CN-121687383-B - Household environment cognition rehabilitation training action guiding system based on visual interaction

CN121687383BCN 121687383 BCN121687383 BCN 121687383BCN-121687383-B

Abstract

The invention relates to the technical field of rehabilitation training, in particular to a home environment cognitive rehabilitation training action guiding system based on visual interaction, which can realize the following steps of acquiring rehabilitation training images of a target patient in each unit training period in each preset rehabilitation training period and acquiring cognitive recovery scores corresponding to each preset rehabilitation training period through mutual coordination among a plurality of modules; the method comprises the steps of determining a target motion vector sequence and a target state value sequence, dividing intervals, clustering training action time periods, determining action proficiency indexes corresponding to each training action time period, and determining target recommended guide values corresponding to training actions represented by each target cluster. According to the invention, the target recommended guiding values corresponding to different training actions are quantized by analyzing the self condition of the target patient in the rehabilitation training process, so that the rationality of the recommendation of the cognitive rehabilitation training actions is improved, and the guiding effect of the cognitive rehabilitation training actions is further improved.

Inventors

DAI YANYAN
ZHAO QIONG
ZHANG XUAN
LI CHENGWEI

Assignees

贵州中医药大学第二附属医院

Dates

Publication Date: 20260508
Application Date: 20260210

Claims (6)

1. A home environment cognition rehabilitation training action guidance system based on visual interaction, the system comprising: the data acquisition module is used for acquiring rehabilitation training images of a target patient in each unit training period in each preset rehabilitation training period and acquiring cognitive recovery scores corresponding to each preset rehabilitation training period; The vector determining module is used for determining a target motion vector sequence and a target state value sequence corresponding to each unit training period based on all rehabilitation training images in each unit training period; the dividing and clustering module is used for dividing intervals of all preset rehabilitation training periods according to the target motion vector sequences and the target state value sequences corresponding to all unit training periods to obtain training action periods, and clustering all training action periods to obtain target clusters; The index determining module is used for determining an action proficiency index corresponding to each training action period according to the similarity between each training action period and other training action periods in the target cluster to which the training action period belongs; the guiding value determining module is used for determining a target recommended guiding value corresponding to the training action represented by each target cluster according to the duration corresponding to different training action time periods in each target cluster, the action proficiency index and the cognitive recovery score corresponding to the preset rehabilitation training period to which the action proficiency index belongs; The method comprises the steps of determining a target motion vector sequence and a target state value sequence corresponding to each unit training period, wherein the target motion vector sequence and the target state value sequence comprise the steps of determining any one preset rehabilitation training period as a marked rehabilitation training period, determining any one unit training period in the marked rehabilitation training period as a marked unit period, extracting limb key points of each frame of rehabilitation training image, determining each frame of rehabilitation training image except a first frame of rehabilitation training image in the marked unit period as a candidate training image, determining local motion vectors corresponding to each limb key point in each frame of candidate training image according to the motion condition of different limb key points in each frame of candidate training image, forming a candidate limb key point set by representing limb key points with the same human body part in all candidate training images, determining the average value of the local motion vectors corresponding to all limb key points in each candidate limb key point set as a target motion vector corresponding to each candidate limb key point set, and forming a target motion vector sequence corresponding to the marked unit period; determining action difference degree between every two unit training periods according to a target motion vector sequence and a target state value sequence corresponding to every two unit training periods, planning intervals of all unit training periods in all preset rehabilitation training periods according to the action difference degree between different unit training periods through an interval DP algorithm, and recording each interval finally obtained by interval planning as a training action period; The method comprises the steps of determining target difference representative values between every two training action time periods according to action difference degrees between different unit training time periods in every two training action time periods, clustering all training action time periods by taking the target difference representative values between different training action time periods as clustering distance measurement, and marking the obtained clustering clusters as target clusters; determining target recommended guidance values corresponding to training actions represented by each target cluster comprises determining target recommended guidance values corresponding to training actions represented by each target cluster according to accumulated values of time lengths corresponding to all training action periods in each target cluster, average values of action proficiency indexes corresponding to all training action periods in each target cluster and average values of cognitive recovery scores corresponding to preset rehabilitation training periods to which all training action periods in each target cluster belong.
2. The home environment cognitive rehabilitation training action guidance system based on visual interaction according to claim 1, wherein the determining the local motion vector corresponding to each limb key point in each frame of candidate training image according to the motion condition of different limb key points in each frame of candidate training image comprises: Determining any frame of candidate training image as a temporary training image, and determining each limb key point in the temporary training image as a temporary key point; Screening out pixel points matched with each temporary key point from a rehabilitation training image of a previous frame of the temporary training image, and taking the pixel points as matched pixel points corresponding to each temporary key point; Screening out pixel points which are the same as the matched pixel points corresponding to each temporary key point from the temporary training image, and taking the pixel points as reference pixel points corresponding to each temporary key point; Taking the distance between each temporary key point and the corresponding reference pixel point as a vector module, taking the direction of each temporary key point pointing to the corresponding reference pixel point as a vector direction, and marking the constructed vector as a local motion vector corresponding to each temporary key point.
3. The home environment cognition rehabilitation training action guidance system based on visual interaction according to claim 1, wherein the constructing the target state value sequence corresponding to the mark unit period according to the pixel coordinates corresponding to the limb key points in the candidate limb key point set comprises: determining point state factors corresponding to each limb key point according to the distance between each limb key point and other limb key points in the rehabilitation training image to which the limb key point belongs; the average value of point state factors corresponding to all limb key points in each candidate limb key point set is determined as a target state value corresponding to each candidate limb key point set; And integrating the target state values corresponding to all candidate limb key points to form a target state value sequence corresponding to the mark unit time period.
4. The home environment cognitive rehabilitation training action guidance system according to claim 1, wherein the determining the action difference degree between each two unit training periods according to the target motion vector sequence and the target state value sequence corresponding to each two unit training periods comprises: And determining the action difference degree between every two unit training periods according to the cosine similarity between the target motion vectors with the same sequence number in the target motion vector sequences corresponding to every two unit training periods and the absolute value of the difference value between the target state values with the same sequence number in the target state value sequences corresponding to every two unit training periods.
5. The home environment cognitive rehabilitation training action guidance system according to claim 1, wherein the determining the target difference representative value between each two training action periods according to the action difference degree between different unit training periods within each two training action periods comprises: Determining any two training action time periods as a first training action time period and a second training action time period respectively; Determining each unit training period in the first training action period as a first unit training period, and determining each unit training period in the second training action period as a second unit training period; Determining the average value of the action difference degrees between each first unit training period and all second unit training periods as the reference difference degree corresponding to each first unit training period; determining the average value of the action difference degrees between each second unit training period and all the first unit training periods as the reference difference degree corresponding to each second unit training period; And determining the average value of all the reference difference degrees as a target difference representative value between the first training action time period and the second training action time period.
6. The home environment cognition rehabilitation training action guidance system based on visual interaction according to claim 1, wherein the determining the action proficiency index corresponding to each training action period according to the similarity between each training action period and other training action periods in the target cluster to which the training action period belongs comprises: determining any training action time period as a marking action time period, determining a target cluster to which the marking action time period belongs as a marking cluster, and determining each training action time period except the marking action time period in the marking cluster as a reference action time period; and determining an action proficiency index corresponding to the marked action time period according to the average value of the target difference representative values between the marked action time period and all the reference action time periods.

Description

Household environment cognition rehabilitation training action guiding system based on visual interaction Technical Field The invention relates to the technical field of rehabilitation training, in particular to a home environment cognition rehabilitation training action guiding system based on visual interaction. Background In practical situations, in order to facilitate selection of the cognitive rehabilitation training actions by the patient, a suitable cognitive rehabilitation training action is often recommended to the patient, so as to guide the patient to perform training of the suitable cognitive rehabilitation training action in a limited time. At present, when recommending objects to users, a method is generally adopted in which the objects are recommended to users according to the suitability of the objects for most users, namely, if a certain object is suitable for the whole situation of most users, the objects are recommended to each user. However, when a cognitive rehabilitation training action suitable for the overall situation of most patients is directly recommended to a target patient, the following technical problems often exist: The situation of a single patient often has self uniqueness and does not necessarily accord with the overall situation of most patients, so that the recommended cognitive rehabilitation training actions suitable for the overall situation of most patients are directly recommended to target patients, and the recommended cognitive rehabilitation training actions are possibly not suitable for the target patients, so that the rationality of recommending the cognitive rehabilitation training actions to the patients is poor, and the guiding effect of the cognitive rehabilitation training actions is poor. Disclosure of Invention The invention provides a home environment cognitive rehabilitation training action guiding system based on visual interaction, aiming at solving the technical problem of poor cognitive rehabilitation training action guiding effect caused by poor rationality of recommending cognitive rehabilitation training actions to patients. In a first aspect, the present invention provides a home environment cognitive rehabilitation training action guidance system based on visual interaction, the system comprising: the data acquisition module is used for acquiring rehabilitation training images of a target patient in each unit training period in each preset rehabilitation training period and acquiring cognitive recovery scores corresponding to each preset rehabilitation training period; The vector determining module is used for determining a target motion vector sequence and a target state value sequence corresponding to each unit training period based on all rehabilitation training images in each unit training period; the dividing and clustering module is used for dividing intervals of all preset rehabilitation training periods according to the target motion vector sequences and the target state value sequences corresponding to all unit training periods to obtain training action periods, and clustering all training action periods to obtain target clusters; The index determining module is used for determining an action proficiency index corresponding to each training action period according to the similarity between each training action period and other training action periods in the target cluster to which the training action period belongs; The guiding value determining module is used for determining a target recommended guiding value corresponding to the training action represented by each target cluster according to the duration corresponding to different training action time periods in each target cluster, the action proficiency index and the cognitive recovery score corresponding to the preset rehabilitation training period to which the action proficiency index belongs. With reference to the first aspect, in one possible implementation manner, the determining, based on all rehabilitation training images in each unit training period, a target motion vector sequence and a target state value sequence corresponding to each unit training period includes: Determining any one preset rehabilitation training period as a marked rehabilitation training period, and determining any one unit training period in the marked rehabilitation training period as a marked unit period; Extracting limb key points from each frame of rehabilitation training image, and determining each frame of rehabilitation training image except the first frame of rehabilitation training image in the marking unit time period as a candidate training image; Determining local motion vectors corresponding to each limb key point in each frame of candidate training image according to the motion conditions of different limb key points in each frame of candidate training image; limb key points with the same human body parts in all candidate training images are formed into a candidate limb key point set;