JPH06259566A - Sense recognizing processor - Google Patents

Abstract

PURPOSE:To provide a sense recognizing processor having a recognition function approximated to human senses. CONSTITUTION:The sense recogniting processor consists of a sense sensor part 1 corresponding to human senses, a processor part 2 for processing the information of the sensor part 1 and recognizing external status, an environment response part 3 for outputting a response corresponding to an environment recognized by the processor part 2, and an environment storing part 4 for storing an environment or the like recognized by the processor part 2 and constituted so as to develope respective senses by respective sensor information. Thus the processor recognizes an environment by a function approximated to recognition by human senses and sends a response corresponding to the environment. A so-called sensor synergy effect for developing the information of a certain sensor information by that of another sensor can also be expected. A response can be changed by a learning function in accordance with user's liking or an environment and the characteristics of a sensor can also be learned. Consequently a non-adjusting system can be constructed by the self-learning function.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a pattern recognition and detection system and the like, and in particular, it is equipped with an artificial sensor that acts on behalf of the human five senses (visual, auditory, olfactory, tactile, taste) using this sensor. The present invention relates to a recognition processor that recognizes an external situation, that is, an environment and outputs a response according to the environment.

[0002]

2. Description of the Related Art An image recognition system for processing and recognizing an image from a TV camera, a voice recognition system for processing and recognizing voice from a microphone, a gas sensor for detecting a gas component, and a pressure for electronically controlling an automobile engine. A technology for detecting a part of information in the outside world such as a sensor has already been developed. An apparatus that combines different types of sensors, processes the signals with a microprocessor, and outputs the signals is described in "Measurement and Control" vol.22, No.12, pp66-67.

[0003]

In the above prior art, human senses such as sight and hearing can be individually and independently recognized, but human senses cannot be integrated and recognized.

Further, although different types of sensors are combined, the detection of the human five senses, the timing of the detection, the learning of the response to the detection result, etc. are not considered.

Therefore, the output of each sensor cannot be comprehensively determined and appropriately dealt with. In addition, it was not possible to learn the environmental response and the output characteristics of the sensor by using the information of other sensors and improving the response to each sensor output.

An object of the present invention is to provide means capable of comprehensively judging the output of each sensor, enhancing the response to each sensor output, and learning the environmental response and the output characteristics of the sensor. It is to provide a sensory recognition processor.

[0007]

In order to achieve the above object, the present invention detects physical quantities such as light information, sound information, odor information, taste information, mechanical quantity information such as weight and pressure, temperature information and the like. The sensory sensor unit, the recognition processor unit that processes information from the sensory sensor unit and recognizes the state of the environment, the environment response unit that outputs a response corresponding to the environment recognized by the recognition processor unit, and the recognition processor unit The present invention proposes a sensory recognition processor including an environment storage unit that stores the created environment.

The environment response section outputs the response to the recognized environment, the means to input the expected value of the response to the recognized environment, and the environment recognized by the processor section based on the expected value of the input response. And means for learning the response to.

It is desirable that the processor section includes a multilayer structure type neural network that processes each input physical quantity and generates an output signal to the environment response section.

The sensory sensor section detects the information from the environment and the optical information based on a command from the processor section.
When it is provided with means for outputting sound information to the environment, it can operate passively or actively.

The environment response unit can be provided with means for inputting an expected value of a response corresponding to the recognized environment and means for inputting an expected value of the sensor output of the sensory sensor unit.

The sensory sensor section may include a condition determination section for outputting a request for recognition processing to the processor section when the detected value of the physical quantity satisfies a predetermined condition.

A plurality of sensory recognition processors are installed, each sensory recognition processor receives the environmental response of another sensory recognition processor, and the processor unit performs recognition processing in accordance with the received environmental response of the sensory recognition processor. A condition determination unit that outputs a request may be provided.

The sensory sensor section, the processor section, the environment response section, and the environment storage section are formed in one wafer or are three-dimensional V.
It is formed in an LSI structure.

In order to achieve the above object, the present invention provides a sensory recognition processor in which the sensory sensor section includes at least a smoke sensor and a temperature sensor, a sprinkler operated by a signal from the environmental response section, and an environmental response section. The present invention proposes an indoor security system including a bell that is operated by a signal from the computer and a means that notifies the central monitoring station through a communication line by a signal from the environmental response unit.

In order to achieve the above object, the present invention also provides a sensory recognition processor in which the sensory sensor section includes at least a liquid level gauge, and a water leak, an oil leak, etc., which is generated from a pipe or the like by a signal from an environmental response section. It proposes a plant monitoring system including means for detecting abnormal conditions such as steam leaks and notifying them to security guards and the like.

In order to achieve the above-mentioned object, the present invention further provides a sensory recognition processor whose sensory sensor section includes at least an odor sensor and a taste sensor, and a signal from the environmental response section to determine the food condition or the shelf life of the food. It proposes a refrigerator equipped with a means for checking and displaying.

In order to achieve the above object, the present invention provides a sensory recognition processor in which a sensory sensor unit includes at least a body temperature sensor and a pulse sensor, and a human health such as a body temperature, a pulse, and a complexion based on a signal from an environment response unit. It proposes a monitoring system consisting of means for checking the condition.

[0019]

In the present invention, the sensory sensor unit detects the external information via the visual sensor, the auditory sensor, the odor sensor, the tactile sensor, the taste sensor, etc., and sends the detection result to the processor unit. The detection result can be sent to the processor unit at all times, but may be sent to the processor unit only when a predetermined condition such as a detection result exceeding a certain threshold value is satisfied. The processor unit recognizes the external environment, that is, the environment based on the information sent from the sensory sensor unit and the environmental data stored in the environment storage unit, and outputs a response corresponding to the environment to the environmental response unit. . The environmental response unit can also take in the response of another sensory recognition processor and create a response signal based on a wider range of information.
The processor unit can change the environmental response and the output characteristic of the sensor by taking in the environmental response of the user's preference and the output characteristic of the sensor of the sensory sensor unit and comparing it with the expected value of the sensor output input separately.

Therefore, it is possible to obtain a sensory recognition processor having means for comprehensively judging the output of each sensor, improving the response to each sensor output, and learning the environmental response and the output characteristic of the sensor.

[0021]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the sensory recognition processor according to the present invention will be described below with reference to FIGS. FIG. 1 is a block diagram showing the configuration of an embodiment of a sensory recognition processor according to the present invention. The sensory recognition processor of this embodiment is
A sense sensor unit including artificial sensors corresponding to the five senses of human beings (visual sense, hearing sense, smell sense, touch sense, taste sense), a processor unit 2 that processes information from the sense sensor unit 1 and recognizes an external situation, and a processor unit The environment response unit 3 outputs a response corresponding to the environment recognized in 2, and the environment storage unit 4 stores the environment recognized in the processor unit. The sensory sensor unit 1 sends the detected external information to the processor unit 2 via the signal line 5. Processor unit 2
Recognizes the situation of the outside world, that is, the environment based on the information of the outside world and the environment data stored in the environment storage unit 4, and sends a response corresponding to the environment through the signal lines 7-1 to 7-n. It is output from the environment response unit 3 via the.

FIG. 2 is a block diagram showing the configuration of an embodiment of the sensory sensor unit 1. The sensory sensor unit 2 converts sensors corresponding to the five human senses, that is, a visual sensor 10, an auditory sensor 11, a tactile sensor 12, an odor sensor 13, a taste sensor 14, and output signals of the respective sensors into data that can be processed by the processor unit 2. And a data conversion unit 15 for converting the data. Details of the structures and characteristics of sensors such as the visual sensor 10, the auditory sensor 11, the tactile sensor 12, the odor sensor 13, the taste sensor 14 are described in the following references and the like. (1) Hiroo Takemura: "CCD Camera Technology", pages 21-39, Radio Technology Co., 1986 (2) Terei Kataoka, Yukio Shibata, Kiyoshi Takahashi, Hiroo Yamazaki Co-edited:
"Sensor Handbook" pages 168-169, 170-171, 210-21
1 page, 330-331, Baifukan, 1986 FIG. 3 is a block diagram showing the configuration of an example in which the sensory sensor unit is formed by a stimulus receptor. In FIG. 2, visual sensors, auditory sensors, tactile sensors, odor sensors, and taste sensors are used as sensors corresponding to the five senses. However, as described on page 181 of the “Sensor Handbook”, Since it is difficult to cover with the five senses, it may be easier to classify according to the type of stimulus. At that time, as shown in FIG. 3, four types of stimulation receptors, that is, mechanoreceptors 100, photoreceptors 101, chemoreceptors 102, and temperature receptors 103 can realize a sensory sensor unit.

Since the output of the sensory sensor unit 1 is usually an analog signal, it is necessary for the data conversion unit 15 to perform A / D conversion into a digital signal. FIG. 4 shows the data conversion unit 15
It is a block diagram showing the composition of one example. The data converter 15 has a data converter 150-1 corresponding to each sensor.
~ 150-5. Each data converter 150-i
Is composed of an amplifier 151-i and an A / D converter 152-i. Signals such as an amplifier gain setting signal, an A / D converter reference voltage, and a conversion clock are supplied from the processor unit 2 to the data converters 150-1 to 150-5 via the signal line 5-0. It As will be described later, the processor unit 2 learns the output characteristics of each sensor from these signals, and according to each characteristic, the gain setting signal of the amplifier, A /
It is used to adjust the reference voltage of the D converter. The amplifier 151-i amplifies the output signal of each sensor and supplies it to the A / D converter 152-i. The A / D converter 152-i converts the output signal of each sensor into a digital signal based on the reference voltage, the conversion clock, and the like.

FIG. 5 is a block diagram showing the configuration of another embodiment of the data conversion unit 15. In the embodiment of FIG. 4, an A / D converter is provided for each sensor and a plurality of conversion clocks are supplied according to the response speed of each sensor, but the configuration of the conversion clock generation means is simplified. Therefore, as shown in FIG. 5, a configuration in which the A / D converter is commonly used is also conceivable. The data conversion unit 15 includes an amplifier unit 153 including amplifiers 151-1 to 151-5 provided for each sensor,
It comprises a hold / selector 154 for selecting one of the amplifiers 151-1 to 151-5 and holding the output signal thereof, and an A / D converter 152 for converting the selected analog signal. In this case also, the signal line 5-
Signals such as an amplifier gain setting signal, a data selection signal, a reference voltage of the A / D converter, and a conversion clock are supplied via 0. The amplifier 151-i amplifies the output signal of each sensor. The hold / selector 154 holds the output signal of the amplifier selected according to the data selection signal and supplies it to the A / D converter 152-i. The A / D converter 152-i converts the held output signal of the amplifier into a digital signal based on the reference voltage and the conversion clock.

FIG. 6 is a block diagram showing the configuration of an embodiment of the data conversion unit 15 when the sensor is also used as an active element. The sensor of the sensory sensor unit 1 shown in FIGS. 2 and 3 is a sensor that realizes only a passive function of detecting information in the external world. On the other hand, for example, if a light emitting element is also built into the visual sensor to emit light, or if an electric signal is applied to a semiconductor microphone used as a hearing sensor to generate vibration (sound), the sensor is activated. It is also possible to use. The data converter 150-i of this embodiment includes amplifiers 151-i, 155-i, A
The D / A converter 152-i and the D / A converter 156-i. The output signal of the sensor is the amplifier 151-i, A / D
It is supplied to the processor unit 2 via the converter 152-i and the signal line 5-i. On the contrary, the output signal from the processor unit 2 includes the signal line 5-j, the D / A converter 156-i, and the amplifier 15
It is supplied to the sensor via 5-i. In that case, through the signal line 5-0, the gain setting signal of the amplifier, the A / D
Signals such as a converter operation command or a D / A converter operation command, a reference voltage, and a conversion clock are supplied.

FIG. 7 is a block diagram showing the configuration of an embodiment of the sensory sensor unit including means for correcting the temperature characteristic of each sensory sensor along with other physical quantity detecting means. In the configuration of the sensory sensor unit shown in FIG.
It cannot detect physical quantities such as infrared rays, humidity, and radiation. Further, since the sensory sensor unit 1 is a part that detects information in the external environment, temperature correction of its characteristics is important. In addition to the components of FIG. 2, the sensory sensor unit 1 of FIG. 7 includes signal lines 17-1 to 17-n for inputting information of an external sensor and a temperature sensor for detecting the external temperature and the internal temperature of the sensory recognition processor. And the visual and auditory sensors can be used both passively and actively.

When recognizing an environment using the sensation recognition processor of the present invention, there are three possible processing timings for the output signal of the sensation sensor unit 1, for example. (1) The signal from the sensory sensor unit 1 is processed at every predetermined sampling time. (2) Processing is performed only when the signal from the sensory sensor unit 1 satisfies a predetermined condition. (3) The signal from the sensory sensor unit 1 is processed according to the response of another sensory recognition processor.

FIG. 8 is a block diagram showing the configuration of an embodiment of the sensory sensor unit 1 in which the condition determining unit 19 is incorporated in the sensory sensor unit 1 of FIG. In order to be able to perform two types of processing (1) and (2) using the sensory recognition processor of the present invention, a condition determination section 19 is provided between the sensory sensor section 1 and the processor section 2. Good. Of course, processor unit 2
It is also conceivable to incorporate the condition determination unit 19 in the. The sense sensor unit of FIG. 8 is different from the condition determination unit 19 of the configuration of FIG.
Further, a signal line 191 for designating a judgment condition, a mask signal line 192, and a signal line 193 for notifying the result of the condition judgment are added to the condition judgment unit 19. Condition determination unit 1
Reference numeral 9 captures the output signals 5-1 to 5-n + 1 of the data converter 15 and makes a determination based on the determination condition supplied from the processor unit 2 via the signal line 191 and via the signal line 193.
The determination result is output to the processor unit 2. At this time, the output signal used for the determination can be selected from the output signals 5-1 to 5-n + 1 of the data converter 15 by the mask supplied from the processor unit 2 via the signal line 192.

FIG. 9 is a block diagram showing the configuration of an embodiment of the condition judging section 19. The condition determination unit 19 is the arithmetic unit 1
94-1 to 194-n + 1 and a logic circuit 195.
The output signals 5-1 to 5-n + 1 of the data converter 15 are conditionally determined by the respective arithmetic units 194-1 to 194-n + 1, integrated by the logic circuit 195, and finally the result of the condition determination. Is output as. This condition determination result is used, for example, as an interrupt signal to the processor unit 2.

FIG. 10 is a block diagram showing the configuration of an embodiment of the arithmetic unit 194-i shown in FIG. Arithmetic unit 194-i
Is composed of a register 1941 for storing a threshold value 1, a register 1942 for storing a threshold value 2 and a comparator 1943.

FIG. 11 is a block diagram showing the configuration of another embodiment of the arithmetic unit 194-i shown in FIG. In this embodiment, a cumulative adder 1944 and a register 1945 are added to the configuration of the embodiment of FIG. Where the register 1945
Is a barrel shift register, adder 1944
Then, the average value of the accumulated results can be obtained.

FIG. 12 is a diagram for explaining the comparison operation function of the comparator 1943 in FIGS. 10 and 11. TH
1 represents the threshold value 1 and TH2 represents the threshold value 2. The four types of comparison functions shown in FIG.
It can be selected by supplying the judgment condition via 1. The comparison function 0 outputs 1 when the input of the comparator 1943 exceeds the threshold value 2. The comparison function 1 outputs 1 when the input of the comparator 1943 is less than or equal to the threshold value 1. The comparison function 2 is
It outputs 1 when the input of the comparator 1943 is between the threshold 1 and the threshold 2. The comparison function 3 is the comparator 194.
When the input of 3 is less than the threshold value 1 or more than the threshold value 2, 1 is output.

If an arithmetic unit having the functions of FIGS. 9 to 12 is used, the image signal is processed when the average brightness of the image signal of the outside world detected by the sensor 10 becomes larger than a predetermined value. Or when the concentration of the gas component detected by the odor sensor 13 exceeds a predetermined value, the processor unit 2
It is possible to interrupt the.

FIG. 13 is a diagram showing the configuration of an embodiment of the logic circuit 195 of FIG. When the logic circuit is configured in this way, the sensor for determining the condition can be selected by the mask. This logic circuit is composed of AND circuits 1951-1 to 195.
1-n + 1 and an OR circuit 1952. Signal line 192
The output of the arithmetic unit in which 1 is designated in the mask supplied from is output as the condition determination result via the OR circuit 1952 and the signal line 193. That is, AN
The output of the D circuits 1951-1 to 1951-n + 1 is the signal line 1
It is output from 93.

FIG. 14 is a block diagram showing the configuration of an embodiment of the processor unit 2 for processing the external information detected by the sensory sensor unit 1. The processor unit 2 can be configured by using, for example, a plurality of processor elements. A high-performance microcomputer is used as the processor element. The processor unit 2 of FIG. 14 is composed of eight processor elements 20-1 to 20-8. These processor elements 20-1 to 20-8 are connected to the environment storage unit 4 and the environment response unit 3 via the bus 21. In this embodiment, the processor element 20-1 operates as a master processor, and exchanges signals such as a conversion clock, an amplifier gain setting signal, a judgment condition signal, a mask signal, and a condition judgment result with the sense sensor unit 1. . Further, the processor elements 20-1 to 20-8 are
Based on the processing algorithm stored in the environment storage unit 9, the sensor information corresponding to each can be independently processed in parallel. The processor element 20-1 integrates the processing results and outputs them to the environment response unit 8. In processing individual sensor information, if other sensor information acquired via the bus 21 is used as a reference, the processing can be advanced.

FIG. 15 is a diagram showing an embodiment in which a neural network is applied as one of the processing algorithms in the processor section 2. This processing algorithm is a multilayer structure type neural network. From the left of FIG. 15, they are called an input layer, an intermediate layer, and an output layer, and ◯ represents a neuron. The advantage of using a neural network is that sensor information with different properties can be processed comprehensively, so you can expect human-like processing results due to the synergy effect.
As will be described later, the output value for the processing result can be changed by learning to suit the user's preference. When a neural network is used, the environment storage unit 4 stores the weight between neurons. Information of each sensor is input to the neuron in the input layer. For example, the density value of each pixel is input from the visual sensor, the sampled voice waveform is input from the auditory sensor, and the odor component is input from the odor sensor. The neuron in the middle layer multiplies the input information by a predetermined weight to obtain the sum thereof, and transmits the sum to the neuron in the output layer through an output function such as a sigmoid function. The output layer neuron also performs the same calculation as the intermediate layer neuron to obtain an output value, and outputs the output value to the environment response unit 3.

FIG. 16 is a block diagram showing the configuration of an embodiment of the environment response unit 3 of FIG. The environmental response unit 3 is D /
An analog response unit 30 including an A converter, a digital response unit 31, and a network controller 32.
The processor unit 2 in FIG. 1 supplies data to be output to the analog response unit 30, the digital response unit 31, and the network controller 32 via the signal line 8 in accordance with the environment recognition result, and the sensory recognition of the present invention is performed. Drives external equipment connected to the processor. Although not shown here, it is also possible to directly drive an external device by using the micromachining technology which is currently attracting attention. The feature of the environment response unit 3 is not only to output a response, but also to process the response result (external response 1 to external response n) of another environment recognition processor from the digital response unit 31 and the network controller 32 and information of the sensory sensor. The processing algorithm (the expected output value of the neural network described above) can be input.

FIG. 17 is a diagram showing an example of a data structure stored in the environment storage unit 4. FIG. 17A shows a case where only the normal environment is stored, and FIG. 17B shows
The case where the environments of a plurality of states and the responses at that time are stored in a pair format is shown. The environment storage method is shown in Fig. 1.
As shown in FIG. 7 (a), information of a sensory sensor in a certain state and a processing algorithm, for example, in the case of a visual sensor, each sensory sensor such as a processing algorithm of a visual image or a modeled image and visual sensor information. It is conceivable to memorize in. In the case of FIG. 17A, the information from the sensory sensor unit 1 is compared with the normal environment data in the environment storage unit 4, and a response may be returned from the environment response unit 3 according to the distance.
On the other hand, in the case of FIG. 17B, the information from the sensory sensor unit 1 is compared with the environmental data of each state of the environmental storage unit 4,
It suffices to return the response in the state of the shortest distance.

FIG. 18 is a block diagram showing a device configuration and a signal flow as a whole when learning the environmental response and the sensor characteristic by using the sensory recognition processor of the present invention. The sensory recognition processor may be switched to the learning mode by, for example, an external response input from the environment response unit 3 or communication. Here, a neural net is used as the processing algorithm of the processor unit 2. First, when learning the environmental response, the signal lines 7-1 to 7
The expected output value of the response is input from the environment response unit 3 via -n. The processor unit 2 executes learning of the neural network using an algorithm such as back propagation based on the information from the sensory sensor unit 1 and the expected output value from the environment response unit 3, and the learning result is obtained. The weight between neurons is stored in the environment storage unit 4.

FIG. 19 is a diagram specifically showing the structure of the neural network. w is the weight between neurons,
For example, wij (k) represents the weight between the i-th neuron in the intermediate layer k-1 and the j-th neuron in the intermediate layer k.

FIG. 20 is a diagram showing an example of storage in the environment storage unit 4. By doing so, it becomes possible to learn the response corresponding to the environment according to the user's preference.

Next, the case of learning the sensor characteristics will be described. In this learning as well, as in the case of learning the environmental response, the desired sensor characteristics may be input from the environmental response unit 3 via the signal lines 7-1 to 7-n. For example, when the expected output value of the sensor as shown by the broken line in FIG. 21 is input and the actual output value of the sensor is as shown by the solid line, the correction coefficient as shown in FIG. 22 is obtained from the difference between the two. When the correction coefficient is obtained, the processor unit 2 determines that the data conversion unit 1 in FIG.
5, the gain of the amplifier or A via the signal line 5-0
It is possible to instruct to change the reference voltage or the like of the / D converter according to the correction coefficient.

Similarly, when the expected output value of the sensor as shown by the broken line in FIG. 23 is input and the actual output value of the sensor is as shown by the solid line, the correction coefficient as shown in FIG. Is obtained. As shown in FIG. 24, when the non-linearity of the correction coefficient is strong, the processor unit 2 can perform the correction by the program processing. If this learning of the correction coefficient is executed in the background, it can be corrected in accordance with the temperature change and the secular change of the sensor. In addition, when the yield of the sensory sensor is low or when a sensory sensor is not used, the evaluation value of the sensor output may be changed.

FIG. 25 is a diagram showing an example of the evaluation value of the sensor output. In this example, the taste sensor is not used at all, the information of the visual sensor is used 100%, and the information of the auditory sensor, the tactile sensor, and the olfactory sensor is used at a reduced rate to some extent. The change in the rate of using the evaluation value can be realized by instructing the data conversion unit 15 to change the gain of the amplifier via the signal line 5-0. By using the self-learning function according to the present invention having such sensor characteristics, a non-adjustment system can be constructed.

FIG. 26 is a block diagram showing a device configuration and a signal flow when an environment is recognized using the sensory recognition processor of the present invention. When recognizing the environment, there are three possible operation modes of the sensory recognition processor. (1) The signal from the sensory sensor unit 1 is processed at every predetermined sampling time. (2) Processing is performed only when the signal from the sensory sensor unit 1 satisfies a predetermined condition. (3) The signal from the sensory sensor unit 1 is processed according to the response of another sensory recognition processor.

FIG. 27 is a diagram showing an example of the processing timing of the processor section. The synchronization process of FIG. 27A corresponds to the above operation mode (1). In this method,
The processor unit 2 issues an internal trigger at predetermined sampling intervals to execute environment recognition processing. FIG. 27 (b)
The asynchronous processing of (1) corresponds to the above operation modes (2) and (3). In this method, the processor unit 2 issues an internal trigger in response to an interrupt signal from the condition determination unit 19 or an interrupt signal generated by the environment response unit 3 accepting an external response, and executes an environment recognition process. .

FIG. 28 is a diagram showing an example of an environment recognition processing procedure which is activated by issuing an internal trigger and is common to all operation modes. The processor unit 2 takes in information from the sensory sensor unit 1 and recognizes the environment. Next, the recognized environment is compared with the normal environment stored in the environment storage unit, and if there is an abnormality, a response corresponding to it is output.

As an implementation form of the sensory recognition processor of the present invention, one shown in FIGS. 29 to 31 can be considered. Figure 2
9 is a diagram showing a structure in which the present invention is realized as hardware when the yield and the degree of integration of the LSI are not so high, and an embodiment in which the sensor and the information processing section are provided as separate chips and are configured by a multi-chip module Is.

FIG. 30 shows an embodiment in which the sensor and the information processing unit are formed in one wafer by the wafer scale integration technique. In this case, since the temperature sensor is also formed within one wafer, it is suitable for application to the embodiments of FIGS. 32 and 33 described later.

FIG. 31 shows an embodiment in which the sensor and the information processing unit are completely integrated into a single chip by using a three-dimensional VLSI to reduce the system size and cost. Also in this case, since the temperature sensor can be formed in one chip, it will be described later with reference to FIGS.
Is suitable for application to the embodiment.

There are various fields of application of the sensory recognition processor of the present invention, such as security, FA, home appliances, and medical care. FIG. 32 is a block diagram showing an example of a device configuration when the present invention is applied to security monitoring in a building. The sensory recognition processor of the present invention is installed at the monitoring target location, and they are connected by a communication line or the like. Recognize whether a suspicious person is intruding from the information from the visual sensor, auditory sensor, and tactile sensor, and at the same time, recognize whether gas leaks, fires, etc. have occurred. 3 activates sprinklers and bells and reports to the central monitoring station via communication lines.

The present invention can be similarly applied to a plant monitoring system for detecting an abnormal state such as a water leak, an oil leak, or a steam leak which occurs from piping of a factory or a plant and notifying a guard or the like. Also, it can be used as a man-machine interface means by incorporating it into a computer or the like, and the computer can be customized according to one's taste.

FIG. 33 is a chart showing the relationship between a combination example of the five-sense sensor and the temperature sensor and its application field. In FIG. 33, it is premised that the same type of sense, for example, a combination of sight and sight is not used, but it goes without saying that it may be used. In each application field, the filled portion means that the sense is not used. For example, in automobile engine control, the taste is not used. In the field of home electric appliances, for example, if this sensory recognition processor is installed in a refrigerator, by utilizing an odor sensor and a taste sensor, it is possible to check the taste of food and the shelf life of the food with a feeling similar to that of a human. On the other hand, if incorporated into a wristwatch, for example, it is possible to monitor health conditions such as body temperature, pulse rate, and complexion with a feeling similar to that of a human. Further, as the processing function of the sensory recognition processor of the present invention is enhanced by self-learning, a simulator that simulates human reaction can be realized.

[0054]

According to the present invention, it is possible to realize a recognition function close to human sense, recognize the situation of the outside world, that is, the environment, and return a response corresponding to the environment.
A synergistic effect of the sensor, which enhances the processing of certain sensor information by using other sensor information, can be expected.

Since the processor of the present invention has a learning function, the response according to the environment can be changed according to the preference of the user of the processor of the present invention.

Further, not only the response but also the characteristics of the sensor can be learned, and the environment can be recognized while changing the evaluation value of the sensor output according to the characteristics. Since the learning of the sensor characteristics is self-learning, an unadjusted system can be constructed.

The sensory recognition processor can be realized by a single chip, and a very small / low cost system can be constructed.

[Brief description of drawings]

FIG. 1 is a block diagram showing the configuration of an embodiment of a sensory recognition processor according to the present invention.

FIG. 2 is a block diagram showing a configuration of an embodiment of a sensory sensor unit.

FIG. 3 is a block diagram showing a configuration of an embodiment in which a sensory sensor unit is formed of a stimulus receptor.

FIG. 4 is a block diagram showing a configuration of an embodiment of a data conversion unit.

FIG. 5 is a block diagram showing the configuration of another embodiment of the data conversion unit.

FIG. 6 is a block diagram showing a configuration of an embodiment of a data conversion unit when the sensor is also used as an active element.

FIG. 7 is a block diagram showing a configuration of an embodiment of a sensory sensor unit including another sensor for detecting a physical quantity and a device for correcting the temperature characteristic of each sensory sensor.

8 is a block diagram showing a configuration of an embodiment of a sensory sensor unit in which a condition determination unit is incorporated in the sensory sensor unit of FIG.

FIG. 9 is a block diagram illustrating a configuration of an embodiment of a condition determination unit.

10 is a block diagram showing the configuration of an embodiment of the arithmetic unit of FIG.

11 is a block diagram showing the configuration of another embodiment of the arithmetic unit of FIG. 9. FIG.

12 is a diagram illustrating a comparison calculation function of the comparators in FIGS. 10 and 11. FIG.

13 is a diagram showing a configuration of an embodiment of the logic circuit of FIG.

FIG. 14 is a block diagram showing a configuration of an embodiment of a processor unit that processes information on the outside world detected by a sensory sensor unit.

FIG. 15 is a diagram showing an embodiment in which a neural network is applied as one of processing algorithms in the processor unit.

16 is a block diagram showing the configuration of an embodiment of the environment response unit in FIG. 1. FIG.

FIG. 17 is a diagram showing an example of a data structure stored in an environment storage unit.

FIG. 18 is a block diagram showing a device configuration and a signal flow as a whole when learning an environmental response and a sensor characteristic by using the sensory recognition processor of the present invention.

FIG. 19 is a diagram specifically showing the structure of a neural network.

FIG. 20 is a diagram illustrating an example of storage in an environment storage unit.

FIG. 21 is a diagram showing an example of output characteristics of a sensor.

22 is a diagram showing an example of a correction curve for the output of the sensor of FIG.

FIG. 23 is a diagram showing another example of the output characteristics of the sensor.

FIG. 24 is a diagram showing an example of a correction curve for the output of the sensor of FIG. 23.

FIG. 25 is a diagram showing an example of evaluation values of sensor output.

FIG. 26 is a block diagram showing a device configuration and a signal flow when an environment is recognized using the sensory recognition processor of the present invention.

FIG. 27 is a diagram illustrating an example of processing timing of a processor unit.

FIG. 28 is a diagram showing an example of an environment recognition processing procedure that is common to all operation modes and is activated by issuing an internal trigger.

FIG. 29 is a diagram showing a structure in which the present invention is realized as hardware when the yield and the degree of integration of the LSI are not so high, and the sensor and the information processing unit are formed as separate chips in a multi-chip module. Here is an example.

FIG. 30 shows an example in which a sensor and an information processing unit are configured in one wafer by a wafer scale integration technique.

FIG. 31 is an embodiment in which a sensor and an information processing unit are completely integrated into one chip by using a three-dimensional VLSI, and the system is downsized and the price is reduced.

FIG. 32 is a block diagram showing an example of a device configuration when the present invention is applied to security monitoring in a building.

FIG. 33 is a chart showing a combination example of a five-sense sensor and a temperature sensor and its application field.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Sensory sensor section 2 Processor section 3 Environmental response section 4 Environmental storage section 5 Signal line 7 Signal line 8 Signal line 9 Signal line 10 Visual sensor 11 Hearing sensor 12 Tactile sensor 13 Odor sensor 14 Taste sensor 15 Data conversion section 17 Signal line 18 Temperature sensor 19 Condition determination unit 20 Processor element 21 Bus 30 Analog response unit 31 Digital response unit 32 Network controller 100 Mechanical receptor 101 Photoreceptor 102 Chemiceptor 103 Temperature receptor 150 Data converter 151 Amplifier 152 A / D converter 153 amplifier 156 D / A converter 192 mask 194 arithmetic unit 195 logic circuit

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Koji Kamejima, Inventor Kojitamachi 502, Tsuchiura City, Ibaraki Prefecture, Institute of Mechanical Research, Hiritsu Manufacturing Co., Ltd. Ceremony Hitachi Energy Research Institute

Claims (13)

[Claims] 1. A sensory sensor section for detecting a physical quantity such as light information, sound information, odor information, taste information, mechanical quantity information such as weight and pressure, temperature information and the like, and an environment for processing information from the sensory sensor section. A sense consisting of a recognition processor unit that recognizes the situation, an environment response unit that outputs a response corresponding to the environment recognized by the recognition processor unit, and an environment storage unit that stores the environment recognized by the recognition processor unit. Recognition processor. 2. The sensory recognition processor according to claim 1, wherein the environment response unit is input with a response output unit for the recognized environment and a unit for inputting an expected value of the response for the recognized environment. And a means for learning the response to the environment recognized by the processor unit based on the expected value of the response. 3. The sensory recognition processor according to claim 1, wherein the processor unit includes a multi-layered neural network that processes each of the input physical quantities and generates an output signal to the environment response unit. Sensory recognition processor characterized by 4. The sensory recognition processor according to claim 1, wherein the sensory sensor unit detects the information from the environment, and optical information based on a command from the processor unit. , A sensory recognition processor characterized by comprising means for outputting sound information to the environment. 5. The sensory recognition processor according to claim 1, wherein the environment response unit inputs an expected value of a response corresponding to the recognized environment, and a sensor of the sensory sensor unit. A sensory recognition processor comprising: means for inputting an expected value of output. 6. The sensation recognition processor according to claim 1, wherein the sensation sensor unit causes the processor unit to perform recognition processing when a detected value of the physical quantity satisfies a predetermined condition. A sensory recognition processor characterized by comprising a condition determination unit for outputting a request. 7. The sensory recognition processor according to claim 1, wherein there are a plurality of sensory recognition processors, and each sensory recognition processor receives an environmental response of another sensory recognition processor. A sensory recognition processor comprising: a condition determination unit that outputs a recognition processing request to the processor unit according to the received environmental response of the sensory recognition processor. 8. The sensory recognition processor according to claim 1, wherein the sensory sensor unit, the processor unit, the environment response unit, and the environment storage unit are formed in one wafer. A sensory recognition processor characterized by being performed. 9. The sensory recognition processor according to claim 1, wherein the sensory sensor unit, the processor unit, the environment response unit, and the environment storage unit have a three-dimensional VLSI structure. A sense recognition processor characterized by being formed. 10. A sensory recognition processor according to claim 1, wherein the sensory sensor section includes at least a smoke sensor and a temperature sensor, and a sprinkler operated by a signal from the environmental response section. An indoor security system comprising: a bell operated by a signal from the environmental response unit; and means for notifying a central monitoring station via a communication line by a signal from the environmental response unit. 11. A sensory recognition processor according to claim 1, wherein the sensory sensor section includes at least a liquid level gauge, and a water leak from a pipe or the like caused by a signal from the environmental response section. A plant monitoring system that includes means for detecting abnormal conditions such as oil leaks and steam leaks and notifying them to security personnel. 12. A sensory recognition processor according to claim 1, wherein the sensory sensor unit includes at least an odor sensor and a taste sensor, and a signal from the environment response unit indicating whether or not food is eaten. A refrigerator provided with means for checking and displaying the expiration date. 13. A sensory recognition processor according to claim 1, wherein the sensory sensor section includes at least a body temperature sensor and a pulse sensor, and a body temperature, a pulse, and a complexion based on a signal from the environment response section. A monitoring system that consists of a means for checking the health condition of humans.