Intelligent closestool suitable for old people

Abstract

The intelligent closestool comprises a closestool cover and a closestool body, the closestool body comprises a feeding cavity, a discharging channel communicated with the feeding cavity, voice interaction equipment arranged at the top of the closestool body, a master control device arranged at the bottom of the closestool body, and a power source connecting wire arranged on one side of the back face of the closestool body; the intelligent closestool has the remarkable beneficial effects in multiple aspects, a plurality of problems in the prior art are practically solved, and the defect that a traditional intelligent closestool and an existing intelligent closestool are single in function is overcome in the health monitoring aspect. The non-contact body temperature monitoring unit is used for continuously monitoring the body temperature, the intelligent sound induction and disease prediction unit is used for analyzing sound characteristics such as cough and wheeze, and the urine detection device is used for detecting urine indexes, so that the comprehensive monitoring of the health condition of the old people is realized; the problem that disease signals are difficult to perceive due to lack of key physiological index monitoring is effectively solved, and the diagnosis and treatment delay risk is greatly reduced. And the disease early warning capability is qualitatively improved.

Description

Technical Field

[0001] The present invention relates to a smart toilet, and more particularly to a smart toilet suitable for the elderly. Background Art

[0002] With the accelerating global aging process, the convenience of life and health management of the elderly have become the focus of social attention. Toileting, as a high-frequency scenario in daily life, is not only a basic physiological need but also an important scenario for monitoring the health status of the elderly. However, traditional toilets and existing smart toilets have significant deficiencies in health monitoring and are difficult to meet the special needs of the elderly:

[0003] (1) Single health monitoring function: The health monitoring of existing smart toilets mostly focuses on basic data such as weight and body fat, lacking continuous monitoring of key physiological indicators such as body temperature and heart rate. Due to the decline in perception ability, the elderly often cannot detect early disease signals such as fever and abnormal heart rate in time, delaying the treatment opportunity.

[0004] (2) Lack of disease warning ability: The elderly often suffer from chronic diseases (such as respiratory diseases and cardiovascular diseases). The sound characteristics such as coughing and wheezing during toileting and changes in body temperature can reflect the development of the disease. However, traditional toilets do not have the function of sound collection and analysis, and cannot identify abnormal physiological signals through non-contact methods, resulting in potential health risks being difficult to be pre-warned in advance.

[0005] (3) Insufficient operation complexity and adaptability: Although some smart toilets have health monitoring functions, their operation interfaces are complex and the sensor contact design (such as the need to manually wear devices), which is extremely unfriendly to the elderly with inconvenient mobility and declining cognitive ability, and the actual utilization rate is low.

[0006] (4) Weak data integration and medical linkage: Existing health monitoring data mostly exist in isolation and do not form an effective linkage with the medical system or guardians. The health data of the elderly during toileting (such as abnormal body temperature and abnormal breathing sounds) cannot be transmitted in time, making it difficult to achieve early intervention in diseases. Summary of the Invention

[0007] The present invention provides a smart toilet suitable for the elderly to solve the above-mentioned existing technical problems.

[0008] The technical solution of the present invention is realized as follows:

[0009] A smart toilet suitable for the elderly includes a toilet lid and a toilet body. The body includes a chamber for receiving materials, a discharge channel communicating with the material receiving chamber, a voice interaction device provided on the top of the body, a total control device provided at the bottom of the body, and a power connection line provided on one side of the back of the body; wherein, the voice interaction device is wirelessly connected to the total control device;

[0010] The toilet seat also includes an annular toilet cover.

[0011] Preferably, one side of the discharge channel is connected to a detection port, the detection port is connected to a urine detection device, and a discharge port is provided below the urine detection device and is connected to the sewer.

[0012] Preferably, at the connection between the detection port and the discharge channel, a separation net with a mesh diameter less than 2 mm is provided.

[0013] Preferably, the voice interaction device includes a voice output port and a voice input port. The voice output port is a speaker, and the voice input port is a microphone.

[0014] Preferably, the annular toilet cover is internally provided with a temperature sensor and a sound collector, and the temperature sensor and the sound collector are wirelessly connected to the master control device.

[0015] Preferably, the master control device includes a control system, the non-contact body temperature monitoring unit, the intelligent sound induction and disease prediction unit, and the multi-modal health data fusion unit;

[0016] Among them, the non-contact body temperature monitoring unit is composed of a sensor module, a signal processing module, a data storage module, and a communication module. Among them,

[0017] The sensor module is composed of a data acquisition sub-module, a data filtering sub-module, and an environmental data correction sub-module. The data acquisition sub-module is wirelessly connected to the temperature sensor and the sound collector to collect the user's temperature data and sound data; the data verification sub-module verifies the validity of the collected original data, eliminates abnormal values outside the reasonable range, and prevents noise data from entering the subsequent processing process; the environmental data correction sub-module introduces a temperature correction model to calibrate the original body temperature value through the real-time collected environmental temperature data;

[0018] The signal processing module processes through a multi-level algorithm to achieve noise reduction, feature extraction, and anomaly detection of body temperature data. Specifically, it includes a data filtering sub-module. The data filtering sub-module adopts a cascaded filtering strategy to effectively smooth high-frequency noise while retaining the trend information of body temperature changes. Further, a second-order Butterworth low-pass filter is used with a cut-off frequency set to 0.1 Hz to filter out high-frequency interference caused by physiological activities such as breathing and heartbeat, highlighting the slow change characteristics of body temperature, and extracting key features from the filtered data; and achieving precise early warning through multi-level decision-making;

[0019] The communication module ensures the efficiency and security of data transmission through protocol design, error handling, and security mechanisms.

[0020] Preferably, the signal processing module further includes an abnormality detection sub-module. When the body temperature ≥ 37.5°C at any time, a fever warning is triggered. When the rapid temperature rise rate ≥ 0.5°C / 3 minutes, a rapid body temperature rise warning is triggered.

[0021] Preferably, the environmental data correction sub-module specifically introduces a temperature correction model to calibrate the original body temperature value through the ambient temperature data collected in real time. The calculation formula is:

[0022] T 补偿后 = T 原始 + k × (T 环境 - T 基准 )

[0023] where k is the compensation coefficient. Under normal temperature and pressure, T 基准 is set to 25 ± 0.5°C. This algorithm eliminates the influence of ambient temperature fluctuations on the measurement results and ensures that the accuracy of body temperature data is better than ±0.3°C.

[0024] Preferably, the disease prediction model module realizes disease prediction through multi-technology fusion based on the preprocessed audio signal and the extracted acoustic features. It is assumed that the voice features corresponding to different diseases conform to the mixture of multiple Gaussian distributions. Using the expectation-maximization algorithm, the acoustic features in the training data are processed to estimate the mean, covariance, and weight of each Gaussian distribution. For the newly collected voice features, calculate their probabilities in each Gaussian distribution, and the disease category corresponding to the distribution with the highest probability is the prediction result.

[0025] The present invention has many significant beneficial effects and effectively solves many problems in the prior art. At the level of health monitoring, it makes up for the deficiencies of traditional and existing intelligent toilets with single functions. Through a non-contact body temperature monitoring unit, it continuously monitors body temperature, and cooperates with an intelligent sound sensing and disease prediction unit to analyze sound characteristics such as coughing and wheezing, as well as a urine detection device to detect urine indicators, realizing all-round monitoring of the health status of the elderly. It effectively solves the problem that it is difficult to detect disease signals due to the lack of monitoring of key physiological indicators, and greatly reduces the risk of medical treatment delay. The disease warning ability has been qualitatively improved. By using the disease prediction model module to deeply analyze sound characteristics and combining body temperature and urine data, abnormal physiological signals can be identified in advance. For example, through technologies such as Gaussian mixture models, the association between sound characteristics and diseases can be accurately judged, and potential risks of common chronic diseases of the elderly such as respiratory diseases and cardiovascular diseases can be warned in advance, changing the current situation that traditional toilets cannot warn of diseases through non-contact methods. The operation adaptability is excellent. The setting of the voice interaction device simplifies the operation process, avoids the troubles brought to the elderly by complex operation interfaces and contact sensors, improves the actual utilization rate of the product, and fully considers the characteristics of the elderly with inconvenient mobility and declining cognitive ability. The data integration and medical linkage are close. The multi-modal health data fusion unit integrates various health data, constructs a fusion model through machine learning algorithms, and generates a comprehensive health score. Once a disease risk is judged, it immediately warns the elderly through the voice interaction device and sends the data to the guardian's mobile phone APP, realizing effective linkage with the medical system and providing strong support for the early intervention of diseases, solving the problem that existing health monitoring data is isolated and difficult to be effectively applied. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 FIG. 6 is a perspective view of an intelligent toilet suitable for the elderly according to the present invention.

[0027] Figure 2 FIG. 10 is a schematic structural diagram of an intelligent toilet suitable for the elderly according to the present invention.

[0028] In the figure, 1 - main body; 2 - toilet lid; 3 - total control device; 4 - urine detection device; 5 - feeding chamber; 6 - discharge channel; 7 - voice interaction device; 8 - partition net. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0029] The drawings are only for illustrative purposes and should not be construed as a limitation of this patent;

[0030] For better illustration of this embodiment, some components in the drawings are omitted, enlarged or reduced, and do not represent the size of the actual product;

[0031] For those skilled in the art, it is understandable that some well-known structures and their descriptions in the drawings may be omitted.

[0032] The technical solution of the present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0033] Embodiment 1

[0034] As Figure 1-2 shown, an intelligent toilet applicable to the elderly according to the present invention includes a toilet lid and a toilet body. The body includes a feeding chamber, a discharge channel communicating with the feeding chamber, a voice interaction device provided on the top of the body, a master control device provided on the bottom of the body, and a power connection line provided on one side of the back of the body;

[0035] Among them, the voice interaction device is connected to the master control device through a wireless signal; the toilet lid further includes an annular toilet seat.

[0036] Preferably, a detection port is communicated with one side of the discharge channel, the detection port is communicated with a urine detection device, and a discharge port is provided below the urine detection device and communicated with the sewer.

[0037] Preferably, a partition net with a mesh diameter less than 2 mm is provided at the connection between the detection port and the discharge channel.

[0038] Preferably, the voice interaction device includes a voice output port and a voice input port. The voice output port is a speaker, and the voice input port is a microphone.

[0039] Preferably, the annular toilet seat is internally provided with a temperature sensor and a sound collector, and the temperature sensor and the sound collector are wirelessly connected to the master control device.

[0040] Preferably, the master control device includes a control system, a non-contact body temperature monitoring unit, an intelligent sound sensing and disease prediction unit, and a multi-modal health data fusion unit;

[0041] Among them, the non-contact body temperature monitoring unit is composed of a sensor module, a signal processing module, a data storage module and a communication module. Among them,

[0042] The sensor module is composed of a data acquisition sub-module, a data filtering sub-module and an environmental data correction sub-module. The data acquisition sub-module is wirelessly connected to the temperature sensor and the sound collector to collect the user's temperature data and sound data; the data verification sub-module verifies the validity of the collected original data, eliminates abnormal values beyond a reasonable range, and avoids noise data from entering the subsequent processing process; the environmental data correction sub-module introduces a temperature correction model to calibrate the original body temperature value through the real-time collected environmental temperature data;

[0043] The described signal processing module processes through a multi-stage algorithm to achieve noise reduction, feature extraction, and anomaly detection of body temperature data. Specifically, it includes a data filtering sub-module that effectively smooths high-frequency noise using a cascaded filtering strategy while retaining the trend information of body temperature changes. Further, a second-order Butterworth low-pass filter with a cut-off frequency set to 0.1 Hz is used to filter out high-frequency interference caused by physiological activities such as breathing and heartbeat, highlighting the slow change characteristics of body temperature, and extracting key features from the filtered data; and precise early warning is achieved through multi-layer decision-making;

[0044] The described communication module ensures the efficiency and security of data transmission through protocol design, error handling, and security mechanisms.

[0045] The signal processing module further includes an anomaly detection sub-module. When the body temperature ≥ 37.5°C at any time, a fever early warning is triggered. When the high temperature rise rate ≥ 0.5°C / 3 minutes, a rapid body temperature increase early warning is triggered.

[0046] The environmental data correction sub-module specifically introduces a temperature correction model to calibrate the original body temperature value through real-time collected environmental temperature data. The calculation formula is:

[0047] T 补偿后 =T 原始 +k×(T 环境 -T 基准 )

[0048] Where k is the compensation coefficient. Under normal temperature and pressure, T 基准 is set to 25 ± 0.5°C. This algorithm eliminates the influence of environmental temperature fluctuations on the measurement results and ensures that the body temperature data accuracy is better than ±0.3°C.

[0049] Preferably, the intelligent sound sensing and disease prediction unit consists of an audio signal preprocessing module, an acoustic feature extraction module, and a disease prediction model module. Among them, the audio signal preprocessing module uses an adaptive filtering algorithm based on recursive least squares (RLS) to estimate the environmental noise in real time and separate the target sound from the mixed signal. Combining the double-threshold method of short-time energy and zero-crossing rate, it automatically identifies the start and end positions of the sound signal, and the threshold parameter is adaptively determined by the Otsu method to avoid manual intervention and improve the robustness in different scenarios;

[0050] The described acoustic feature extraction module includes converting the sound signal into a frequency-domain representation, extracting features reflecting frequency distribution and energy concentration, especially collecting the numerical changes of sounds of daily users throughout the seasons and time periods of the day. Such numerical changes of sounds are normal changes, and the sound numerical values include the numerical values of timbre and volume. Numerical values exceeding the normal changes are extracted and marked as abnormal, and the user's sounds during this period are collected at a high frequency. The average value of the changing numerical values is taken as the final output determination numerical value;

[0051] The described disease prediction model module realizes disease prediction through multi-technology fusion based on the preprocessed audio signal and the extracted acoustic features. It is assumed that the sound features corresponding to different diseases conform to the mixture of multiple Gaussian distributions. Using the Expectation-Maximization (EM) algorithm, the acoustic features in the training data are processed to estimate the mean, covariance, and weight of each Gaussian distribution. For the newly collected sound features, calculate their probabilities in each Gaussian distribution, and the disease category corresponding to the distribution with the highest probability is the prediction result. For example, when analyzing the sound features of cough, if the probability of a certain feature vector in the Gaussian distribution representing pneumonia is significantly higher than other distributions, it is predicted that pneumonia may be present. As time goes by and data accumulates continuously, new sound data and corresponding disease labels are regularly collected, and the incremental learning algorithm is used to update the model so that the model can adapt to the changes in sound features of different seasons and different individuals, continuously improving the prediction performance. For example, during the high-incidence season of influenza, more sound data of influenza patients are collected to update the model to better identify the sound features related to influenza.

[0052] The described multi-modal health data fusion unit provides a more comprehensive and accurate health assessment of the elderly by integrating health data from different monitoring modules. Connect the body temperature data to obtain the body temperature data from the non-contact body temperature monitoring unit in real time, including the original body temperature measurement value, the body temperature value corrected by the environmental data, and the body temperature change trend information. These data can intuitively reflect the fever situation of the elderly and are an important basis for judging whether there is inflammation or infection in the body. Integrate the sound data; the cough, wheezing and other sound feature data output by the intelligent sound sensing and disease prediction unit are connected to this unit. For example, the time-domain features of sound (such as short-time energy, zero-crossing rate), frequency-domain features (spectral centroid, spectral kurtosis), and time-frequency domain features (Mel Frequency Cepstral Coefficient MFCC), etc., can be used to analyze the respiratory health status, such as whether there are diseases such as colds and pneumonia;

[0053] Urine data collection. The urine analysis data collected by the urine detection device, including indicators such as urine protein, urine sugar, and occult blood, is of great significance for monitoring the kidney function, diabetes condition, and urinary system diseases of the elderly. These data are connected to the multi-modal health data fusion unit in real time through the connection of the detection port to the urine detection device.

[0054] The multimodal health data fusion unit assigns weights to each data modality according to the contribution degree of each modality data to different diseases or health conditions. For example, when judging respiratory diseases, voice feature data may have a higher weight, while when evaluating kidney function, urine test data has a higher weight. The multimodal data is fused into a comprehensive feature vector by weighted summation. The formula is: F = w1D1 + w2D2 + w3D3, where F is the fused feature vector, D1, D2, and D3 represent body temperature data, voice data, and urine data respectively, w1, w2, and w3 are the corresponding weights, and w1 + w2 + w3 = 1.

[0055] The multimodal health data fusion unit uses machine learning algorithms to build a fusion model, such as using the random forest algorithm. The multimodal data is used as input features, and disease labels (such as healthy, cold, pneumonia, diabetes, etc.) are used as output labels to train the random forest model. During the training process, the model automatically learns the associations and patterns between each modality data, so as to achieve more accurate disease prediction and health assessment. For example, through learning, it is found that when the body temperature rises, the voice feature shows abnormal coughing, and the urine sugar index in the urine increases at the same time, it may indicate diabetes complicated with infection.

[0056] The multimodal health data fusion unit generates a comprehensive health score for the elderly according to the fused data and the model output. This score comprehensively considers various health information such as body temperature, voice, and urine, and reflects the overall health status of the elderly in an intuitive way. For example, the health score is on a 100-point scale. A score of 80 or above indicates good health, 60 - 80 indicates potential health risks, and below 60 requires timely attention and further examination.

[0057] When the fused data and the model determine that the elderly have a disease risk, the warning mechanism is triggered. For example, if the body temperature continues to be higher than 37.5°C, the voice feature shows a cough pattern related to pneumonia, and the urine protein index in the urine increases abnormally, the system will send a warning to the elderly through the voice interaction device and send the health data to the guardian's mobile APP to remind them to seek medical attention in time.

[0058] The fused health data and the health assessment results are stored in the local master control device and uploaded to the cloud server regularly for long-term tracking of the elderly's health status and data retrospective analysis. Encryption technology is used for data storage to ensure the privacy and security of the elderly.

[0059] Optimize the fusion model and algorithm based on actual medical diagnosis results and changes in the health status of the elderly. For example, if it is found that the model misjudges a certain disease, the model can be retrained and adjusted by adding more accurate case data to continuously improve the performance and accuracy of the multi-modal health data fusion unit.

[0060] The voice interaction device of the present invention plays a crucial role in many aspects, greatly improving the convenience and functionality of the product.

[0061] From an operational perspective, it significantly reduces the usage threshold. Due to the inconvenient mobility and decreased cognitive ability of the elderly, a complex operation interface will become an obstacle to using a smart toilet. The voice interaction device provides a simple and direct operation method. The elderly only need to say corresponding instructions, such as "open the toilet lid", "adjust the water temperature", "activate the flushing function", etc., and the smart toilet can quickly respond and execute the operation, allowing the elderly to easily complete various toilet-related operations without manually operating complex buttons or touchscreens, improving the usage experience.

[0062] In terms of health monitoring and disease warning, the voice interaction device also plays an important role. When various monitoring units of the smart toilet, such as the non-contact body temperature monitoring unit, the intelligent sound sensing and disease prediction unit, and the multi-modal health data fusion unit, detect abnormal data and determine that the elderly are at risk of disease, the voice interaction device can immediately send a warning message to the elderly. For example, when the system monitors that the elderly's body temperature continuously exceeds 37.5°C and the voice characteristics show a cough pattern related to pneumonia, the voice interaction device will send a voice prompt: "Your health condition may be abnormal. Your body temperature is on the high side and the cough characteristics are suspected of pneumonia. Please pay attention in time or contact your family." This intuitive voice reminder method ensures that the elderly can timely understand their own health status.

[0063] In addition, the voice interaction device can also provide information query services. The elderly can ask about their own health data by voice, such as "What is my body temperature today?" "Are my urine test indicators normal in the last week?" etc. The smart toilet will feedback the corresponding health data information through the voice interaction device, facilitating the elderly to keep track of their own health dynamics at any time. At the same time, the device can also provide some popular science knowledge about health. For example, when the elderly ask "What are the possible causes of coughing?", the device will give the popular science content related to the disease, enhancing the elderly's understanding of health problems.

[0064] The terms describing the positional relationship in the drawings are for illustrative purposes only and should not be construed as a limitation of this patent;

[0065] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, rather than limitations on the implementation manners of the present invention. For those of ordinary skill in the art, based on the above description, other different forms of changes or modifications can be made. It is not necessary and impossible to enumerate all the implementation manners here. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the claims of the present invention.

Claims

1. An intelligent toilet applicable to the elderly, comprising a toilet lid and a toilet body, characterized in that, The body includes a chamber for feeding, a discharge channel communicating with the feeding chamber, a voice interaction device provided at the top of the body, a master control device provided at the bottom of the body, and a power connection line provided on one side of the back of the body; Among them, the voice interaction device is wirelessly connected to the master control device; The toilet cover further includes an annular toilet seat.

2. The intelligent toilet applicable to the elderly according to claim 1, wherein One side of the discharge channel is communicated with a detection port, the detection port is communicated with a urine detection device, and a discharge port is provided below the urine detection device, which is communicated with the sewer.

3. The intelligent toilet applicable to the elderly according to claim 1, wherein At the connection between the detection port and the discharge channel, a partition net with a mesh diameter less than 2 mm is provided.

4. The intelligent toilet applicable to the elderly according to claim 1, characterized in that, The voice interaction device includes a voice output port and a voice input port. The voice output port is a speaker, and the voice input port is a microphone.

5. The intelligent toilet applicable to the elderly according to claim 1, characterized in that, The annular toilet seat is internally provided with a temperature sensor and a sound collector, and the temperature sensor and the sound collector are wirelessly connected to the master control device.

6. The intelligent toilet applicable to the elderly according to claim 1, characterized in that, The master control device includes a control system, the non-contact body temperature monitoring unit, the intelligent sound sensing and disease prediction unit, and the multi-modal health data fusion unit; Among them, the non-contact body temperature monitoring unit is composed of a sensor module, a signal processing module, a data storage module and a communication module. Among them, The sensor module is composed of a data acquisition sub-module, a data filtering sub-module and an environmental data correction sub-module. The data acquisition sub-module is wirelessly connected to the temperature sensor and the sound collector to collect the user's temperature data and sound data; the data verification sub-module verifies the validity of the collected raw data, eliminates abnormal values outside the reasonable range, and prevents noise data from entering the subsequent processing process; the environmental data correction sub-module introduces a temperature correction model to calibrate the original body temperature value through the real-time collected environmental temperature data; The signal processing module processes through a multi-stage algorithm to achieve noise reduction, feature extraction and abnormal detection of body temperature data. Specifically, it includes a data filtering sub-module. The data filtering sub-module adopts a cascaded filtering strategy to effectively smooth high-frequency noise while retaining the trend information of body temperature changes. Further, a second-order Butterworth low-pass filter is used, and the cut-off frequency is set to 0.1 Hz to filter out high-frequency interference caused by physiological activities such as breathing and heartbeat, highlighting the slow change characteristics of body temperature, and extracting key features from the filtered data; and precise early warning is achieved through multi-layer decision-making; The communication module ensures the efficiency and security of data transmission through protocol design, error handling and security mechanisms.

7. The intelligent toilet applicable to the elderly according to claim 6, characterized in that, The signal processing module further includes an abnormal detection sub-module. When the body temperature ≥ 37.5 °C at any time within the abnormal detection sub-module, a fever early warning is triggered. When the temperature rise rate ≥ 0.5 °C / 3 minutes, a rapid body temperature rise early warning is triggered.

8. The intelligent toilet applicable to the elderly according to claim 6, characterized in that, The environmental data correction sub-module specifically introduces a temperature correction model to calibrate the original body temperature value through the real-time collected environmental temperature data. The calculation formula is: T 补偿后 = T 原始 + k × (T 环境 T 基准 ) where k is the compensation coefficient, and at normal temperature and pressure, T 基准 is set to 25 ± 0.5 °C. This algorithm eliminates the influence of environmental temperature fluctuations on the measurement results and ensures that the body temperature data accuracy is better than ±0.3 °C.

9. The intelligent toilet applicable to the elderly according to claim 6, characterized in that, The disease prediction model module realizes disease prediction through multi-technology fusion based on the pre-processed audio signal and the extracted acoustic features. It is assumed that the voice features corresponding to different diseases conform to the mixture of multiple Gaussian distributions.

10. The intelligent toilet applicable to the elderly according to claim 9, characterized in that, The disease prediction model module uses the Expectation-Maximization algorithm to process the acoustic features in the training data, estimate the mean, covariance, and weight of each Gaussian distribution, and for the newly collected voice features, calculate their probabilities in each Gaussian distribution. The disease category corresponding to the distribution with the highest probability is the prediction result.

Images