Method for predicting pet behavior and smart glasses using the same

Abstract

A method for predicting pet behavior and smart glasses applying the same are disclosed. The method for predicting pet behavior includes the following steps: Obtaining temporal analysis data of a pet's image based on an image. Obtaining temporal analysis data of a pet's sound based on an audio source. Predicting at least one behavior the pet is about to perform using a Bayesian classifier prediction model based on the image and sound temporal analysis data. Displaying real-time information on a transparent display unit of a smart glasses based on the at least one behavior.

Claims

1. A method for predicting pet behavior, comprising: An image analysis unit obtains temporal analysis data of a pet based on an image, the temporal analysis data including a plurality of actions corresponding to a plurality of time points; an audio analysis unit obtains temporal analysis data of the pet based on an audio signal; a Bayesian classifier prediction model predicts at least one action that the pet is about to perform based on the temporal analysis data of the image and the temporal analysis data of the audio; and a transparent display unit displays real-time information on a transparent display unit of a smart glasses based on the at least one action; wherein after the image analysis unit obtains the actions based on the image, the Bayesian classifier prediction model predicts the at least one action that the pet is about to perform based on the actions and the temporal analysis data of the audio.

2. The method for predicting pet behavior as described in claim 1, wherein the image time-series analysis data includes at least one of the following: and a plurality of expressions corresponding to the plurality of time points.

3. The method for predicting pet behavior as described in claim 1, wherein the steps for obtaining the temporal data of the image include: A single frame captures elements to obtain a plurality of consecutive frames at a plurality of time points; A background removal element removes the background from each image. A pet recognition model identifies the pet in each frame; a feature point detection element detects multiple limb nodes and multiple facial feature points of the pet in each frame; an action recognition model identifies multiple actions at those time points based on the limb nodes; an expression recognition model identifies multiple expressions at those time points based on the facial feature points; and a frame timing linking element obtains the image timing analysis data based on the actions and expressions.

4. The method for predicting pet behavior as described in claim 1, wherein the sound temporal analysis data includes at least one of the following: a plurality of amplitudes corresponding to a plurality of time points, and a plurality of pitches corresponding to the plurality of time points.

5. The method for predicting pet behavior as described in claim 1, wherein the steps for obtaining the temporal analysis data of the sound include: A conversion element converts the audio signal into a plurality of spectra over a plurality of time intervals; a spectrum analysis element analyzes the spectra to obtain a plurality of amplitudes and a plurality of pitches over the time intervals; and an audio timing linking element obtains audio timing analysis data based on the amplitudes and pitches.

6. The method for predicting pet behavior as described in claim 1, wherein the time of the image time-series analysis data and the time of the sound time-series analysis data correspond to each other.

7. The method for predicting pet behavior as described in claim 1, wherein predicting the at least one behavior that the pet is about to perform is a prediction of a series of behaviors that the pet is about to perform.

8. A smart pair of glasses, comprising: An image capturing unit is used to acquire an image; an image analysis unit is connected to the image capturing unit, which is used to acquire temporal analysis data of a pet based on the image; an audio capturing unit is used to acquire an audio signal; an audio analysis unit is connected to the audio capturing unit, which is used to acquire temporal analysis data of the pet based on the audio signal; a Bayesian classifier prediction model is used to predict at least one behavior that the pet will perform based on the temporal analysis data of the image and the temporal analysis data of the audio; and a transparent display unit is used to display real-time information based on the at least one behavior; wherein after the image analysis unit acquires the actions based on the image, the Bayesian classifier prediction model predicts the at least one behavior that the pet will perform based on the actions and the temporal analysis data of the audio.

9. The smart glasses as described in claim 8, wherein the image analysis unit comprises: A frame-capturing element is used to obtain multiple consecutive frames at multiple points in time from the image; A background removal element is used to remove the background from each of the images; A pet recognition model for identifying the pet in each frame; a feature point detection element for detecting multiple limb nodes and multiple facial feature points of the pet in each image; a motion recognition model for recognizing multiple motions at certain time points based on the limb nodes; a facial expression recognition model for recognizing multiple facial expressions at certain time points based on the facial feature points; and a frame timing linking element for obtaining frame timing analysis data based on the motions and expressions.

10. The smart glasses as claimed in claim 9, wherein the sound analysis unit comprises: A conversion element for converting the audio signal into a plurality of spectra over a plurality of time intervals; a spectrum analysis element for analyzing the spectra to obtain a plurality of amplitudes and a plurality of pitches over the time intervals; and an audio timing linking element for obtaining audio timing analysis data based on the amplitudes and pitches.

Images