Waste identification method, waste identification device, and waste identification program

An excreta identification method performed by a computer, the method comprising:

• acquiring sound data collected by a microphone arranged in a toilet;
• identifying which of defecation, urination, flatulating, and the generation of an environmental sound in the toilet has occurred from the acquired sound data by inputting as an input value the acquired sound data into an identification model that has been subjected to machine learning,
• in which sound data including any of a defecation sound, a urination sound, a flatulating sound, and the environmental sound in the toilet is input into the identification model during the machine learning as an input value, and
• in which an output value of the identification model during the machine learning identifies which of the defecation, the urination, the flatulating, and the generation of the environmental sound in the toilet has occurred;
• and
• outputting an identification result as an output value of the identification model, identifying which of the defecation, the urination, the flatulating, and the generation of the environmental sound in the toilet has occurred, wherein
• the identification model is subjected to machine learning in which sound data of the sound of water splashing on the microphone during the urination, the defecation, the flatulating, or the generation of the environmental sound is an input value input into the identification model, and in which the output value represents the state of the water splashing on the microphone, and
• in the identification operation, the identification model determines that water is splashed on the microphone by inputting the acquired sound data to the identification model.

An excreta identification device comprising:

• an acquisition unit that acquires sound data collected by a microphone arranged in a toilet;
• an identification unit that identifies which of defecation, urination, flatulating, and the generation of an environmental sound in the toilet has occurred from the acquired sound data by inputting as an input value the acquired sound data into an identification model that has been subjected to machine learning
• in which sound data including any of a defecation sound, a urination sound, a flatulating sound, and the environmental sound in the toilet is input into the identification model during the machine learning as an input value, and
• in which an output value of the identification model during the machine learning identifies which of the defecation, the urination, the flatulating, and the generation of the environmental sound in the toilet has occurred;
• and
• an output unit that outputs an identification result as an output value of the identification model, identifying which of the defecation, the urination, the flatulating, and the generation of the environmental sound in the toilet has occurred, wherein
• the identification model correlates the urination sound with the speed at which the urine is excreted, and classifies input urination sounds into plural sound levels, each sound level corresponding to an excretion speed of a urine stream, wherein the identification model is subjected to machine learning
• in which each of the plural sound levels of the urination sounds is input as an input value to the identification model during machine learning and each input urination sound level is correlated with a corresponding urination excretion speed, and
• in which an output value of identification model during machine learning is the excretion speed associated with a corresponding input urination sound level, and
• in the identifying operation the identification model identifies the sound level of the urination sound in the acquired sound data in the event a urination sound is present in the acquired sound data,
• the device further comprising:
• a measuring unit that measures the duration of continuous urination based on the duration of the urination sound;
• and
• a calculation unit that calculates a voided volume of urine voided during urination using the identified sound level of urination, the urination excretion speed corresponding to the identified sound level, and the measured duration.

A non-transitory computer readable recording medium storing an excreta identification program causing a computer to function to:

• acquire sound data collected by a microphone arranged in a toilet;
• identify which of defecation, urination, flatulating, and the generation of an environmental sound in the toilet has occurred from the acquired sound data by inputting as an input value the acquired sound data into an identification model that has been subjected to machine learning
• in which sound data including any of a defecation sound, a urination sound, a flatulating sound, and the environmental sound in the toilet is input into the identification model during the machine learning as an input value, and
• in which an output value of the identification model during the machine learning identifies which of the defecation, the urination, the flatulating, and the generation of the environmental sound in the toilet has occurred;
• and
• output an identification result as an output value of the identification model, identifying which of the defecation, the urination, the flatulating, and the generation of the environmental sound in the toilet has occurred, wherein
• the identification model correlates the urination sound with the speed at which the urine is excreted, and classifies input urination sounds into plural sound levels, each sound level corresponding to an excretion speed of a urine stream, wherein the identification model is subjected to machine learning
• in which each of the plural sound levels of the urination sounds is input as an input value to the identification model during machine learning and each input urination sound level is correlated with a corresponding urination excretion speed, and
• in which an output value of identification model during machine learning is the excretion speed associated with a corresponding input urination sound level, and
• in the identifying operation the identification model identifies the sound level of the urination sound in the acquired sound data in the event a urination sound is present in the acquired sound data,
• the program further causing the computer to function to:
• measure the duration of continuous urination based on the duration of the urination sound;
• and
• calculate a voided volume of urine voided during urination using the identified sound level of urination, the urination excretion speed corresponding to the identified sound level, and the measured duration.