CN115527266A - User identification method and device based on gait information, electronic equipment and storage medium - Google Patents
User identification method and device based on gait information, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN115527266A CN115527266A CN202211110679.0A CN202211110679A CN115527266A CN 115527266 A CN115527266 A CN 115527266A CN 202211110679 A CN202211110679 A CN 202211110679A CN 115527266 A CN115527266 A CN 115527266A
- Authority
- CN
- China
- Prior art keywords
- user
- data
- gait
- model
- gait data
- 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
- G06V40/23—Recognition of whole body movements, e.g. for sport training
- G06V40/25—Recognition of walking or running movements, e.g. gait 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
- G06V10/40—Extraction of image or video features
-
- 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
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
- G06V10/765—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects using rules for classification or partitioning the feature space
-
- 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
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Psychiatry (AREA)
- Social Psychology (AREA)
- Human Computer Interaction (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The embodiment of the invention discloses a user identification method, a device, electronic equipment and a storage medium based on gait information, wherein the method comprises the following steps: acquiring gait data of a user to be identified to obtain the gait data; based on a current identity recognition model obtained by pre-training, carrying out user recognition processing on the gait data to obtain a recognition result; adding a user tag to the gait data when the identification result does not contain a target user tag; performing model training on the current identity recognition model based on the gait data and a corresponding user label to obtain an updated current identity recognition model, and performing user recognition on the gait data based on the updated current identity recognition model; and under the condition that the identification result contains a target user tag, obtaining the user identity of the user to be identified based on the target user tag. The technical scheme of the embodiment of the invention can improve the accuracy of user identity identification.
Description
Technical Field
The present invention relates to the field of artificial intelligence technologies, and in particular, to a method and an apparatus for identifying a user based on gait information, an electronic device, and a storage medium.
Background
The user identity recognition is one of basic services supporting the safety of the Internet of things, and the current user identity recognition mainly adopts a vision-based recognition mode and finishes the user identity recognition by extracting the visual features of the user through a camera. However, visual-based identification is limited by challenges such as ambient lighting, and cannot work properly in dark environments. Although face information is widely used in various scenes such as payment and password recognition, the recognition of face information is affected by light, face occlusion, and the like, and gait information recognition is more private.
The existing gait recognition technology is not high in precision of user identity recognition, and therefore user experience is affected. Therefore, a user identification method based on gait information is needed to solve the above technical problems.
Disclosure of Invention
The embodiment of the invention provides a user identification method and device based on gait information, electronic equipment and a storage medium, and solves the technical problems.
In a first aspect, an embodiment of the present invention provides a method for identifying a user based on gait information, where the method includes:
acquiring gait data of a user to be identified to obtain the gait data;
based on a current identity recognition model obtained by pre-training, carrying out user recognition processing on the gait data to obtain a recognition result;
adding a user tag to the gait data when the identification result does not contain a target user tag; performing model training on the current identity recognition model based on the gait data and a corresponding user label to obtain an updated current identity recognition model, and performing user recognition on the gait data based on the updated current identity recognition model;
and under the condition that the identification result contains a target user tag, obtaining the user identity of the user to be identified based on the target user tag.
In a second aspect, an embodiment of the present invention further provides a device for identifying a user based on gait information, where the device includes:
the gait data acquisition module is used for acquiring gait data of the user to be identified to acquire the gait data;
the identification result acquisition module is used for carrying out user identification processing on the gait data based on a current identity identification model obtained by pre-training to obtain an identification result;
the model updating module is used for adding a user tag to the gait data under the condition that the identification result does not contain a target user tag; performing model training on the current identity recognition model based on the gait data and a corresponding user label to obtain an updated current identity recognition model, and performing user recognition on the gait data based on the updated current identity recognition model;
and the user identity determining module is used for obtaining the user identity of the user to be identified based on the target user label under the condition that the identification result contains the target user label.
In a third aspect, an embodiment of the present invention further provides an electronic device, where the electronic device includes:
one or more processors;
a storage device to store one or more programs,
when executed by the one or more processors, cause the one or more processors to implement a method of user identification based on gait information as in any embodiment of the invention.
In a fourth aspect, an embodiment of the present invention further provides a storage medium containing computer-executable instructions, where the computer-executable instructions are configured to perform a method for identifying a user based on gait information according to any embodiment of the present invention when executed by a computer processor.
In the embodiment of the invention, gait data of a user to be identified is acquired, the gait data is acquired, and the user identification processing is carried out on the gait data based on a current identity identification model acquired by pre-training to acquire an identification result. And under the condition that the identification result does not contain the target user label, adding a user label to the gait data, performing model training on the current identity identification model based on the gait data and the corresponding user label, obtaining the updated current identity identification model, and performing user identification on the gait data based on the updated current identity identification model, so that when the gait data of the user to be identified is obtained again, the user identity of the user to be identified can be obtained based on the gait data, and the user experience is improved. Under the condition that the identification result contains the target user label, the user identity of the user to be identified is obtained based on the target user label, and compared with face identification, fingerprint identification and the like, the influence of factors such as light or whether the finger has water stain does not need to be considered, so that the user identity identification is more accurate.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
Wherein:
fig. 1 is a schematic flowchart of a user identification method based on gait information in an embodiment;
fig. 2 is a schematic diagram of a gait data acquisition device in another embodiment;
fig. 3 is a schematic structural diagram of a user identification device based on gait information in another embodiment;
fig. 4 is a schematic structural diagram of an electronic device in another embodiment.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be obtained by a person skilled in the art without inventive step based on the embodiments of the present invention, are within the scope of protection of the present invention.
In an embodiment of the present invention, a method for identifying a user based on gait information is provided, where the method is applicable to the case of identifying the gait information of the user, and the method may be performed by a user identification device based on gait information, and the device may be implemented in a form of software and/or hardware.
As shown in fig. 1, a user identification method based on gait information according to an embodiment of the present invention includes:
and S110, acquiring gait data of the user to be identified to obtain the gait data.
The user to be identified refers to a user needing identity identification. The gait data can be gait data of CSI signals acquired through WIFI, and can also be gait image data acquired through image acquisition equipment.
In the embodiment of the invention, the gait data of the user to be identified is collected, optionally, a collection period can be preset, and the gait data of the user to be identified is collected according to the collection period. And the gait data of the user to be identified prepares for the subsequent identification of the user to be identified.
Optionally, a WIFI transmitter and a receiver are arranged in the target area, the user to be identified walks between the WIFI transmitter and the receiver, and the receiver acquires gait data. The target area may be an area of a hallway, corridor, or the like. The WIFI transmitter and the receiver are arranged on two sides of the passageway, and when a user to be identified passes through the passageway, the receiver collects gait data of the user to be identified.
For example, the embodiment of the present invention designs an experiment for acquiring gait data of a user to be identified, where the experiment device includes: hp8530p notebook computer, built-in 5300 802.11n card is as the receiver, and a WIFI router Netgear R7000 is as the transmitter. The experimental environment includes: three WIFI transmitters and three receivers are respectively placed on two sides of the corridor, an experimenter moves between the WIFI transmitters and the receivers arranged in the experiment, the experimenter sends out signals through the WIFI transmitters, and gait data of the experimenter are received through the receivers, and the experimenter is shown in fig. 2. T1, T2 and T3 denote WIFI transmitters, and R1, R2 and R3 denote receivers. It should be understood that the gait data includes 9 sets of data, each set of data is collected for no more than 5 seconds, and the 9 collected sets of data are transformed in a matrix form to obtain a matrix of 9 × 30 × 3200. 9 denotes 9 groups of data, 30 denotes 30 subcarriers included in each group of data, and 3200 denotes 3200 CSI streams included per subcarrier. By the gait data acquisition method, the gait data of the experimenter can be acquired.
Optionally, because the WIFI signal itself has characteristics of environmental noise and CSI phase instability, when acquiring gait data, it is necessary to preprocess the gait data to obtain gait data that can be input into the current identification model. Specifically, the embodiment of the present invention performs outlier removal on 30 subcarriers of each group by a median filtering method. And in order to enhance information of the gait data and improve the identification effect of the gait data, short-time Fourier transform is carried out on the collected gait data to obtain a spectrogram, and relative key data in a frequency band of 20-80Hz is intercepted and used as the gait data input into the current identity identification model.
And S120, based on the current identity recognition model obtained through pre-training, carrying out user recognition processing on the gait data to obtain a recognition result.
The current identity recognition model is a model which can obtain a user identity recognition result after gait data are processed. The identification result may be a user tag or an unknown user, i.e. no user tag.
Specifically, the gait data is subjected to user identification processing through a current identity identification model obtained through pre-training to obtain an identification result, and preparation is made for subsequently determining the user identity of the user to be identified according to the identification result or updating the model.
S130, adding a user tag to the gait data under the condition that the identification result does not contain a target user tag; and performing model training on the current identity recognition model based on the gait data and the corresponding user label to obtain an updated current identity recognition model, and performing user recognition on the gait data based on the updated current identity recognition model.
The step of identifying the gait data of the user to be identified by the identification module, wherein the step of not including the target user label means that the identification result corresponding to the gait data is an unknown user, and in the current identity identification model, the gait data of the user to be identified is not trained, so that the user label of the gait data of the user to be identified cannot be obtained.
In the embodiment of the invention, under the condition that the identification result does not contain the target user label, the user label is added to the gait data, the current identity identification model is trained based on the gait data and the user label corresponding to the gait data, the updated current identity identification model is obtained, and the user identification is carried out on the gait data based on the updated current identity identification model. Therefore, when the gait data of the user to be identified is collected again, the updated current identity recognition model can recognize the user label of the user to be identified, and the identity recognition efficiency and accuracy of the user to be identified are improved.
S140, under the condition that the identification result contains the target user label, obtaining the user identity of the user to be identified based on the target user label.
The target user label is the user label of the user to be identified, which is obtained after the gait data is processed through the current identity identification model. The user tags correspond to user identities one to one.
In the embodiment of the invention, under the condition that the identification result comprises the target user label, the user identity of the user to be identified is obtained based on the target user label, the user identity of the user to be identified can be known based on the collected gait data, the problems of light and the like are required to be considered relative to image identification and the like, the user identity is determined through the gait data, and the accuracy of user identity determination can be improved. Optionally, the user identity corresponding to the user tag may be determined according to the corresponding relationship between the user tag and the user identity and the user tag.
In the embodiment of the invention, gait data of a user to be identified is acquired, the gait data is acquired, and the user identification processing is carried out on the gait data based on a current identity identification model acquired by pre-training to acquire an identification result. And under the condition that the identification result does not contain the target user label, adding the user label to the gait data, performing model training on the current identity identification model based on the gait data and the corresponding user label to obtain an updated current identity identification model, and performing user identification on the gait data based on the updated current identity identification model, so that when the gait data of the user to be identified is obtained again, the user identity of the user to be identified can be obtained based on the gait data, and the user experience is improved. Under the condition that the identification result contains the target user label, the user identity of the user to be identified is obtained based on the target user label, and compared with face identification, fingerprint identification and the like, the influence of factors such as light or whether the finger has water stain does not need to be considered, so that the user identity identification is more accurate.
In another embodiment of the present invention, the performing user identification processing on the gait data based on the current identity identification model obtained by pre-training to obtain an identification result includes: performing feature extraction on the gait data based on a current convolution network model in the current identity recognition model to obtain a feature vector; and carrying out user identification processing on the characteristic vector based on a classification model in the identity identification model to obtain an identification result.
The current identity recognition model comprises a current convolution network model and a classification model. Optionally, the classification model may include any one of a random forest model, a KNN model, a naive bayes model, a support vector set model, and a regression analysis model. The current convolutional network model may be any one of a LeNet network model, an Alexnet network model, and a ResNet network model. Optionally, in the embodiment of the present invention, resNet is selected as the current convolutional network model, and the number of convolutional layers may be 18.
Specifically, feature extraction is performed on gait data based on a current convolution network model to obtain feature vectors, and user identification processing is performed on the feature vectors based on a classification model to obtain an identification result. The features in the gait data can be well extracted through the current convolution network model obtained through training, the recognition result corresponding to the gait data is obtained through classification of the classification model, and the accuracy of determining the recognition result is improved. It should be understood that the classification model may also be obtained after training.
In another technical solution of the embodiment of the present invention, the performing, by the feature vector based on a classification model in the identity recognition model, user recognition processing to obtain a recognition result includes: classifying the feature vectors based on a classification model in the identity recognition model to obtain the probability that the feature vectors belong to each user label; determining the probability with the highest numerical value and larger than a preset threshold value in all the probabilities as a target probability; and determining the user label corresponding to the target probability as a target user label to obtain the identification result.
In the embodiment of the invention, the feature vectors are classified according to the classification model in the identity recognition model to obtain the probability that the feature vectors belong to each user tag, the probability with the highest numerical value and larger than the preset threshold value in each probability is determined as the target probability, the user tag corresponding to the target probability is determined as the target user tag, and the recognition result is further obtained.
Illustratively, referring to table 1, the user tags include user a, user B, user C, user D, and user E, and the preset threshold is 40%. The probability that the feature vector A belongs to each user label is 20%,30%,70%,80% and 95% in sequence, the highest value in the probability is 95%, the corresponding user label is the user E, and is greater than a preset threshold value, 95% is determined as the target probability, 95% of the corresponding user E is determined as the target user label, and then the identification result is obtained as the user E.
TABLE 1
In another embodiment of the present invention, before the performing user identification processing on the gait data based on the current identity recognition model obtained by pre-training to obtain an identification result, the method further includes: carrying out triple sampling on the training data set to obtain a plurality of triple data; processing each triple data through an initial feature extraction submodel in the initial convolutional network model to obtain a plurality of gait feature vectors; processing each gait feature vector through a triple loss function in the initial convolutional network model to obtain loss values of a plurality of triple data, and obtaining an integral loss value according to the loss value of each triple data; and adjusting the initial feature extraction submodel based on the overall loss value until the overall loss value is within a preset range so as to obtain the current convolutional network model.
The triple sampling is to obtain two gait data of the same user from a training data set, mark one of the two gait data as an anchor point sample, mark the other one as a positive sample, and then obtain the gait data of one other user from the training data set as a negative sample, wherein the positive sample is the gait data of the same user, and the negative sample is the gait data of other users except the anchor point sample. This obtains a plurality of triple data. The initial convolutional network model is a ResNet18 network model, the loss function in the initial convolutional network model is replaced by a triple loss function, and the hidden dimension of the triple loss function can be set to be 8. The activation function may be a ReLU function. Because the triple loss function belongs to a metric learning method, a deep metric learning mode is obtained by combining the initial convolution network model and the triple loss function. The method for deep measurement learning can improve the accuracy of feature vector extraction, so that the information in the feature vector is richer, the characteristics of gait data can be more represented, namely the gait features are more accurately extracted, and the accuracy of identifying the user label based on the gait data is higher. And the characteristics of the unknown user can be identified, so that the gait recognition model can identify the unknown user.
Specifically, each triplet data is processed through an initial feature extraction submodel in an initial convolutional network model to obtain a plurality of gait feature vectors, each gait feature vector is processed through a triplet loss function in the initial convolutional network model to obtain loss values of the plurality of triplet data, the loss values of each triplet data are summed to obtain an average value, the average value is used as an overall loss value, the initial feature extraction submodel is adjusted based on the overall loss value until the overall loss value is within a preset range, and the current convolutional network model is obtained. Alternatively, for a method of obtaining an overall loss value based on the loss values of each triplet of data, the variance, or mean square deviation, of these loss values may also be obtained. It should be understood that the preset range can be set according to actual conditions.
Optionally, the gait data x is subjected to convolution transformation through a feature extraction submodel of the initial convolutional neural network to obtain a feature embedding space e (x; θ), where e () represents convolution and θ represents convolution parameters, such as a weight of each convolutional layer. The process of embedding the features into the space, namely the process of mapping the features, is as follows: the gait data x is normalized and mapped to a d-dimensional hypersphere space of an embedding space through L2, namely, the space after the mapping of the ternary loss function can be expressed as e (x; theta) epsilon R d And satisfy | | e (x; theta) | calculation of sulphur 2 And =1. For example, when the semantic similarity d (X1, Y1, θ) of X1 and Y1 is calculated by the triplet loss function, the distance network between the embedded features can be calculated using euclidean distance:
in the embodiments of the present invention, for convenience of explanation, the following will be referred to as D X1Y1 To represent d (X1, Y1, θ), for the anchor sample a, the positive samplep, negative sample n, equation 1 can be obtained:
wherein [ a ]] + If a is larger than 0, a is taken, and if a is smaller than zero, zero is taken. Alpha denotes a manually set threshold value. Where α is the border block of positive and negative samples, and all triplet gait data (a, p, n) are e.t. T represents a WIFI gait data set. It can be achieved by the above equation 1 that the anchor sample is closer to the positive sample than the negative sample.
The embodiment of the invention optimizes the triple loss function, which is shown in a formula 2:
where B represents the total number of batches, and at the time of sampling, the data in each batch is sampled. a is the anchor sample, p is the positive sample, n is the negative sample, α is the border block of the positive and negative samples, i is the batch number, which can be 1,2,3, \8230; \8230, B. As for
Is selected based on
The euclidean distance of (b) is to select a negative sample having an appropriate distance from the positive sample according to the euclidean distance, and the information content of the negative sample is richer. The setting can play the role of a triple loss function, and the problem of over-fitting of positive samples is avoided.
Alternatively, due to the triple loss function, a closer euclidean distance between the same type of data relative to the distance between different types of data may be achieved. Therefore, the classification model can select the KNN model, and the distance between every two categories in the feature space is mapped by combining the convolution network model, so that the depth measurement learning effect is better. For example, for target data φ, L (φ) = (1- μ) L 1 (φ)+μL 2 (φ) where μ is a parameter, representing the ratio of push to pull, push representing an increase in distance between two different data types, and pull representing a decrease in distance between data of the same data type.
For L 1 () Indicating that data is pushed to a position closer to data of the same data type as it is, that is, the minimum distance of the target data from other data of the same data type is calculated, and the closer the distance is, the higher the similarity between the target data and the other data is understood as:
L 1 (φ)＝∑ i,j∈T(i) D i,j t (i) is the target neighbor of data i and data j, and is also an anchor point for the same user's gait data as data i and j. The above equation represents pulling the data i closer to the target neighbor T (i).
For L 2 () Represents pushing data farther away than data of a different data type:
L 2 (φ)＝∑ a,n,ya≠yn [m+D a,T(a) -D an ] +
where Da, T (a) represents the distance between data a and the target neighbor of data a, and the target neighbor T (a) refers to an anchor point of the gait data of the same user adjacent to data a. ya ≠ yn represents that data a and data n belong to anchor points of gait data of different users, and y represents the user corresponding to the gait data. And m is a parameter, and when the distance between a and n is larger than m, n is pushed to a position farther away from a. It should be understood that the two formulas L described above 1 Phi and L 2 The data mentioned in (φ) refer to gait data. The two data are of different data types, i.e. the two data do not belong to the same user.
The gait data of the user to be identified is identified, if the identification result is that the target user label is not included, the gait data is added with the user label, the gait data and the corresponding user label are added into the training data set to obtain a new training data set, the new training data set is subjected to triple sampling to obtain new triple gait data, and then the trained current convolutional network model is subjected to model training again according to the new triple gait data to obtain an updated current convolutional network model.
In another embodiment of the present invention, the clustering each target gait data to obtain a plurality of data categories includes: and periodically acquiring each target gait data in the preset time interval according to the preset time interval, so that the current convolutional network model in the current identity recognition model is periodically updated on the basis of each target gait data.
It should be appreciated that the point in time at which the current convolutional network model is updated is after the acquisition of each target gait data.
In the embodiment of the invention, each target gait data in the preset time interval is regularly acquired according to the preset time interval, and the current convolutional network model in the current identity recognition model is regularly updated based on each target gait data, so that the current convolutional network model can be suitable for the user to be recognized corresponding to each target gait data, and further the current identity recognition model can recognize the user to be recognized corresponding to each target gait data.
Optionally, after obtaining each target gait data, adding a user tag to each target gait data, and directly using each target gait data and the corresponding user tag to train the current convolutional network model, or of course, adding each target gait data to a training data set to obtain a new training data set, and training the current convolutional network model based on the new training data set to obtain an updated current convolutional network model.
Optionally, based on the gait data and the corresponding user label, performing model training on the current identity recognition model, including: model training is carried out on the current convolution network model in the current identity recognition model based on the gait data and the corresponding user label, and the method specifically comprises the following steps: determining a plurality of target identification results which do not contain target user tags; acquiring target gait data corresponding to each target identification result; clustering the target gait data to obtain a plurality of data categories; adding a user tag to each data category; adding each user label and at least one target gait data corresponding to the user label into a training data set to obtain a new training data set; and performing model training on the current convolution network model in the current identity recognition model based on the new training data set.
The algorithm for clustering the target gait data includes, but is not limited to, a K-means algorithm, a DBSCAN algorithm, a spectral clustering algorithm, a Gaussian Mixture Model (GMM), and the like.
In the embodiment of the invention, when the model training is carried out on the current identity recognition model, the model training is carried out on the current convolution network model in the current identity recognition model. Specifically, a plurality of target identification results which do not include target user tags are determined, target gait data corresponding to the target identification results are obtained, the target gait data are clustered to obtain at least one data category, user tags are added to the data categories, and each user tag and at least one target gait data corresponding to the user tag are added to a training data set to obtain a new training data set; model training is carried out on the current convolution network model based on the new training data set, so that the current convolution network model can learn the features in the target gait data, and further the features in the target gait data can be recognized when the gait data are collected again.
In another embodiment of the present invention, a user identification apparatus based on gait information is provided, and the user identification apparatus based on gait information provided in the embodiment of the present invention can execute the user identification method based on gait information provided in any embodiment of the present invention, and has a functional module corresponding to the execution method and a beneficial effect. Referring to fig. 3, the apparatus includes: a gait data acquisition module 310, a recognition result acquisition module 320, a model update module 330 and a user identity determination module 340; wherein:
the gait data acquisition module 310 is configured to acquire gait data of a user to be identified, and acquire the gait data; a recognition result obtaining module 320, configured to perform user recognition processing on the gait data based on a current identity recognition model obtained through pre-training, to obtain a recognition result; a model updating module 330, configured to add a user tag to the gait data when the recognition result does not include a target user tag; performing model training on the current identity recognition model based on the gait data and a corresponding user label to obtain an updated current identity recognition model, and performing user recognition on the gait data based on the updated current identity recognition model; a user identity determining module 340, configured to, if the recognition result includes a target user tag, obtain the user identity of the user to be recognized based on the target user tag.
Further, the identification result obtaining module 320 in the embodiment of the present invention is further configured to:
performing feature extraction on the gait data based on a current convolution network model in the current identity recognition model to obtain a feature vector; and carrying out user identification processing on the characteristic vector based on a classification model in the identity identification model to obtain an identification result.
Further, the identification result obtaining module 320 in the embodiment of the present invention is further configured to:
classifying the feature vectors based on a classification model in the identity recognition model to obtain the probability that the feature vectors belong to each user label; determining the probability with the highest numerical value and larger than a preset threshold value in all the probabilities as a target probability; and determining the user label corresponding to the target probability as a target user label to obtain the identification result.
Further, the apparatus in the embodiment of the present invention further includes:
the current convolutional network model training module is used for carrying out triple sampling on a training data set to obtain a plurality of triple data; processing each triple data through an initial feature extraction submodel in the initial convolutional network model to obtain a plurality of gait feature vectors; processing each gait feature vector through a triple loss function in the initial convolutional network model to obtain loss values of a plurality of triple data, and obtaining an integral loss value according to the loss value of each triple data; and adjusting the initial feature extraction submodel based on the overall loss value until the overall loss value is within a preset range so as to obtain the current convolutional network model.
Further, the model updating module 330 in the embodiment of the present invention is further configured to:
performing model training on a current convolutional network model in the current identity recognition model based on the gait data and a corresponding user label, and specifically configured to: determining a plurality of target identification results which do not contain target user tags; acquiring target gait data corresponding to each target identification result; clustering each target gait data to obtain a plurality of data categories;
adding a user tag to each data category; adding each user label and at least one target gait data corresponding to the user label into a training data set to obtain a new training data set; and performing model training on the current convolutional network model in the current identity recognition model based on the new training data set.
Further, the model updating module 330 in the embodiment of the present invention is further configured to:
and periodically acquiring each target gait data in the preset time interval according to the preset time interval, so that the current convolutional network model in the current identity recognition model is periodically updated on the basis of each target gait data.
Further, in this embodiment of the present invention, the current convolutional network model includes a ResNet network model, and the classification model includes a KNN model.
In the embodiment of the invention, gait data of a user to be identified is acquired to obtain the gait data, and the gait data is identified and processed by the user based on the current identity identification model obtained by pre-training to obtain an identification result. And under the condition that the identification result does not contain the target user label, adding the user label to the gait data, performing model training on the current identity identification model based on the gait data and the corresponding user label to obtain an updated current identity identification model, and performing user identification on the gait data based on the updated current identity identification model, so that when the gait data of the user to be identified is obtained again, the user identity of the user to be identified can be obtained based on the gait data, and the user experience is improved. Under the condition that the identification result contains the target user label, the user identity of the user to be identified is obtained based on the target user label, and compared with face identification, fingerprint identification and the like, the influence of factors such as whether light or fingers have water stains does not need to be considered, so that the user identity identification is more accurate.
It should be noted that, the modules included in the apparatus are merely divided according to functional logic, but are not limited to the above division as long as the corresponding functions can be implemented; in addition, specific names of the functional modules are only for convenience of distinguishing from each other, and are not used for limiting the protection scope of the embodiment of the present invention.
In another embodiment of the present invention, an electronic device is provided. Fig. 4 is a schematic structural diagram of an electronic device according to an embodiment of the present invention. FIG. 4 illustrates a block diagram of an exemplary electronic device 50 suitable for use in implementing embodiments of the present invention. The electronic device 50 shown in fig. 4 is only an example, and should not bring any limitation to the functions and the scope of use of the embodiment of the present invention.
As shown in fig. 4, the electronic device 50 is in the form of a general purpose computing device. The components of the electronic device 50 may include, but are not limited to: one or more processors or processing units 501, a system memory 502, and a bus 503 that couples the various system components (including the system memory 502 and the processing unit 501).
The system memory 502 may include computer system readable media in the form of volatile memory, such as Random Access Memory (RAM) 504 and/or cache memory 505. The electronic device 50 may further include other removable/non-removable, volatile/nonvolatile computer system storage media. By way of example only, storage system 506 may be used to read from and write to non-removable, nonvolatile magnetic media (not shown in FIG. 4, and commonly referred to as a "hard drive"). Although not shown in FIG. 4, a magnetic disk drive for reading from and writing to a removable, nonvolatile magnetic disk (e.g., a "floppy disk") and an optical disk drive for reading from or writing to a removable, nonvolatile optical disk (e.g., a CD-ROM, DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to the bus 503 by one or more data media interfaces. Memory 502 may include at least one program product having a set (e.g., at least one) of program modules that are configured to carry out the functions of embodiments of the invention.
A program/utility 508 having a set (at least one) of program modules 507 may be stored, for instance, in memory 502, such program modules 507 including, but not limited to, an operating system, one or more application programs, other program modules, and program data, each of which examples or some combination thereof may comprise an implementation of a network environment. Program modules 507 generally perform the functions and/or methodologies of embodiments of the invention as described herein.
The electronic device 50 may also communicate with one or more external devices 509 (e.g., keyboard, pointing device, display 510, etc.), with one or more devices that enable a user to interact with the electronic device 50, and/or with any devices (e.g., network card, modem, etc.) that enable the electronic device 50 to communicate with one or more other computing devices. Such communication may occur via input/output (I/O) interfaces 511. Also, the electronic device 50 may communicate with one or more networks (e.g., a Local Area Network (LAN), a Wide Area Network (WAN), and/or a public network such as the Internet) via the network adapter 512. As shown, the network adapter 512 communicates with the other modules of the electronic device 50 over the bus 503. It should be appreciated that although not shown in FIG. 4, other hardware and/or software modules may be used in conjunction with electronic device 50, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, among others.
The processing unit 501 executes various functional applications and data processing, for example, a user identification method based on gait information provided by an embodiment of the present invention, by running a program stored in the system memory 502.
In another embodiment of the present invention, there is also provided a storage medium containing computer executable instructions which, when executed by a computer processor, perform a method of user identification based on gait information, the method comprising:
acquiring gait data of a user to be identified to obtain the gait data; based on a current identity recognition model obtained by pre-training, carrying out user recognition processing on the gait data to obtain a recognition result; adding a user tag to the gait data when the identification result does not contain a target user tag; model training is carried out on the current identity recognition model based on the gait data and a corresponding user label, an updated current identity recognition model is obtained, and user recognition is carried out on the gait data based on the updated current identity recognition model; and under the condition that the identification result contains a target user tag, obtaining the user identity of the user to be identified based on the target user tag.
Computer storage media for embodiments of the present invention may take the form of any combination of one or more computer-readable media. The computer readable medium may be a computer readable signal medium or a computer readable storage medium. A computer readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples (a non-exhaustive list) of the computer readable storage medium would include the following: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this document, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.
A computer readable signal medium may include a propagated data signal with computer readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take many forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.
Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.
Computer program code for carrying out operations for embodiments of the present invention may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, smalltalk, C + + or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet service provider).
The above disclosure is only for the purpose of illustrating the preferred embodiments of the present invention, and it is therefore to be understood that the invention is not limited by the scope of the appended claims.
Claims (10)
1. A user identification method based on gait information is characterized by comprising the following steps:
acquiring gait data of a user to be identified to obtain the gait data;
based on a current identity recognition model obtained by pre-training, carrying out user recognition processing on the gait data to obtain a recognition result;
adding a user tag to the gait data when the identification result does not contain a target user tag; performing model training on the current identity recognition model based on the gait data and a corresponding user label to obtain an updated current identity recognition model, and performing user recognition on the gait data based on the updated current identity recognition model;
and under the condition that the identification result contains a target user tag, obtaining the user identity of the user to be identified based on the target user tag.
2. The method for recognizing a user based on gait information according to claim 1, wherein the step of performing user recognition processing on the gait data based on a current identity recognition model obtained by pre-training to obtain a recognition result comprises:
performing feature extraction on the gait data based on a current convolution network model in the current identity recognition model to obtain a feature vector;
and carrying out user identification processing on the characteristic vector based on a classification model in the identity identification model to obtain an identification result.
3. The gait information-based user identification method according to claim 2, wherein the performing user identification processing on the feature vectors based on the classification model in the identity identification model to obtain an identification result comprises:
classifying the feature vectors based on a classification model in the identity recognition model to obtain the probability that the feature vectors belong to each user label;
determining the probability with the highest numerical value and larger than a preset threshold value in all the probabilities as a target probability;
and determining the user label corresponding to the target probability as a target user label to obtain the identification result.
4. The method for recognizing a user based on gait information as claimed in claim 1, wherein before the step of performing a user recognition process on the gait data based on the pre-trained current identity recognition model to obtain a recognition result, the method further comprises:
carrying out triple sampling on the training data set to obtain a plurality of triple data;
processing each triple data through an initial feature extraction submodel in the initial convolutional network model to obtain a plurality of gait feature vectors;
processing each gait feature vector through a triple loss function in the initial convolutional network model to obtain loss values of a plurality of triple data, and obtaining an integral loss value according to the loss value of each triple data;
and adjusting the initial feature extraction submodel based on the overall loss value until the overall loss value is within a preset range so as to obtain the current convolutional network model.
5. The gait information-based user recognition method according to claim 1, wherein the model training of the current identification model based on the gait data and the corresponding user label comprises:
performing model training on a current convolutional network model in the current identity recognition model based on the gait data and a corresponding user label, specifically comprising:
determining a plurality of target identification results which do not contain target user tags;
acquiring target gait data corresponding to each target identification result;
clustering each target gait data to obtain a plurality of data categories;
adding a user tag to each data category;
adding each user label and at least one target gait data corresponding to the user label into a training data set to obtain a new training data set;
and performing model training on the current convolutional network model in the current identity recognition model based on the new training data set.
6. The gait information-based user identification method according to claim 5, wherein the clustering of each of the target gait data to obtain a plurality of data categories comprises:
and periodically acquiring each target gait data in the preset time interval according to the preset time interval, so that the current convolutional network model in the current identity recognition model is periodically updated on the basis of each target gait data.
7. A method of gait information based user identification according to any of claims 2-3, characterized in that the current convolutional network model comprises a ResNet network model and the classification model comprises a KNN model.
8. A user identification device based on gait information, comprising:
the gait data acquisition module is used for acquiring gait data of the user to be identified to acquire the gait data;
the identification result acquisition module is used for carrying out user identification processing on the gait data based on a current identity identification model obtained by pre-training to obtain an identification result;
the model updating module is used for adding a user tag to the gait data under the condition that the identification result does not contain a target user tag; model training is carried out on the current identity recognition model based on the gait data and a corresponding user label, an updated current identity recognition model is obtained, and user recognition is carried out on the gait data based on the updated current identity recognition model;
and the user identity determining module is used for obtaining the user identity of the user to be identified based on the target user label under the condition that the identification result contains the target user label.
9. An electronic device, characterized in that the electronic device comprises:
one or more processors;
a storage device for storing one or more programs,
when executed by the one or more processors, cause the one or more processors to implement the gait information based user identification method of any one of claims 1-7.
10. A storage medium containing computer-executable instructions, which when executed by a computer processor, are configured to perform a method of user identification based on gait information according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211110679.0A CN115527266A (en) | 2022-09-13 | 2022-09-13 | User identification method and device based on gait information, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211110679.0A CN115527266A (en) | 2022-09-13 | 2022-09-13 | User identification method and device based on gait information, electronic equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115527266A true CN115527266A (en) | 2022-12-27 |
Family
ID=84696974
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211110679.0A Pending CN115527266A (en) | 2022-09-13 | 2022-09-13 | User identification method and device based on gait information, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115527266A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116821778A (en) * | 2023-08-30 | 2023-09-29 | 之江实验室 | Gait recognition method and device based on WIFI signals and readable storage medium |
-
2022
- 2022-09-13 CN CN202211110679.0A patent/CN115527266A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116821778A (en) * | 2023-08-30 | 2023-09-29 | 之江实验室 | Gait recognition method and device based on WIFI signals and readable storage medium |
| CN116821778B (en) * | 2023-08-30 | 2024-01-09 | 之江实验室 | Gait recognition method and device based on WIFI signals and readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109583332B (en) | Face recognition method, face recognition system, medium, and electronic device | |
| EP2795538A1 (en) | A method, an apparatus and a computer software for context recognition | |
| CN112071322B (en) | End-to-end voiceprint recognition method, device, storage medium and equipment | |
| CN109255289B (en) | Cross-aging face recognition method based on unified generation model | |
| US10423817B2 (en) | Latent fingerprint ridge flow map improvement | |
| CN112070058A (en) | Face and face composite emotional expression recognition method and system | |
| CN111898550B (en) | Expression recognition model building method and device, computer equipment and storage medium | |
| CN112926403A (en) | Unsupervised pedestrian re-identification method based on hierarchical clustering and difficult sample triples | |
| CN114741517A (en) | Training method, device, equipment and medium of text classification model and text classification method, device and equipment | |
| CN115527266A (en) | User identification method and device based on gait information, electronic equipment and storage medium | |
| CN112668482A (en) | Face recognition training method and device, computer equipment and storage medium | |
| CN106971203B (en) | Identity recognition method based on walking characteristic data | |
| US10755074B2 (en) | Latent fingerprint pattern estimation | |
| CN111738009B (en) | Entity word label generation method, entity word label generation device, computer equipment and readable storage medium | |
| CN115312033A (en) | Speech emotion recognition method, device, equipment and medium based on artificial intelligence | |
| CN113657248A (en) | Training method and device for face recognition model and computer program product | |
| CN113743239A (en) | Pedestrian re-identification method and device and electronic equipment | |
| CN117113146A (en) | Main body classification processing method, related device and medium | |
| CN113177479B (en) | Image classification method, device, electronic equipment and storage medium | |
| WO2022252089A1 (en) | Training method for object detection model, and object detection method and device | |
| WO2021151354A1 (en) | Word recognition method and apparatus, computer device, and storage medium | |
| CN114373212A (en) | Face recognition model construction method, face recognition method and related equipment | |
| CN114419480A (en) | Multi-person identity and action association identification method and device and readable medium | |
| CN110059180B (en) | Article author identity recognition and evaluation model training method and device and storage medium | |
| CN112651996A (en) | Target detection tracking method and device, electronic equipment and storage medium |
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 |