CN115050095A - Human body posture prediction method based on Gaussian process regression and progressive filtering - Google Patents
Human body posture prediction method based on Gaussian process regression and progressive filtering Download PDFInfo
- Publication number
- CN115050095A CN115050095A CN202210628615.3A CN202210628615A CN115050095A CN 115050095 A CN115050095 A CN 115050095A CN 202210628615 A CN202210628615 A CN 202210628615A CN 115050095 A CN115050095 A CN 115050095A
- Authority
- CN
- China
- Prior art keywords
- human body
- position state
- joint point
- gaussian process
- process regression
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/20—Movements or behaviour, e.g. gesture recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/766—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using regression, e.g. by projecting features on hyperplanes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/103—Static body considered as a whole, e.g. static pedestrian or occupant recognition
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Human Computer Interaction (AREA)
- Psychiatry (AREA)
- Social Psychology (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to the field of human body posture prediction, in particular to a human body posture prediction method based on Gaussian process regression and progressive filtering, wherein a state transition model is established by utilizing the Gaussian process regression method, and the reliability of measured information is checked by using a predicted value so as to improve the accuracy of prior information; estimating the position state based on a progressive filtering method to achieve faster and more accurate performance; and finally, predicting the position states of all the joint points of the human body. The invention has the beneficial effects that: compared with the existing human body posture prediction, the method is based on Gaussian process regression and progressive filtering, the accuracy of state prediction is effectively improved, and the robustness of adverse factors such as errors of the sensor is improved.
Description
Technical Field
The invention relates to the field of human body posture prediction, in particular to a human body posture prediction method based on Gaussian process regression and progressive filtering.
Background
The human body posture prediction has important application value in the fields of human-computer interaction, behavior recognition, video monitoring, sports games and the like. In particular, with the development of artificial intelligence, 3D vision and other technologies, human posture prediction technology is applied in the fields of industry, agriculture, sports and the like, for example, for collecting athlete motion information, and combining human physiology, physics and artificial intelligence methods, thereby forming intelligent human motion prediction.
Due to the complexity and randomness of human motion, it is difficult for conventional modeling methods to accurately describe the motion of the human body. With the development of AI technologies such as deep learning, a plurality of effective methods such as a recurrent neural network and Gaussian process regression are provided for human motion prediction. On the other hand, the visual sensor is a sensor commonly used in the field of human body posture prediction at present, however, the visual sensor is easily affected by factors such as ambient light, shading and the like, and uncertainty of visual human body posture detection is caused. In turn, it is difficult for the conventional kalman filtering method to obtain a satisfactory noise suppression effect. Because the human body posture prediction model, the visual posture detection and other aspects have uncertainty, the human body posture prediction precision is greatly reduced. At present, no technology exists for robustly and accurately solving the problem of human body posture prediction in complex scenes and motion.
For example, chinese patent CN113569627A, published 2021, 10 months and 29 days, discloses a human body posture prediction model training method, which comprises: acquiring a labeled training set and an unlabeled training set, wherein the labeled training set comprises a plurality of first human body images containing labeled data, and the labeled data is used for representing real posture information in the first human body images; inputting the first human body image into a generator in a human body posture prediction model to obtain a corresponding first human body posture prediction result, and calculating a first loss value of the generator according to the labeling data and the first human body posture prediction result; inputting the second human body image into a generator to obtain a corresponding second human body posture prediction result; calculating a second loss value corresponding to the discriminator in the human body posture prediction model according to the first human body image, the annotation data, the second human body image and the second human body posture prediction result; and optimizing the generator and the discriminator according to the first loss value and the second loss value to obtain a human body posture prediction model. According to the technical scheme, the human body posture is predicted according to the image, a large number of human body image labeling samples need to be prepared in advance, the sample labeling is time-consuming and labor-consuming, and the accuracy is limited.
Disclosure of Invention
In order to overcome the defects of uncertainty of measurement information caused by illumination sensitive change of a sensor and the like, difficulty in modeling of human body movement and the like, the invention provides a human body posture prediction method based on Gaussian process regression and progressive filtering.
The technical scheme adopted by the invention is as follows: a human body posture prediction method based on Gaussian process regression and progressive filtering comprises the following steps:
step 1) using the collected three-dimensional position states and corresponding postures of each human body joint point as a training data set, and respectively establishing a position state transfer model of each human body joint point by utilizing Gaussian process regression;
step 2) determining the initial position state of the jth personal body joint pointAnd initial covariance,,The number of the monitored human body joint points is obtained;
step 3) establishing an observation model to obtain a measurement value of the sensor at the moment k to the jth personal body joint pointAnd zero mean and covariance ofMeasurement noise of;
Step 4) reading the measurement information of the sensor to obtain the position state of the j-th personal body joint point at the k momentMeasured value ofBased on the measured value of the sensor readAnd the predicted value of the position state output by the observation model is the prior estimated position state, and the measured value at the moment k is judgedIf the position is not in the confidence interval of the predicted value, the step 5) is carried out, and if the position is not in the confidence interval, the measured value is replaced by the priori estimated position stateRe-executing step 4);
step 5) progressive introduction of measurement valuesCorrecting the current position state to obtain the position stateEstimated value and covariance ofJudging whether the over-estimation condition occurs or not, if the over-estimation condition does not occur, continuing to correct, otherwise, skippingStep 5) is carried out;
step 6) using the position state transition model to obtain the position state of the j-th personal body joint point at the current moment according to the step 5)And its covariancePredicting the position state of the jth individual's body joint point at the next timeAnd its covariance;
And 7) repeating the steps 4) to 6) for each human body joint point to finish posture prediction.
Preferably, in step 1), the training data set is collectedIn whichIn order to input a set of vectors,in order to output the set of vectors,indicating the amount of data, and the new input is recorded as the position state at the k-th timeSelecting a predetermined kernel functionUsing Gaussian process regression to establishThe position state transition model of each joint point of the human body is as follows:
wherein the content of the first and second substances,representing the state transition function, k =1,2, … being a discrete time sequence,is the position state of the human body joint point,、andcoordinate values of the human body joint point at the moment of k +1 on an x axis, a y axis and a z axis respectively,、andthe speeds of the human body joint points at the moment k on the x axis, the y axis and the z axis respectively,zero mean and covariance ofGaussian noise.
Preferably, said bits are used in step 6)Setting state transition model for predicting position state of each joint point of human body at next momentAnd its varianceThe method comprises the following steps:and。
preferably, in step 3), the observation model is established as follows:
wherein the content of the first and second substances,,the measurement value of the sensor for the joint point j at the time k is shown.
Preferably, in step 5), the measurement values are introduced progressivelyThe method for correcting the current position state comprises the following steps:
step 51) setting iteration parametersSetting the initial value to be 0, setting the maximum value of the gradual lead-in measurement updating step number to be N, and setting the iteration step lengthIteration parameter,Representing the number of iterations;
step 52) calculatingAt the time of the second iteration, the position state measurement value of the j-th human joint point at the k momentPredicted value, covariance ofAnd covariance:
Wherein the content of the first and second substances,andrespectively corresponding sigma points and weight values;
step 53) measuring value of position state of j-th individual body joint point at time kIn the first placeSecondary filter gain:
Step 54) calculating the position state of the jth personal body joint point at the time kAnd its covarianceIn the first placeThe value of the sub-iteration:
step 55) judging whether the over-estimation condition occurs or not, if not, judging that the over-estimation condition occursAfter adding 1, the correction is continued from step 52), otherwise, step 6) is entered.
Preferably, in step 55), the method for determining whether the over-estimation condition occurs includes:
order toIf, ifAnd is provided withIf not, judging that the overestimation condition does not occur, otherwise, judging that the overestimation condition occurs.
Preferably, the monitored joints of the human body are one or more of a nodding joint, a thoracic joint, a shoulder joint, an elbow joint, a wrist joint, a sacral joint, a hip joint, a knee joint, and an ankle joint.
Preferably, the position state and the measurement value both refer to three-dimensional coordinate values in a rectangular coordinate system.
Preferably, the Gaussian noise in step 2)Is zero mean and covariance ofThe noise of the gaussian noise of (a),
preferably, in step 4), the predicted position state of the observation model is a position state of the joint point j predicted by the observation model created by the gaussian process regression methodAnd its covariance。
The invention has the beneficial effects that: compared with the existing human body posture prediction, the method is based on Gaussian process regression and progressive filtering, the accuracy of state prediction is effectively improved, and the robustness of adverse factors such as errors of the sensor is improved.
Other features and advantages of the present invention will be disclosed in more detail in the following detailed description of the invention and the accompanying drawings.
Drawings
The invention is further described below with reference to the accompanying drawings:
fig. 1 is a schematic flow chart of a human body posture prediction method according to an embodiment of the present invention.
Detailed Description
The technical solutions of the embodiments of the present invention are explained and illustrated below with reference to the drawings of the embodiments of the present invention, but the following embodiments are only preferred embodiments of the present invention, and not all embodiments. Based on the embodiments in the implementation, other embodiments obtained by those skilled in the art without any creative effort belong to the protection scope of the present invention.
In the following description, the appearances of the indicating orientation or positional relationship such as the terms "inner", "outer", "upper", "lower", "left", "right", etc. are only for convenience in describing the embodiments and for simplicity in description, and do not indicate or imply that the device or element being referred to must have a particular orientation, be constructed and operated in a particular orientation, and are not to be construed as limiting the present invention.
Example (b):
a method for predicting a human body posture based on gaussian process regression and progressive filtering, please refer to fig. 1, which includes:
step 1) using the collected three-dimensional position states and corresponding postures of each human body joint point as a training data set, and respectively establishing a position state transfer model of each human body joint point by utilizing Gaussian process regression;
step 2) determining the initial position state of the jth personal body joint pointAnd initial covariance,,For the number of joint points of the body being monitored, Gaussian noiseIs zero mean and covariance ofGaussian noise of (2);
step 3) establishing an observation model to obtain a measurement value of the sensor at the moment k to the jth personal body joint pointAnd zero mean and covariance ofOf (2) measuring noise;
Step 4) reading the measurement information of the sensor to obtain the position state of the j-th personal body joint point at the time kMeasured value ofBased on the measured value of the sensor readAnd a predicted value of the position state output by the observation model, wherein the predicted value of the position state is the prior estimated position state, and the measured value at the moment k is judgedWhether the estimated position is within the confidence interval of the predicted value, if the estimated position is within the confidence interval, the step 5) is carried out, and if the estimated position is not within the confidence interval, the measured position is replaced by the priori estimated position stateValue ofRe-executing step 4);
step 5) progressive introduction of measurement valuesCorrecting the current position state to obtain the position stateEstimated value and covariance ofJudging whether an over-estimation condition occurs or not, if not, continuing to correct, otherwise, jumping out of the step 5);
step 6) using a position state transition model to obtain the position state of the jth personal body joint point at the current moment according to the step 5)And its covariancePredicting the position state of the jth individual's body joint point at the next timeAnd its covariance;
And 7) repeating the steps 4) to 6) for each human body joint point to finish posture prediction.
In this embodiment, one or more of the human joints being monitored are nodding, thoracic, shoulder, elbow, wrist, sacral, hip, knee, and ankle. The position state and the measurement value both refer to three-dimensional coordinate values in a rectangular coordinate system. The embodiment predicts the human body posture not the posture which the human body will but is not made at a certain moment in the future, but predicts the correct human body posture according to the measurement value of the included noise of the posture which the human body has made. In which the human body posture is to the position and posture of the human body part, not just the posture of the human body. Human body gestures play an important role in the fields of human-computer interaction, behavior recognition, video surveillance, sports games and the like.
In step 1), a training data set has been collectedWhereinIn order to input a set of vectors,in order to output the set of vectors,indicating the amount of data, and the new input is recorded as the position state at the k-th timeSelecting a predetermined kernel functionUsing Gaussian process regression to establishThe position state transition model of each joint point of the human body is as follows:
wherein, the first and the second end of the pipe are connected with each other,indicating a state transitionThe function, k =1,2, … is a discrete time series,is the position state of the human body joint point,、andcoordinate values of the human body joint point at the moment of k +1 on an x axis, a y axis and a z axis respectively,、andthe speeds of the human body joint points at the moment k on the x axis, the y axis and the z axis respectively,zero mean and covariance ofGaussian noise.
In step 3), the established observation model is as follows:
wherein the content of the first and second substances,,the measurement value of the joint point j by the sensor at the time k is shown.
In step 5), the measurement values are introduced progressivelyThe method for correcting the current position state comprises the following steps:
step 51) setting iteration parametersSetting the initial value to be 0, setting the maximum value of the gradual lead-in measurement updating step number to be N, and setting the iteration step lengthIteration parameter,Representing the number of iterations;
step 52) calculating the position state measurement value of the jth human body joint point at the k moment in the (m + 1) th iterationPredicted value, covariance ofAnd covariance:
Wherein the content of the first and second substances,andrespectively corresponding sigma points and weight values;
step 53) measuring value of position state of j-th individual body joint point at time kIn the first placeSecondary filter gain:
Step 54) calculating the position state of the jth personal body joint point at the time kAnd its covarianceIn the first placeThe value of the sub-iteration:
step 55) judging whether the over-estimation condition occurs or not, if not, judging that the over-estimation condition occursAfter adding 1, the correction is continued from step 52), otherwise, step 6) is entered.
In step 55), the method for judging whether the over-estimation condition occurs comprises the following steps:
order toIf, ifAnd isIf so, judging that the over-estimation condition does not occur, otherwise, judging that the over-estimation condition occurs.
Step 6), predicting the position state of each joint point of the human body at the next moment by using the position state transition modelAnd its varianceThe method comprises the following steps:and。
while the invention has been described with reference to specific embodiments thereof, it will be understood by those skilled in the art that the invention is not limited thereto, and may be embodied in many different forms without departing from the spirit and scope of the invention as set forth in the following claims. Any modification which does not depart from the functional and structural principles of the present invention is intended to be included within the scope of the claims.
Claims (10)
1. A human body posture prediction method based on Gaussian process regression and progressive filtering is characterized in that,
the method comprises the following steps:
step 1) using the collected three-dimensional position states and corresponding postures of each human body joint point as a training data set, and respectively establishing a position state transfer model of each human body joint point by utilizing Gaussian process regression;
step 2) determining the initial position state of the jth personal body joint pointAnd initial covariance,,The number of the monitored human body joint points is obtained;
step 3) establishing an observation model to obtain a measurement value of the sensor at the moment k to the jth personal body joint pointAnd zero mean and covariance ofOf (2) measuring noise;
Step 4) reading the measurement information of the sensor to obtain the position state of the j-th personal body joint point at the time kMeasured value ofBased on the read measured values of the sensorsAnd a predicted value of the position state output by the observation model, wherein the predicted value of the position state is the prior estimated position state, and the measured value at the moment k is judgedIf the position is not in the confidence interval of the predicted value, the step 5) is carried out, and if the position is not in the confidence interval, the measured value is replaced by the priori estimated position stateRe-executing step 4);
step 5) progressive introduction of measurement valuesCorrecting the current position state to obtain the position stateEstimated value and covariance ofJudging whether an over-estimation condition occurs or not, if not, continuing to correct, otherwise, jumping out of the step 5);
step 6) using the position state transition model to obtain the position state of the j-th personal body joint point at the current moment according to the step 5)And its covariancePredicting the position state of the jth individual's body joint point at the next timeAnd its covariance;
And 7) repeating the steps 4) to 6) for each human body joint point to finish posture prediction.
2. The human body posture prediction method based on Gaussian process regression and progressive filtering according to claim 1,
step 1), collected training data setWhereinIn order to input a set of vectors,in order to output the set of vectors,indicating the amount of data, and the new input is recorded as the position state at the k-th timeSelecting a predetermined kernel functionUsing Gaussian process regression to establishThe position state transition model of each joint point of the human body is as follows:
wherein the content of the first and second substances,representing the state transition function, k =1,2, … are discrete time sequences,is the position state of the human body joint point,、andcoordinate values of the human body joint point at the moment of k +1 on an x axis, a y axis and a z axis respectively,、andthe speeds of the human body joint points at the moment k on the x axis, the y axis and the z axis respectively,zero mean and covariance ofGaussian noise.
3. The human body posture prediction method based on Gaussian process regression and progressive filtering according to claim 2,
4. the human body posture prediction method based on Gaussian process regression and progressive filtering according to any one of claims 1 to 3,
in step 3), the established observation model is as follows:
5. The human body posture prediction method based on Gaussian process regression and progressive filtering according to any one of claims 1 to 3,
in step 5), the measurement values are introduced progressivelyThe method for correcting the current position state comprises the following steps:
step 51) setting iteration parametersSetting the initial value to be 0, setting the maximum value of the gradual lead-in measurement updating step number to be N, and setting the iteration step lengthIteration parameter,Representing the number of iterations;
step 52) calculatingAt the time of the second iteration, the position state measurement value of the jth individual body joint point at the k momentPredicted value, covariance ofAnd covariance:
Wherein the content of the first and second substances,andrespectively corresponding sigma points and weight values;
step 53) measuring value of position state of j-th individual body joint point at time kIn the first placeSecondary filter gain:
Step 54) calculating the position state of the j-th personal body joint point at the time kAnd its covarianceIn the first placeThe value of the sub-iteration:
6. The human body posture prediction method based on Gaussian process regression and progressive filtering according to claim 5,
in step 55), the method for determining whether the over-estimation condition occurs comprises the following steps:
7. The human body posture prediction method based on Gaussian process regression and progressive filtering according to any one of claims 1 to 3,
one or more of a human joint nod, thoracic, shoulder, elbow, wrist, sacral, hip, knee, and ankle that are monitored.
8. The human body posture prediction method based on Gaussian process regression and progressive filtering according to any one of claims 1 to 3,
and the position state and the measurement value both refer to three-dimensional coordinate values under a rectangular coordinate system.
10. The human body posture prediction method based on Gaussian process regression and progressive filtering according to any one of claims 1 to 3,
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210628615.3A CN115050095A (en) | 2022-06-06 | 2022-06-06 | Human body posture prediction method based on Gaussian process regression and progressive filtering |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210628615.3A CN115050095A (en) | 2022-06-06 | 2022-06-06 | Human body posture prediction method based on Gaussian process regression and progressive filtering |
Publications (1)
Publication Number | Publication Date |
---|---|
CN115050095A true CN115050095A (en) | 2022-09-13 |
Family
ID=83159469
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210628615.3A Pending CN115050095A (en) | 2022-06-06 | 2022-06-06 | Human body posture prediction method based on Gaussian process regression and progressive filtering |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115050095A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116627262A (en) * | 2023-07-26 | 2023-08-22 | 河北大学 | VR interactive device control method and system based on data processing |
-
2022
- 2022-06-06 CN CN202210628615.3A patent/CN115050095A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116627262A (en) * | 2023-07-26 | 2023-08-22 | 河北大学 | VR interactive device control method and system based on data processing |
CN116627262B (en) * | 2023-07-26 | 2023-10-13 | 河北大学 | VR interactive device control method and system based on data processing |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110070074B (en) | Method for constructing pedestrian detection model | |
CN110660083B (en) | Multi-target tracking method combined with video scene feature perception | |
CN109341703B (en) | Visual SLAM algorithm adopting CNNs characteristic detection in full period | |
CN113091738B (en) | Mobile robot map construction method based on visual inertial navigation fusion and related equipment | |
CN108961308B (en) | Residual error depth characteristic target tracking method for drift detection | |
CN111259735B (en) | Single-person attitude estimation method based on multi-stage prediction feature enhanced convolutional neural network | |
CN114676956A (en) | Old man's risk early warning system that tumbles based on multidimensional data fusion | |
Cheng et al. | A vision-based robot grasping system | |
CN114495267A (en) | Old people falling risk assessment method based on multi-dimensional data fusion | |
CN109409418A (en) | A kind of winding detection method based on bag of words | |
CN111462184B (en) | Online sparse prototype tracking method based on twin neural network linear representation model | |
CN109934849A (en) | Online multi-object tracking method based on track metric learning | |
CN115050095A (en) | Human body posture prediction method based on Gaussian process regression and progressive filtering | |
CN109544632B (en) | Semantic SLAM object association method based on hierarchical topic model | |
CN110163112B (en) | Examinee posture segmentation and smoothing method | |
CN112785630A (en) | Multi-target track exception handling method and system in microscopic operation | |
CN116524062A (en) | Diffusion model-based 2D human body posture estimation method | |
Mohedano et al. | Camera localization usingtrajectories and maps | |
CN112036291A (en) | Kinematic data model construction method based on motion big data and deep learning | |
CN112270357A (en) | VIO vision system and method | |
Yuan et al. | Robust instruments position estimation using improved kernelized correlation filter for substation patrol robots | |
CN113743341B (en) | Human body posture real-time estimation method based on self-adaptive model | |
Xie et al. | Research on Human Pose Capture Based on the Deep Learning Algorithm | |
Cui et al. | Mobile Big Data Analytics for Human Behavior Recognition in Wireless Sensor Network Based on Transfer Learning | |
CN115342817B (en) | Method, system, equipment and medium for monitoring track tracking state of unmanned tracked vehicle |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |