Information processor

Abstract

To enable prediction of the duration of body odor.SOLUTION: A processor comprises a storage unit for a model capable of obtaining a predicted value for the duration of body odor, and a model generation unit that generates a model by machine learning using a plurality of data sets, each of which is obtained at a plurality of measurement time points different from each other, including data sets that associate a detected value of an odor component of urine by an odor sensor, a measurement time point at which the urine odor is measured, and a body odor level at the measurement time point with one another.SELECTED DRAWING: Figure 2

Description

[Technical field]

[0001] The present embodiment relates to an information processing device. [Background technology]

[0002] Conventionally, it is known that a semiconductor gas sensor using metal oxide is used to measure body odor emitted from the human body and to determine the strength of the odor based on the detection value of the odor components. Patent Document 1 describes a method in which an odor component in a detection space is measured by an odor detection unit, and the odor level of a user who plans to use a vehicle is predicted based on the odor detection value output by the odor detection unit and the planned start time of use input by an input unit. Patent Document 2 describes a method in which urine from a user is collected and the concentration of a specific component is measured by a sensing means incorporated in a urine collection unit. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent Publication No. 2022-141210 [Patent Document 2] JP 2006-029819 A Summary of the Invention [Problem to be solved by the invention]

[0004] Even if an odor sensor is placed close to the armpit of a human body and the body odor level is directly measured, the measurement accuracy is insufficient. This is because the concentration of odorous components contained in body odor is low. Also, there are various types of body odor, and the duration of body odor varies depending on the type of odor. In the past, there was an issue of not being able to know the duration of body odor.

[0005] One embodiment aims to be able to predict the duration of body odor. [Means for solving the problem]

[0006] According to one embodiment, an information processing device includes a storage unit for a model capable of obtaining a predicted value of the duration of body odor, and a model generation unit that generates a model by machine learning using a plurality of data sets obtained at a plurality of different measurement times, including a data set that associates the detection value of urine odor components detected by an odor sensor, the measurement time when the urine odor was measured, and the body odor level at the measurement time, and stores the model in the storage unit. Effect of the Invention

[0007] According to the present invention, it is possible to predict the duration of body odor. [Brief description of the drawings]

[0008] [Figure 1] FIG. 1 is a diagram showing a configuration of a duration prediction system according to the first embodiment. [Diagram 2] FIG. 2 is a diagram illustrating a functional configuration of the sensor controller according to the first embodiment. [Diagram 3] FIG. 3 is a diagram showing an example of odor sensor data. [Figure 4] FIG. 4 is a diagram for explaining an example of a body odor level. [Diagram 5] FIG. 5 is a diagram showing examples of multiple data sets that the model generation unit uses for machine learning. [Figure 6] FIG. 6 is a diagram showing the relationship between the duration actually measured and the predicted value of the duration according to the duration prediction model. [Figure 7] FIG. 7 is a diagram for explaining an example of prediction of duration by the duration prediction model. [Figure 8] FIG. 8 is a flow chart showing a series of processes when a subject measures the odor of urine. [Figure 9] FIG. 9 is a flowchart showing the operation of the model generating unit. [Figure 10]FIG. 10 is a diagram illustrating an example of a data set input to the sensor controller according to the second embodiment. [Figure 11] FIG. 11 is a diagram illustrating a functional configuration of the sensor controller according to the second embodiment. As illustrated in FIG. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0009] Hereinafter, an information processing device for predicting the duration of body odor and a duration prediction system including the information processing device according to an embodiment will be described in detail with reference to the accompanying drawings. Note that the present invention is not limited to these embodiments.

[0010] (First embodiment) FIG. 1 is a diagram showing a configuration of a duration prediction system 1000 according to the first embodiment.

[0011] The duration prediction system 1000 includes an odor sensor 1, a sensor controller 2, and an information terminal 3. The odor sensor 1 and the sensor controller 2 are connected by a wired or wireless communication path to enable mutual transfer of information. The sensor controller 2 and the information terminal 3 are connected by a wired or wireless communication path to enable mutual transfer of information.

[0012] The sensor controller 2 is an example of an information processing device that predicts the duration of body odor.

[0013] Odors contain odorants. Odorants can be single or complex molecules. Examples of such molecules include alcohol, allicin, the odorant of garlic, and acetone, which is found in breath. Odorants can be detected as odors by the human sense of smell if they are present at a certain concentration or higher.

[0014] Body odor is an odor emitted by living organisms. For example, body odor can be measured by bringing an odor sensor close to the skin, and the body odor level can be determined based on the detected values ​​of the odor components. The body odor level is an index that indicates the strength of the body odor. For example, data relating the detected values ​​of odor components of body odor to the subjective odor strength judged by a person when smelling the body odor can be collected, and a learning model can be generated using machine learning. By using this learning model, an unknown body odor can be measured with an odor sensor, and the unknown body odor level can be determined based on the detected values ​​of the odor components.

[0015] In the above-mentioned determination method, the concentration of odor components emitted from the human body is very small, so it is prone to variation depending on the measurement time and measurement location, making it difficult to obtain measurement reproducibility of the odor sensor. Therefore, there is a problem in measuring the odor by bringing the odor sensor close to a living body and accurately determining the body odor level. In this embodiment, the odor emitted from urine is measured with the odor sensor 1, and data is collected that associates the detection value of the odor components, the body odor level, and the time of measurement, and a learning model is generated using machine learning.

[0016] As shown in Figure 3, the level of body odor decreases over time. For this reason, if the odor is measured at multiple points when the body odor level is changing over time, and a model trained on the detection value of the odor sensor, the date and time of measurement by the odor sensor, and the body odor level is used, the duration of the body odor can be estimated. If the duration of the body odor is known, measures can be taken according to the duration of the body odor, which is convenient. For example, deodorizing measures can be taken.

[0017] As described above, using the machine learning model, the sensor controller 2 predicts the duration of body odor based on measurements of the subject's urine odor.

[0018] For example, the subject may collect urine in a cup or the like, and measure the odor by bringing the odor sensor 1 close to the urine collected in the cup.

[0019] 1, odor sensor 1 is installed near toilet 2000. When a subject urinates into toilet 2000, he or she operates odor sensor 1 to instruct measurement. In response to the instruction, odor sensor 1 uses a pump to draw in air around the water surface of toilet 2000 via tube 11. Then, odor sensor 1 measures the odor components that make up the urine odor contained in the drawn in air, and obtains an output value.

[0020] In addition to the pump, the odor sensor 1 includes an odor sensor element and a chamber in which the odor sensor element is housed. Gas containing the odor to be measured (here, gas sucked from around the water surface of the toilet bowl 2000) is introduced into the chamber by the pump. This exposes the odor sensor element to the gas containing the odor to be measured. The odor sensor element then measures the odor components contained in the gas.

[0021] As the odor sensor element, a QCM (Quartz Crystal Microbalance) type sensor element, a MEMS (Micro Electro Mechanical Systems) type sensor element, an FBAR (Film Bulk Acoustic Resonator) type sensor element, or any other sensor element may be adopted. The odor sensor element selectively responds to a single or complex molecule contained in the odor component.

[0022] The number of odor sensor elements included in the odor sensor 1 is not limited to one. The odor sensor 1 may include a plurality of sensor elements each having a different reactivity. These plurality of sensor elements may have different reactivities, for example, by having sensitive films each containing a different material.

[0023] The odor sensor outputs the reaction when it measures the odor to the sensor controller 2. This output value is corrected with a reference value for calibration. Disturbances such as temperature and humidity may also be corrected with the output values ​​of a separately provided temperature sensor and humidity sensor. The detection values ​​of odor components obtained in this way are used for machine learning. The detection values ​​of odor components are sometimes simply called detection values.

[0024] The sensor controller 2 executes control of the odor sensor 1. This may include obtaining detected values ​​of odor components (hereinafter, referred to as odor sensor data), aggregating the odor sensor data, processing the odor sensor data, etc. Here, the sensor controller 2 calculates a predicted value of the duration of the subject's body odor based on the odor sensor data.

[0025] The information terminal 3 is a personal computer, a PDA (Personal Digital Assistant), a smartphone, etc. A user including a subject can operate the information terminal 3 to perform various inputs to the sensor controller 2 and obtain information output from the sensor controller 2 from the information terminal 3.

[0026] The user can input a request for a predicted value of the duration of the body odor from the information terminal 3 to the sensor controller 2. In response to the request, the sensor controller 2 can display the predicted value of the duration of the body odor on a display device provided in the information terminal 3.

[0027] FIG. 2 is a diagram showing a functional configuration of the sensor controller 2 according to the first embodiment.

[0028] The sensor controller 2 includes a model generation unit 211, a storage unit 212, a duration prediction unit 213, and an output unit 214. The model generation unit 211, the duration prediction unit 213, and the output unit 214 are realized by a processor. The storage unit 212 is realized by a volatile memory or a non-volatile memory.

[0029] The model generation unit 211 generates the duration prediction model 100. The duration prediction model is a trained model configured to be capable of acquiring a predicted value of the duration of a subject's body odor from odor sensor data.

[0030] A plurality of data sets acquired at different times are prepared in order to generate the duration prediction model 100. Each data set includes odor sensor data, acquisition date and time, and body odor level.

[0031] Each data set includes a body odor level acquired simultaneously with the odor sensor data. The body odor level is information that quantifies the strength of the body odor felt by a person. A data set can be considered as data that associates the detection value of an odor component calculated based on the output value when the odor of urine is measured by the odor sensor 1, the strength of the body odor, and the acquisition date and time, which is the measurement time when the odor of urine was measured. The model generation unit 211 generates the duration prediction model 100 by machine learning using multiple data sets acquired at different times.

[0032] The model generation unit 211 stores the generated duration prediction model 100 in the storage unit 212 .

[0033] When new odor sensor data is input, the duration prediction unit 213 calculates a predicted value of the subject's body odor duration based on the newly input odor sensor data and its acquisition date and time. The duration prediction unit 213 obtains a predicted value of the subject's body odor duration by using the duration prediction model 100.

[0034] The output unit 214 outputs the predicted value of the duration obtained by the duration prediction unit 213 .

[0035] The method of outputting the predicted value of the duration is not limited to a specific method. The output unit 214 may transmit the predicted value of the duration to the information terminal 3, or may store the predicted value of the duration in an internal or external storage device of the sensor controller 2.

[0036] Next, the duration prediction model 100 will be described in more detail.

[0037] FIG. 3 is a diagram showing an example of odor sensor data.

[0038] Case #1 shows the transition of odor sensor data when the odor of urine is measured on a certain day. Case #2 shows the transition of odor sensor data when the odor of urine is measured on another day when the body odor is strong. As shown in the figure, the magnitude of the sensor value of the odor sensor data obtained from the odor of urine differs depending on the case. Also, the time required for the odor sensor data to decay to a predetermined level, for example, 30, differs depending on the case.

[0039] FIG. 4 is a diagram for explaining an example of a body odor level.

[0040] FIG. 4 shows the transition of the odor sensor data acquired in each of the cases shown in FIG. 3, and the transition of the body odor level determined at the same timing as the odor sensor data was acquired.

[0041] In case #1, the body odor level was 80 when the odor sensor data was first acquired. Then, 9 hours later, the body odor level had decreased to 30.

[0042] In case #2, the body odor level was 150 when the odor sensor data was first acquired. Then, 12 hours later, the body odor level had decreased to 30.

[0043] When odor sensor data is input, the duration prediction model 100 outputs the time from when the odor sensor data is input until the body odor level reaches a predetermined small value as a predicted value of the duration. The predetermined small value is referred to as the body odor tolerance value.

[0044] Here, the body odor tolerance is set to 30. In other words, the duration is set to the time until the body odor level reaches 30.

[0045] 5 is a diagram showing an example of multiple data sets used by the model generation unit 211 for machine learning. In this figure, the odor sensor 1 is assumed to have four sensor elements. Each of odor sensor data #1, odor sensor data #2, odor sensor data #3, and odor sensor data #4 is a detection value obtained by a different one of the four sensor elements.

[0046] FIG. 5 shows four data sets acquired on a certain day. Specifically, data set DS1 acquired at 12:00, data set DS2 acquired at 15:00, data set DS3 acquired at 18:00, and data set DS4 acquired at 21:00 are shown. Each data set includes the acquisition date and time, the detection value for each sensor element, and the body odor level. Note that in this figure, the date is omitted from the data of the acquisition date and time.

[0047] According to these four data sets, odor sensor data #1 decreases from 12:00 to 21:00, such as 100, 70, 50, and 30. Odor sensor data #2 decreases from 12:00 to 21:00, such as 60, 50, 40, and 30. Odor sensor data #3 decreases from 12:00 to 21:00, such as 40, 30, 25, and 20. Odor sensor data #4 remains constant at 20 from 12:00 to 21:00. Body odor levels decrease from 12:00 to 21:00, such as 80, 50, 40, and 30.

[0048] The model generating unit 211 generates a duration prediction model 100 that indicates a relationship that enables the duration to be output when the odor sensor data and the acquisition date and time are input, based on these data sets DS1 to DS4.

[0049] The machine learning algorithm used by the model generation unit 211 is not limited to a specific algorithm.

[0050] 6 is a diagram showing the relationship between the duration obtained by actually measuring the time and the predicted value of the duration by the duration prediction model 100. It can be seen that the predicted value of the duration by the duration prediction model 100 is very close to the duration used in the machine learning. In other words, the duration prediction model 100 can predict the duration with high accuracy.

[0051] Fig. 7 is a diagram for explaining an example of duration prediction by the duration prediction model 100. For reference, this diagram shows four data sets DS1 to DS4 used in machine learning and the predicted values ​​of duration shown in Fig. 6.

[0052] After the duration prediction model 100 is generated, when new odor sensor data DN is input, the duration prediction unit 213 calculates the duration corresponding to the input odor sensor data DN and its acquisition date and time using the duration prediction model 100. According to the example shown in Fig. 7, when a set of 30, 40, 30, and 20, which are the measurement values ​​of odor sensor data #1 to #4, is input as odor sensor data DN, a predicted duration value of 2.85 hours is obtained from the input odor sensor data DN.

[0053] Next, the operation of the duration prediction system 1000 according to the first embodiment will be described.

[0054] FIG. 8 is a flowchart showing a series of processes when a subject measures the odor of urine for machine learning.

[0055] When a subject urinates, the subject measures the odor of the urine using the odor sensor 1 (S101). Next, the subject has another person judge the body odor level of the subject, and inputs the body odor level obtained by the judgment (S102). The body odor level is input, for example, via the information terminal 3. Note that the method of inputting the body odor level is not limited to this.

[0056] The sensor controller 2 acquires odor sensor data from the odor sensor 1, acquires the body odor level from the information terminal 3, and acquires the current time as the acquisition date and time. Then, a data set including the odor sensor data, the acquisition date and time, and the body odor level is stored in a predetermined storage area in the sensor controller 2 (S103). Then, the series of processes for measuring the odor of urine is completed.

[0057] The subject performs the series of processes shown in Fig. 8 several times a day, so that the sensor controller 2 stores a plurality of data sets acquired at different times.

[0058] FIG. 9 is a flowchart showing the operation of the model generating unit 211.

[0059] The model generation unit 211 generates the duration prediction model 100 by machine learning using a plurality of data sets (S201). The model generation unit 211 stores the generated duration prediction model 100 in the storage unit 212 (S202). Then, the operation of the model generation unit 211 ends.

[0060] When making the sensor controller 2 predict the duration, the subject performs the following process.

[0061] When a subject urinates, the subject measures the urine odor using the odor sensor 1. The odor sensor data is input to the sensor controller 2. The subject requests the sensor controller 2 to predict the duration.

[0062] The request for prediction of duration is executed, for example, via the information terminal 3. The method for requesting prediction of duration is not limited to this.

[0063] When the duration prediction unit 213 receives the odor sensor data, its acquisition date and time, and a request for predicting the duration, it calculates a predicted value of the duration based on the odor sensor data and the duration prediction model 100.

[0064] The output unit 214 outputs the predicted value of the duration. The predicted value of the duration is output to the information terminal 3. The method of outputting the predicted value of the duration is not limited to this.

[0065] In the above, the description of the case where there are multiple subjects has been omitted. When there are multiple subjects, the acquisition of the data set, the generation of the duration prediction model 100, and the calculation of the predicted value of the duration may be performed individually for each subject. For this purpose, the data set and the duration prediction model 100 may be configured to include the ID of the subject.

[0066] The relationship between the odor sensor data obtained by the odor sensor 1 and the body odor level, the rate at which the body odor level decreases, and the like may differ from person to person. By acquiring a data set, generating the duration prediction model 100, and calculating the predicted value of the duration individually for each subject, it is possible to predict the duration using the duration prediction model 100 optimized for each subject. This improves the accuracy of the duration prediction.

[0067] It should be noted that the acquisition of the data set, the generation of the duration prediction model 100, and the calculation of the predicted value of the duration do not necessarily have to be performed individually for each subject.

[0068] In the above description, the sensor controller 2 generates the duration prediction model 100 and predicts the duration using the duration prediction model 100. The information processing device that executes the generation of the duration prediction model 100 and the prediction of the duration is not limited to the sensor controller 2. One or both of the generation of the duration prediction model 100 and the prediction of the duration may be executed in an information processing device other than the sensor controller 2.

[0069] In addition, the odor sensor used to obtain the data set used in machine learning and the odor sensor used to obtain the odor sensor data used during prediction do not necessarily need to be the same sensor.

[0070] As described above, according to the first embodiment, the sensor controller 2 includes a storage unit 212 and a model generation unit 211. The model generation unit 211 generates a duration prediction model capable of acquiring a predicted value of the duration of body odor from the measurement value by the odor sensor 1 through machine learning using a plurality of data sets each acquired at different times. Each of the plurality of data sets includes odor sensor data which is a measurement value of the odor of urine by the odor sensor 1, and the body odor level at the time when the odor of urine was measured. The model generation unit 211 stores the generated duration prediction model 100 in the storage unit 212.

[0071] Therefore, by using the duration prediction model 100, it becomes possible to predict the duration of body odor from odor sensor data.

[0072] Furthermore, according to the first embodiment, the sensor controller 2 includes a duration prediction unit 213 and an output unit 214. When odor sensor data is input, the duration prediction unit 213 calculates a predicted value of the body odor duration based on the input odor sensor data and the duration prediction model 100. The output unit 214 outputs the predicted value of the body odor duration.

[0073] Therefore, by using the duration prediction model 100, it becomes possible to predict the duration of body odor from odor sensor data.

[0074] Second embodiment In the second embodiment, differences from the first embodiment will be described, and descriptions of the same matters as in the first embodiment will be omitted or will be briefly described. Note that the sensor controller of the second embodiment will be referred to as sensor controller 2a.

[0075] FIG. 10 is a diagram showing an example of a data set input to the sensor controller 2a according to the second embodiment.

[0076] Fig. 10 shows details of the datasets DS5 to DS12. The datasets DS5 to DS8 contain meal information indicating "garlic dishes". The datasets DS9 to DS12 contain meal information indicating "sake, sashimi".

[0077] The meal information is information indicating the content of the meal that the subject has just eaten. For example, if the subject eats meals in the morning, afternoon, and evening, and the odor is measured immediately after waking up and before breakfast, the meal information is information indicating the content of the dinner of the previous day.

[0078] Thus, in the second embodiment, the data set for machine learning includes food information in addition to the odor sensor data, the acquisition date and time, and the body odor level. Based on these data sets, the model generation unit 211a generates the duration prediction model 100a, which is a trained model capable of obtaining a predicted value of the body odor duration from the odor sensor data and the food information.

[0079] FIG. 11 is a diagram showing a functional configuration of a sensor controller 2a according to the second embodiment.

[0080] The sensor controller 2a includes a model generating unit 211a, a storage unit 212, a duration predicting unit 213a, and an output unit 214. The model generating unit 211a, the duration predicting unit 213a, and the output unit 214 are realized by a processor. The storage unit 212 is realized by a volatile memory or a non-volatile memory.

[0081] The model generation unit 211a generates the duration prediction model 100a by machine learning using multiple data sets acquired at different times. Each of the multiple data sets includes odor sensor data, acquisition date and time, body odor level, and diet information. The model generation unit 211a stores the generated duration prediction model 100a in the storage unit 212.

[0082] When odor sensor data, its acquisition date and time, and meal information are newly input, the duration prediction unit 213a calculates a predicted value of the subject's body odor duration based on the newly input odor sensor data and meal information. The duration prediction unit 213a obtains a predicted value of the subject's body odor duration by using the duration prediction model 100a.

[0083] The output unit 214 outputs the predicted value of the duration obtained by the duration prediction unit 213a.

[0084] The strength of a subject's body odor and the duration of the subject's body odor may vary depending on the contents of their diet. By adding dietary information as data for machine learning and prediction, the accuracy of the duration prediction is improved.

[0085] Although some embodiments of the present invention have been described, these embodiments are presented as examples and are not intended to limit the scope of the invention. These novel embodiments can be implemented in various other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their modifications are included in the scope and spirit of the invention, and are included in the scope of the invention and its equivalents described in the claims. [Explanation of symbols]

[0086] 1 odor sensor, 2, 2a sensor controller, 3 information terminal, 11 tube, 100, 100a duration prediction model, 211, 211a model generation unit, 212 storage unit, 213, 213a duration prediction unit, 214 output unit, 1000 duration prediction system, 2000 toilet bowl, DN sensor data, DS1 to DS12 datasets.

Claims

1. A storage unit for a model capable of obtaining a predicted value of the duration of body odor; a model generation unit that generates the model by machine learning using a plurality of data sets acquired at a plurality of different measurement times, the data set including a data set that associates the detection value of the urine odor component detected by the odor sensor, the measurement time when the urine odor was measured, and the body odor level at the measurement time, and stores the model in the storage unit; An information processing device comprising:

2. The data set is a data set in which the detection value of a urine odor component, a measurement time when the urine odor was measured, a body odor level at the measurement time, and diet information are associated with each other. The information processing device according to claim 1 .

Images