JP4224991B2 - Network speech recognition system, network speech recognition method, network speech recognition program - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a network technology for operating a plurality of electric devices by voice recognition, and centrally processing voices of operators commanded to a plurality of devices connected to a line by a central processing unit. The present invention relates to a network voice recognition system, a network voice recognition method, and a network voice recognition program for transmitting a predetermined operation output to an electric device.
[0002]
[Prior art]
Japanese Unexamined Patent Publication No. 56-88503 is available as a technology for handling voice recognition. This is because the heating device is equipped with a voice recognition unit and a voice synthesis unit, and the two voice recognition modes and one voice synthesis mode are switched sequentially, so that the operation proceeds as if the operator and the device interact. This is intended to prevent malfunction due to misrecognition.
[0003]
The input speech is extracted and stored in the memory. In the recognition processing, a standard pattern stored in advance in the ROM is compared with the pattern stored in the memory, and a pattern having a high degree of similarity is recognized as the voice command. Next, the mode is switched to the voice synthesis mode, and a predetermined voice message is issued to the operator based on the voice recognized by the voice synthesis synthesizer. When the operator listens to this message and inputs the next voice, the device again shifts to the second voice recognition mode, and determines whether a predetermined voice is input.
[0004]
With this configuration, the system can surely proceed with the control. It is possible to prevent malfunctions such as starting heating that is not instructed by misrecognition or interrupting heating that has not been completed.
[0005]
This is an application by the inventor of the present invention, but at that time, the technical background at the time when it was difficult to achieve perfect recognition even with specific speaker speech recognition (a method of registering the language of the target person to be recognized in advance) is disclosed. It is an invention.
[0006]
In recent years, with the remarkable progress of semiconductor technology and the development of software technology and network technology, information devices such as personal computers and mobile phones have spread widely in the home. In addition, electric appliances called internet appliances are gradually appearing on the market.
[0007]
Under such circumstances, speech recognition technology, which has been attracting attention for some time, is now entering the practical stage as a technology that enables good communication between humans and machines. In particular, it has become cheaper, CPU processing speed has increased dramatically, and speech recognition technology has evolved, so unspecified speaker speech recognition, which has become extremely difficult 10 years ago (no need for language registration in advance) Can be recognized at a high recognition rate. It can be said that a computer equipped with a microphone instead of a keyboard has entered a practical level. Computers with such voice recognition software pre-installed at the time of factory shipment are not uncommon.
[0008]
However, a heating device as disclosed in JP-A-56-88503 has not yet been commercialized. This is despite the rapid spread of speech recognition technology in the world of personal computers. This comes primarily from the economics of speech recognition systems. With a personal computer, both the fast CPU and large memory increase the value of the personal computer itself. With a personal computer, you can introduce voice recognition without increasing the price. Moreover, even if some of them are misrecognized, they may be re-input. It is equivalent to simply mistouching the keyboard. However, in a heating device, misrecognition can lead to excessive heating and combustion in the cabinet. When a speech recognition system is incorporated in the heating device, the recognition accuracy is more important, and if this is increased, the system cost will be further increased.
[0009]
As a related technique of the present invention, Japanese Patent Laid-Open No. 2000-111057. issue The gazette is shown. This discloses a data processing device for kitchen appliances. A kitchen appliance, for example, a microwave oven, is provided with a microphone that generates digital data in response to an operator's voice and voice recognition means for extracting a digital signal from the microphone. Input to the microphone are food items consumed in the cabinet, and the main body is provided with storage means for storing digital signals to generate a list of processed items. This digital data is also transferred to the remote management system.
[0010]
With this configuration, when a food item consumed in the warehouse is input to the microphone by voice, the voice recognition software loaded on the appliance body identifies it and generates an e-mail message that is forwarded to the Internet. used. Messages transferred to the Internet are used to replenish and deliver items.
[0011]
Even in this configuration, the kitchen appliance is provided with voice recognition means, and it cannot be denied that the cost of the kitchen appliance has been increased. Moreover, what is handled by voice recognition is only information on the name of a food item, and even if it is misrecognized, an order for a supplemental item is not made accurately. It does not include a function for operating / stopping the device main body disclosed in JP 2000-1111057 A.
[0012]
[Problems to be solved by the invention]
In the present invention, a plurality of electrical devices connected to a network can be operated by a voice-friendly operation system called voice recognition, always using the latest voice recognition technology, and maintaining a high recognition rate while economically. Aiming to realize a usable system.
[0013]
[Means for Solving the Problems]
In order to solve the conventional problem, the network speech recognition system of the present invention converts speech into an electrical signal. Microphone And an electric device having an individual ID with an individual ID and connected to the electric device via a line. Built-in voice recognition means Central processing means, The operator of the electrical device inputs an operation command from the microphone, The electrical device sends a voice signal and the individual ID to the central processing unit, and the central processing unit performs voice recognition processing of the transmitted voice signal and identifies the electrical device that has sent the voice signal. The voice list registered for each electrical device is searched from the audio signal to determine whether the operation command is accepted by the electrical device. If the operation command is accepted, the electrical device is confirmed and accepted. If it is checked and accepted Based on processing results Predetermined by the electrical device and the central processing means The predetermined operation output For the electrical equipment It is configured to transmit. As a result, a voice recognition means that realizes a high recognition rate, which is costly, can be placed in the central processing means, and it recognizes voice data transmitted from electrical devices connected by a line one after another and operates based on the recognition result A signal can be returned.
[0014]
DETAILED DESCRIPTION OF THE INVENTION
The present invention converts sound into an electrical signal. Microphone And an electrical device with an individual ID and connected to this device via a line Built-in voice recognition means Central processing means, The operator of the electrical device inputs an operation command from the microphone, The electrical device sends a voice signal and the individual ID to the central processing unit, and the central processing unit performs voice recognition processing of the transmitted voice signal and identifies the electrical device that has sent the voice signal. The voice list registered for each electrical device is searched from the audio signal to determine whether the operation command is accepted by the electrical device. If the operation command is accepted, the electrical device is confirmed and accepted. If it is checked and accepted Based on processing results Predetermined by the electrical device and the central processing means The predetermined operation output For the electrical equipment A system configured to transmit. As a result, the customer can economically realize easy operation by voice recognition.
[0015]
Also The apparatus includes a plurality of electric devices, and each device does not need to have voice recognition means, and can use a simple operation by voice recognition more economically. All devices can be operated with the same operation system.
[0016]
Also In addition, the electrical equipment further has an A / D converter, which is configured to convert voice into a digital signal and transmit it to the central processing means via a line. It becomes easy to shorten.
[0017]
Also The speech recognition process performed by the central processing means is a speech recognition process based on linear predictive analysis (LPC), vector quantization (VQ), and hidden Markov model (HMM), and features are extracted by linear predictive analysis (LPC). The obtained feature vector can be converted into a finite number of symbols by vector quantization (VQ). By using this, a hidden Markov model (HMM), that is, a phonemic probability model (probability automaton) can be created to recognize a phoneme unit.
[0018]
Also The central processing unit is configured to synthesize a predetermined voice from the memory and transmit it via a line when it is necessary to convey a message to the electric equipment based on the voice recognition processing result, and control the equipment while having a conversation. Can proceed.
[0019]
Also The central processing means is configured to have a hardware for performing speech recognition and a program for performing software processing. Instead of performing software processing for all, the hardware and software share the speech recognition processing in a well-balanced manner. A system can be realized. In particular, the burden on the central processing unit can be reduced.
[0020]
Also , Electric The electrical device converts the sound into an electrical signal; and And individual ID and And the central processing means was sent Identify the electrical device by individual ID Performing voice recognition processing of the voice signal; The voice list registered for each electrical device is searched from the audio signal to determine whether the operation command is accepted by the electrical device. When the operation command is accepted, the electrical device is confirmed and is now accepted If it is checked and accepted Based on processing results Predetermined by the electrical device and the central processing means The predetermined operation output For the electrical equipment A network recognition method comprising a transmitting step and a network recognition method. As a result, the customer can economically realize easy operation by voice recognition.
[0021]
A step of converting the sound into an electric signal by the electric device; And individual ID and And the central processing means has been sent Identify the electrical device by individual ID Performing voice recognition processing of the voice signal; The voice list registered for each electrical device is searched from the audio signal to determine whether the operation command is accepted by the electrical device. When the operation command is accepted, the electrical device is confirmed and is now accepted If it is checked and accepted Based on processing results Predetermined by the electrical device and the central processing means The predetermined operation output For the electrical equipment A network speech recognition program comprising a transmitting step. As a result, the customer can economically realize easy operation by voice recognition.
[0022]
【Example】
Embodiments of the present invention will be described below with reference to the drawings.
[0023]
Example 1
FIG. 2 is a connection diagram showing the configuration of the network speech recognition system in the first embodiment of the present invention. A group of electric devices installed in a home includes a microphone 1 that is a means for converting sound into an electric signal, a speaker 2 that restores a synthesized sound signal to sound, and a connection means 3 that is connected to a network. ing.
[0024]
Therefore, for example, the operation unit of the laundry 4 has only the microphone 1, the speaker 2, and the display window, and there is no timer or keyboard as in the conventional case. The connection means 3 is built in the device body.
[0025]
The microwave oven 5, the rice cooker 6, and the refrigerator 7 are exactly the same, and the operation unit has a microphone 1, a speaker 2, and a display window as necessary. The TV 8 is the same, but the TV originally has the speaker 2 as an original function.
[0026]
There is a central processing means 9 connected to these devices via a connection means 3 via a line. This controls not only a home network but also a large number of home electrical devices via lines. The voice recognition means 10 is incorporated.
[0027]
FIG. 1 is a block diagram showing the configuration of a network speech recognition system according to the first embodiment of the present invention. The laundry 4 is provided with a control means 11 in addition to the microphone 1, the speaker 2 and the connection means 3 described above, and controls the motor and the heater 12 based on a voice command. That is, the control means 11 transmits the operation command input from the microphone 1 to the central processing means 9 via the connection means 3 and the line. This voice data is connected to the central processing means 9 via the connection means 14 and the ID recognition means 15 identifies which address (electrical device) at which location (home) and recognizes that it is a home laundry called A. After that, the command content is analyzed by the voice recognition means 10. In the speech recognition, the similarity with the speech pattern recorded in the storage means 16 is compared, or the sentence is estimated from the preceding and following words or phonemes. As a result of the recognition process, what operation the device should take is transmitted to the laundry 4 again. In response to this output, the control means 11 starts, changes or stops energization of the motor and the heater 12. The control means 11 outputs the synthesized voice message transmitted from the speaker 2 as necessary.
[0028]
The same applies to the microwave oven 5 or lower. In the microwave oven 5, the magnetron 13 that is a heat source is controlled by the control means 11. Of course, whether it is the laundry 4 or the microwave oven 5, the controlled object is not the only one. These are the controlled main blocks. The data output from the central processing means 9 may be data for directly controlling a predetermined controlled block, or data obtained by coding a certain operation mode. These are once decoded by the control means 11 to control the controlled block.
[0029]
With this configuration, the network speech recognition system allows the operator and the device to interact with each other through the network like a heating device having a speech recognition unit and a speech synthesis unit as disclosed in JP-A-56-88503. The operation can be performed as if it were, and the operation of the device can be executed reliably. Of course, speech synthesis is not essential for the present invention.
[0030]
(Example 2)
FIG. 3 is a connection diagram showing the configuration of the network voice recognition system in the second embodiment of the present invention. Each home, such as a home A17 and a home B18, is connected to the central processing means 9 via the Internet. As in FIG. 2, an electrical device group installed in a home A <b> 17 includes a microphone that is a means for converting sound into an electrical signal and a speaker that restores a synthesized speech signal to speech. However, the connection means for connecting to the network does not have individual electric devices. These devices are connected to a hub switch 19, and the hub switch 19 is connected to a router. That is, a single router is shared as a connection means with the central processing means 9. The home B18 has the same configuration.
[0031]
With this configuration, the present invention can be utilized economically by sharing an expensive modem at present. However, in preparation for the realization of future home appliances, the development of an inexpensive OS that simplifies the communication protocol and executes the connection to the Internet that can be installed in an 8-bit microcomputer is being promoted. A nasty modem will be a laugh after a few years. In short, the configuration shown in FIG. 3 is realistic for the time being, but the configuration shown in FIG. 1 will become common in the future.
[0032]
This selection is a comparison between the connection cost to the network (hardware and software such as a usage fee) and the hardware cost of the voice recognition means. If the former exceeds the latter, the present invention actually has no users. However, as will be described later, the speech recognition processing has high-speed arithmetic processing, a large amount of storage means, and technological innovations of daily progress, and according to the present invention, these can be shared by the central processing means, while updating to the latest technology, Each electric device may be provided with a very simple control means for controlling the operation of a heat source, a motor and the like. In some cases, all control can be performed by the central processing means. The premise of the present invention is that the network cost is steadily decreasing and the voice processing technology is evolving.
[0033]
(Example 3)
FIG. 4 is a block diagram showing the configuration of the network speech recognition system according to the third embodiment of the present invention. In contrast to the first embodiment shown in FIG. 1, an A / D converter 21 is connected to the microphone 1, and voice is converted into a digital signal. Send to. An audio signal transmitted by such a configuration is strong against noise. That is, if transmission data is divided into several frames and an even or odd parity check is performed for each frame, it becomes easy to exclude noise. If a digital signal is used, processing such as compression is easy. An A / D converter is inexpensive, and a configuration in which the A / D converter is individually mounted on each electric device is beneficial in terms of practical use in order to increase reliability.
[0034]
Further, in this embodiment, when it becomes necessary for the central processing means 9 to transmit a message to the electric device based on the voice recognition processing result, a predetermined voice is synthesized from the storage means 16 and is sent to the electric device via a line. It is the structure which transmits. Further, each electric device includes a D / A converter 22. The voice synthesized by the central processing unit 9 may be transmitted as analog data on a line. However, if it is handled as digital data, it is resistant to noise and easy to perform processing such as compression. That is, the same effect as that obtained by digitizing the voice command in voice recognition can be obtained.
[0035]
(Example 4)
FIG. 5 is a block diagram showing the configuration of the speech recognition processing of the network speech recognition system in the fourth embodiment of the present invention.
[0036]
Human voices are very vague, especially in Japanese, where independence between words is poor, and there are many factors that hinder accurate recognition, such as individual differences and regionality (dialect). Meanwhile, voice recognition processing is realized by making full use of pattern recognition technology. Specific speaker recognition that registers words to be recognized in advance and unspecified speaker recognition that recognizes the speech of an unspecified number of people. Of course, the latter is technically difficult but practical. Sex high.
[0037]
Speech recognition includes word unit recognition and phoneme unit recognition. In word-by-word recognition, a computer analyzes the speech, extracts features, and creates a time series of features. Then, the similarity is compared with the characteristic time-series word dictionary in the processing means, and the result is output as a recognition result.
[0038]
In recognition of phoneme units, input speech is converted into a phoneme symbol string and replaced with a word string. This is parsed and converted to a string. Furthermore, logic analysis and semantic analysis are performed to generate sentences. Since it recognizes the situation before and after the voice is uttered and the individual devices connected to the central processing means, it can also be used to understand simple language such as possible commands and impossible commands. It is extremely difficult, so a high-speed, large-capacity computer and dedicated hardware hybrid system can function efficiently.
[0039]
In the present embodiment, a method based on HMM (Hidden Markov Model) that is resistant to fluctuations of unspecified speakers is employed. HMM is a widely used technique in recent years, and realizes speech recognition in units of phonemes based on this. This technology is, for example, in the 2000 Second IP Award and IP Award “Unspecified Speaker Phonological Level Speech Recognition / Learning Circuit Based on Hidden Markov Model (Kazuhiro Nakamura et al., Nara Institute of Science and Technology)” Indicated.
[0040]
In FIG. 5, the electrical electrical signal (analog) is replaced with a digital signal by the A / D converter 21. The A / D converter 21 may be either in each electric device or in the central processing means. The effect when placed in the apparatus has already been described with reference to FIG.
[0041]
The audio data converted into the digital signal is sent to the controller 23 and sequentially stored in the LPC memory 24. At the same time, it is sent out to the LPC analysis unit 25 one frame at a time, and speech features are extracted using a linear prediction analysis (LPC) technique. That is, a feature vector is calculated. A feature vector is obtained by calculating a linear prediction coefficient based on speech time-series data and an autocorrelation function. The obtained feature vector is recorded in the LPC memory.
[0042]
Next, in the VQ unit 26, the feature vector series obtained by the LPC analysis is converted into a finite number of symbols. Vector quantization (VQ). For each feature vector, a codebook (a table having a symbol number / representative vector pair as an element) is searched and mapped to the symbol of the representative vector having the minimum distance. The codebook is generated by applying a clustering method to a large amount of learning data and stored in the LPC memory 24.
[0043]
Following the above processing, the HMM unit 27 performs voice recognition. The HMM is a phoneme probability model (probability automaton), and one HMM is configured for each phoneme, and a probability parameter is stored in the HMM memory 28. The phoneme indicating the symbol sequence and the highest probability is selected. And the result of this recognition process is output.
[0044]
In the present embodiment, the speech recognition is performed by hardware, but most of the processing is matrix calculation and probability calculation processing that are repeated endlessly. If the central processing means has sufficient processing capability, the LPC analysis unit, VQ unit, HMM unit, etc. can be replaced with software processing by a program. In addition, a hybrid system of hardware and software can be used to balance processing speed and economy.
[0045]
(Example 5)
FIG. 6 is a flowchart of a program having a configuration in which the central processing means in the fifth embodiment of the present invention is a program process.
[0046]
First, digitized audio data is input (step 100). The ID data of the device is added before or after the voice data, and the ID data is recognized and the device to which the voice data is sent is specified (step 101). The ID is given as a unique ID when the electrical device is shipped from the factory. When the electrical device is installed in a household, the individual information of the household is automatically or registered. And stored in the central processing means. By accessing such a database, the central processing means can specify the accessed home and the configuration of the network device in the home (step 102).
[0047]
Subsequently, the voice recognition process starts. First, audio data is segmented for each frame and features are extracted (step 103). Then, the feature vector is quantized (step 104), and the probability of what the data is close to is calculated for each phoneme (step 105). Then, the HMM having the maximum probability is obtained (step 106). The input speech is confirmed and the speech recognition process ends (step 107). The voice recognition process during this period is exactly the same as that executed in hardware in the example of FIG.
[0048]
Now, the confirmed voice is searched by searching a voice list (text table) registered for each device (step 108). This is to determine whether or not the command is accepted by the device. For example, since it is specified from the ID of which model of which maker the voice data is input from the ID, the inspection is performed in accordance with the function and specification of the microwave oven. In the case of the microwave oven, a command “baked in an oven at 180 ° C. for 35 minutes” can be accepted, but this command cannot be executed in a microwave oven that does not have a heater function. In addition, even if a command to “select automatically” comes to the microwave oven, it cannot be handled. This voice list (text table) may be stored in the memory of the central processing unit every time a new product is released, or the central processing unit searches the server of the manufacturer for data via the network. May be. First, it is checked whether or not the voice is registered in the table (step 109). If the voice is not found by this search, an error message is outputted as a voice (step 110). A message can be issued in a detailed manner according to the content of the error.
[0049]
If it is a registered voice, then the stage (operating state) of the device is confirmed (step 111). Since all commands are commanded to the central processing means by voice, the central processing means can identify what the device is now in. Even if there is no mistake in the voice data, it is checked whether it can be accepted now (step 112). Even if the microwave oven is not in operation and you are ordered to “stop”, you cannot respond. Also, it cannot be accepted while the instruction received immediately before is being executed.
[0050]
Furthermore, since the central processing means knows the configuration of the home network device, it can also check the operation of other devices connected to the network (step 113). For example, when the rice cooker has just begun to operate and it is 40 minutes after cooking, when the instruction to reheat is given to the microwave, it asks the operator by voice output whether they can accept it, and calls the instruction again Is also possible.
[0051]
After such processing, a sound notifying that reception has been completed is output (step 114), and an operation output for causing the device to perform a predetermined operation is retrieved from the table (step 115). For example, a microwave oven has loads such as a magnetron, a motor, and a lamp as a heat source, and an operation state chart indicating which of these loads is turned on and which is turned off is stored in advance as a table. When handling these states with 4-digit hexadecimal data, it is necessary to determine in advance on the equipment side and the central processing means that the 01A6 operates the magnetron at full output, the turntable motor rotates, the lamp turns off, etc. The operation command is output (step 116).
[0052]
With the above program, the central sled means can control the electric equipment connected to the net by voice recognition.
[0053]
【The invention's effect】
As described above, according to the present invention, a plurality of electrical devices connected to a network can be operated with a human-friendly operation system called voice recognition, and a high recognition rate is always maintained using the latest voice recognition technology. However, an economically usable system can be realized.
[Brief description of the drawings]
FIG. 1 is a block diagram showing the configuration of a network voice recognition system in a first embodiment of the present invention.
FIG. 2 is a connection diagram showing the configuration of the network voice recognition system.
FIG. 3 is a connection diagram showing a configuration of a network voice recognition system according to a second embodiment of the present invention.
FIG. 4 is a block diagram showing the configuration of a network speech recognition system in a third embodiment of the present invention.
FIG. 5 is a block diagram showing the configuration of speech recognition processing of the network speech recognition system in the fourth embodiment of the present invention.
FIG. 6 is a flowchart of a program having a configuration in which the central processing means in the fifth embodiment of the present invention is a program process.

Claims (8)

Has a electrical device having a separate ID provided with a microphone for converting sound into electrical signals, and a central processing unit with a built-in voice recognition means are connected with the electrical equipment and the line, the operator of the electrical device Inputs an operation command from the microphone, the electric device sends a voice signal and the individual ID to the central processing means, and the central processing means performs voice recognition processing of the transmitted voice signal and the voice. It is an operation command that is accepted by identifying the electrical device that has sent the signal, searching the voice signal for each electrical device from the audio signal, and determining whether the operation command is accepted by the electrical device. If the stipulated by electrical equipment in advance the electric device based on if the processing result is verified is accepted examined whether the conditions which are accepted now as the central processing unit Network speech recognition system configured to transmit a predetermined operating output to the electrical device.   The network voice recognition system according to claim 1, comprising a plurality of the electric devices.   2. The electric apparatus comprising means for converting the sound into an electric signal further comprises an A / D converter, and is configured to convert the sound into a digital signal and transmit it to the central processing means via a line. Network voice recognition system.   The network speech recognition system according to claim 1, wherein the speech recognition processing performed by the central processing unit is speech recognition processing based on linear prediction analysis (LPC), vector quantization (VQ), and hidden Markov model (HMM).   2. The network according to claim 1, wherein the central processing unit is configured to synthesize a predetermined voice from the memory and transmit it via a line when a message needs to be transmitted to the electrical device based on a voice recognition processing result. Speech recognition system.   2. The network voice recognition system according to claim 1, wherein the central processing unit includes hardware for performing voice recognition and a program for performing software processing. The electrical device converts the sound into an electrical signal, the electrical device transmits the audio signal and the individual ID, and the central processing unit identifies the electrical device by the transmitted individual ID and Performing a voice recognition process; searching a voice list registered for each electric device from the voice signal to determine whether the electric device is an operation command to be accepted; It is composed of a step of transmitting a predetermined operation output determined in advance by the electric device and the central processing unit to the electric device based on a processing result if the device is confirmed and checked and accepted . Network speech recognition method. The electrical device converts the sound into an electrical signal, the electrical device transmits the speech signal and the individual ID, and the central processing unit identifies the electrical device by the transmitted individual ID, A step of performing speech recognition processing; a speech list registered for each electrical device is searched from the speech signal to determine whether the electrical command is accepted; If the device is confirmed and checked to be accepted, and if it is accepted, a predetermined operation output determined in advance by the electric device and the central processing unit based on a processing result is transmitted to the electric device . Network speech recognition program.

Images