined feature with each other. 34, the unregistered voice is registered in the storage means.
The audio processing device according to any one of the above. 35. A trust value storage means for holding a trust value as a parameter indicating a device condition which affects the ease of storage, wherein the control means responds to the trust value held in the trust value storage means. 35. The voice processing device according to claim 26, wherein the range of the random number generated by said random number generation means is regulated. 36. The control unit according to claim 36, wherein the control unit transmits the recognized voice or a voice obtained by performing a predetermined process on the recognized voice to notify registration of the recognized voice in the storage unit. The audio processing device according to any one of claims 26 to 35, wherein the audio processing device outputs the sound from an output unit. 37. A notification sound storage means for storing a predetermined sound, and the sound output means outputs the sound stored in the notification sound storage means in accordance with the result of the collation. The audio processing device according to any one of the above. 38. The control device according to claim 26, wherein in the third operation state, the sound is recognized as a sound on condition that a level of the sound input from the sound input device is within a predetermined range. 37. The audio processing device according to any one of 37. 39. The control means, on the condition that the level of the sound input from the sound input means is in a predetermined range in the third operation state and the level in the predetermined range is maintained for a predetermined time. The voice processing device according to any one of claims 26 to 38, wherein the sound is recognized as voice. 40. The voice processing apparatus according to claim 26, wherein the control unit shifts the operation state to another operation state including the voice recognition state when the voice is not recognized within a predetermined time. .