33 results about "Nearest neighbor classifier" patented technology

The invention discloses an image classification method for small sample class incremental learning based on prototype features. The method comprises the following steps of: 1, training an end-to-end network based on a ResNet framework by using image data # imgabs0 # of a base class task; 2, for the end-to-end network, discarding a fully connected classifier of the end-to-end network, adopting a distance-based nearest neighbor classifier, and freezing parameters of a feature extractor; 3, inputting the image features into a prototype correction module, and optimizing the prototype correction module through a cross entropy loss function and a prototype comparison loss function; 4, continuously training a prototype correction module by using the image data, and generating a new category of prototype features with higher discrimination by the prototype correction module; and 5, classifying the distance between the image features and the prototype features by using a nearest neighbor classifier. According to the method, the distribution of the classes in the feature space is accurately described, so that the image classification precision in a small sample class incremental learning scene is remarkably improved.

The invention relates to the technical field of control and adjustment systems, in particular to an LED display matrix dynamic control system, which comprises a time sequence feature extraction module, a self-adaptive threshold calibration module, a self-adaptive threshold calibration module, a self-adaptive threshold calibration module, a self-adaptive threshold calibration module, a self-adaptive threshold calibration module and a self-adaptive threshold calibration module, wherein the time sequence feature extraction module is used for collecting time domain waveforms of scanning signals and data signals; a display scene is recognized based on a k-nearest neighbor classifier, a composite reference value is generated in combination with historical optimization parameters, a hierarchical compensation execution module is used for executing a three-level compensation strategy according to the composite reference value to correct time sequence deviation, a dynamic model updating module is used for optimizing scene recognition and threshold generation parameters through a genetic algorithm, and a scene library is periodically updated. All the modules cooperate to form a closed-loop link, static, dynamic, high-speed motion and temperature change scenes are dynamically adapted, and the time sequence synchronous control precision and the system robustness of the row driving circuit and the column driving circuit are improved.

The invention discloses a small sample HSI classification method based on prototype mask comparison and cross-stage feature refinement, and the method comprises the steps: carrying out the meta-task training in a source domain and a target domain in a crossing manner, carrying out the spectral dimension unification of samples of a support set and a query set in each meta-task, carrying out the multi-dimensional feature fusion resistant to the dependence of adjacent pixels, and carrying out the classification of the multi-dimensional features. Obtaining support features and query features; constructing positive and negative sample pairs representing a prototype under each meta-task, and performing prototype mask comparison through graph convolution; according to the method, feature redundancy elimination refining is carried out based on emphasis mapping of a cross-domain shared multi-layer perceptron, cross-stage feature refining is carried out on support features and query features through the multi-layer perceptron and directional calibration coordination, refined support features and refined query features are obtained, and different losses are optimized in two training stages; in the test stage, small samples of a target domain are used, and a K-nearest neighbor classifier is trained for classification prediction through features extracted after dimensionality unified mapping, a feature extraction module and a multi-layer perceptron in sequence.

The invention discloses a face recognition method based on adaptive sine angle robust principal component analysis, and belongs to the technical field of face recognition, and the method comprises the steps: obtaining a face image, and carrying out the preprocessing of the face image; constructing a robust principal component analysis model based on an adaptive sine angle, and training the robust principal component analysis model by using the training set; solving an optimal projection matrix by using a non-greedy iterative algorithm; mapping the original face features to a discrimination space by using the optimal projection matrix; and performing face recognition by using a nearest neighbor classifier, and performing robustness evaluation on a recognition result. According to the method, the adaptive sine angle is added, abnormal values can be effectively suppressed in the face of noise data, the influence of the abnormal values on the projection direction is reduced, and the method plays a key role in improving robustness.

The invention discloses a memory enhancement action recognition method and system on a Riemannian manifold and a storage medium, and the method comprises the steps: 1, collecting human body action data, and representing the human body action data as a third-order tensor; 2, expanding along three modes to obtain three corresponding matrixes; 3, calculating by adopting a human short-term memory mechanism to obtain a memory enhanced weight matrix; 4, decomposing the weight matrix through a principal component analysis method to obtain a base vector with a weight; 5, recombining, normalizing and mapping to a unit hyper-sphere, and reserving an angle relation; 6, learning modal weight parameters through a Monte Carlo Markov algorithm, and calculating geometric differences between points on the hypersphere; and 7, carrying out human body action classification by adopting a K-nearest neighbor classifier. The method effectively solves the problem of time information loss in a complex scene, and is suitable for various application scenes such as medical health, virtual reality, physical training and the like.

The invention discloses a small sample automatic modulation identification method and system based on a multi-stage regularization Y-shaped frame in the technical field of wireless communication signal processing. The method comprises the following steps: carrying out Gramer angle field transformation on a received I / Q signal sequence to obtain a two-dimensional GAF image set; performing data enhancement on the two-dimensional GAF image set, dividing the enhanced GAF images with known modulation category labels in the enhanced GAF image set into a support set, and dividing all the remaining enhanced GAF images with unknown modulation category labels into a query set; performing feature extraction on the support set and the query set by using a trained feature extraction model based on a channel space attention relation network CSARN to obtain a to-be-identified feature vector of the support set and a to-be-identified feature vector of the query set, and performing modulation mode classification by using a nearest neighbor category mean classifier to obtain a to-be-identified feature vector of the support set and a to-be-identified feature vector of the query set; and obtaining modulation identification results of the enhanced GAF images of all the unknown modulation category labels.

Techniques for the use of verbal commands in human-robot communication. The number of tasks the robot can perform is limited to a specific set, while providing syntactic flexibility to users. The system includes two components: a speech recognizer for speech-to-text conversion and a natural language understanding module that maps the text to a command for the robot. After speech is transcribed to text, a nearest neighbor classifier can be applied in the high dimensional space of embedding tokens. Multiple variants of each command are provided in a database of reference embeddings, and the classifier can identify the k nearest reference embedding tokens to determine the command. The text similarity model allows for quick detection solutions to be deployed locally on a robot or other device. Local deployment reduces potential latency caused by a cloud connection, which can be important in many assistant robot applications.

The invention discloses an open environment radiation source individual identification method based on manifold learning, and belongs to the technical field of radiation source individual identification. According to the method, firstly, high-dimensional signal data are mapped to a low-dimensional Grassmann manifold space, and the internal structure of the data is kept while redundant features in the high-dimensional data are removed; and then aligning the open manifold structures of the source domain and the target domain in the low-dimensional manifold space, so that the method can measure the distance between the prototypes of the source domain and the target domain in the unified manifold space. And finally, performing cross-domain prototype matching by replacing Euclidean distance with geodesic distance on a Grassmann manifold, labeling a pseudo-label for target domain data according to a matching result, and jointly using source domain label data and target domain pseudo-label data as a reference set of a nearest neighbor classifier so as to obtain a nearest neighbor classifier. And accurate identification of known and unknown types of radiation source individuals in an open environment is realized through a nearest neighbor classification strategy.

The application provides a small sample named entity recognition model training method, comprising the following steps: S1, obtaining a training set, a training set type description set, a support set and a support set type description set; S2, mining clue words in each sample on the training set and the support set respectively and performing clue word labeling to obtain the training set and the support set containing named entity labels and clue word labels respectively; S3, performing multi-round iterative training on a basic named entity recognition model until convergence by using the training set and the training set type description set processed in step S2; and S4, performing migration training on the basic named entity recognition model trained in step S3 until convergence by using the support set and the support set type description set processed in step S2, to obtain a small sample named entity recognition model composed of an encoder and a nearest neighbor classifier.

The invention discloses a small-size image recognition method, device and equipment, a medium and a product in the technical field of computer vision, and the method comprises the steps: obtaining a small-size training image, and carrying out the preprocessing of the small-size training image, and generating a training image matrix; splicing all the training image matrixes to generate an image augmentation matrix; constructing a low-rank and low-dimension combined 2DERPCA optimization model of enhanced robust principal component analysis, and inputting the image augmentation matrix into the 2DERPCA optimization model to calculate an optimal projection matrix; obtaining a to-be-recognized small-size image, and preprocessing the to-be-recognized small-size image to generate a to-be-recognized image matrix; performing feature extraction on the to-be-recognized image matrix and the training image matrix by using the optimal projection matrix to obtain corresponding low-dimensional features; inputting the low-dimensional features into a nearest neighbor classifier for identification; according to the method, feature extraction can be carried out on a picture containing noise, the influence caused by the noise is effectively suppressed, and the robustness of recognition is improved.

The present invention discloses a method, device and equipment for classifying radar echo signals, including: performing data preprocessing; selecting the nearest neighbor set when the autocorrelation degree of each sample in each pair of non-connectable constraint pairs is the largest as the corresponding set of same-class samples, thereby expanding the constraint pair set; generating a connection matrix from the expanded constraint pair set, calculating the objective function after each feature in the candidate feature subset is incorporated into the feature subset, selecting the feature index with the smallest objective function, updating the target feature subset, and generating training samples by performing feature selection on the original sample data set; using the training nearest neighbor classifier to classify unknown radar echo signals. The present invention considers that the nearest neighbor samples around each sample are different, and adaptively finds the most suitable nearest neighbor set for each sample to ensure that the connection between the elements is close, which is convenient for generating the optimal feature subset; considering the correlation between features and the local geometric structure of data, the accuracy of classifying radar echo signals is improved.

The invention discloses a medical clinical data feature extraction method, system and device and a medium, and relates to the technical field of data processing. Comprising the steps of obtaining a medical clinical data set; a moss breeding mechanism of a moss optimization algorithm is replaced by a biogeography-based learning strategy, the implicit effect of the moss optimization algorithm is removed, and dual-mode propagation search is introduced to obtain a biological learning moss optimization algorithm; a clinical feature subset most related to sample classification is searched in the feature extraction process of the medical clinical data set through the biological learning moss optimization algorithm; and carrying out data classification on the clinical feature subset most related to sample classification by adopting a K-nearest neighbor classifier to obtain an optimal clinical feature subset of the medical clinical data. According to the method, the defect that the moss optimization algorithm is insufficient in local development capability of the optimal solution is overcome, and the algorithm convergence precision is effectively improved.

The invention discloses a peanut kernel producing area identification method based on a near infrared spectrum. The method comprises the following steps: (1) collecting near infrared diffuse reflection spectrum data of peanut kernel samples from different producing areas by using a portable near infrared spectrometer; (2) carrying out preprocessing on the spectral data by utilizing a multiple scatter correction (MSC) method and a Savitzky-Golay smoothing filter; (3) extracting identification information of the spectral data by using a fuzzy identification analysis method; and (4) classifying by using a K neighbor classifier. According to the method, linear identification information of the near infrared spectrum value domain space and the null space of the peanut kernels can be extracted, the K-nearest neighbor classifier is adopted for classification, and the peanut kernel producing area can be rapidly and accurately identified. The peanut kernel producing area identification method based on the near infrared spectrum can be used for accurately identifying the peanut kernel producing area.

The application relates to the technical field of feature selection learning. The glass type recognition method provided by the application obtains a feature optimal subset by using a constructed semi-supervised feature algorithm to perform feature selection on obtained glass feature data, filters data to form a training set by using the feature optimal subset, trains a nearest neighbor classifier, and performs type recognition on unknown glass data. In the constructed semi-supervised feature algorithm, the weight of features in a feature data set is obtained by calculating a constraint pair, the change of sample features in a weighted feature space is considered, more effective local information is obtained, and the classification performance of the model on glass is improved on the premise of saving cost.

The invention specifically discloses a rotary machine fault diagnosis method and equipment based on multi-graph cooperation of an intrinsic model, and belongs to the technical field of fault diagnosis. According to the method, while multi-manifold structure features of a data set are fully considered, a graph embedding thought is introduced, dimensionality reduction is performed on a fault feature set, and a processed low-dimensional feature set is input into a K-nearest neighbor classifier, so that fault mode identification is realized. According to the method, while high-dimensional nonlinear fault feature information is effectively extracted, the difficulty of equipment fault classification is reduced, and the accuracy of fault identification is improved. Besides, the constructed diagnosis model keeps good fault identification accuracy in different noise environments and training environments, the fault identification capability is effectively improved, and the method has good dimension reduction effect, noise resistance and stability.

The invention relates to the technical field of image processing, and relates to a palmprint recognition method and device based on a GPGPU, and a medium. The method comprises the following steps: acquiring an original palmprint image through a data acquisition system; preprocessing the original palmprint image to extract an interested palmprint area; the preprocessed palm print image is processed by a feature extraction system, and palm print features are extracted by using an improved KL transformation algorithm; the feature vectors are recorded and classified, palm print features are classified through a nearest neighbor classifier, and palm print recognition is achieved; the collected original palmprint image data, the preprocessed palmprint image data and the palmprint feature data after feature extraction are stored in a distributed storage system of the GPGPU, and the data storage and access process is optimized by using the parallel computing capability of the GPGPU and multi-channel transmission resources. According to the method, the parallel processing advantage of the GPGPU is utilized, and the operation efficiency can be remarkably improved when the palmprint image is processed.

The invention discloses a therapeutic peptide prediction method based on a nearest neighbor classifier, and belongs to the field of biological information. According to the method, peptide sequence features are extracted from QSP740 and CPP400 data sets by using a UniRep model, a kernel risk sensitive loss function is fused on the basis of a K-nearest neighbor algorithm to construct an objective function, and a multi-Laplacian matrix is constructed in combination with three similarity matrixes of an RBF function, cosine similarity and a Pearson's correlation coefficient to enhance model robustness. And finally, classifying the peptide sequence to be predicted by minimizing the objective function and evaluating the performance. According to the method, features are automatically extracted through deep learning, the model robustness is improved through a kernel risk loss function, and the uncertainty of data is processed by means of multi-view Laplace regularization, so that compared with an existing method, the treatment peptide type can be more accurately predicted, and the classification precision and the anti-noise capability can be improved.

Systems and methods for predicting an epileptic seizure with a sufficient window time before its onset to allow the patient to locate a safe place or take appropriate precautionary actions. The method is presented through adroit integration of maximal overlap wavelet packet transform, homogeneity index, and a K-Nearest Neighbors classifier to predict an epileptic event twenty minutes before its onset using electrocardiogram (ECG) signals. The effectively predicts an epileptic seizure 20 minutes prior to its onset with an accuracy of 93.25%.

This invention discloses a method, system, device, and medium for feature extraction from medical clinical data, relating to the field of data processing technology. It includes: acquiring a medical clinical dataset; replacing the moss reproduction mechanism of the moss optimization algorithm with a biogeographical learning strategy, removing the cryptic effect of the moss optimization algorithm, and introducing a bimodal propagation search to obtain a bio-learning moss optimization algorithm; using the bio-learning moss optimization algorithm to find the subset of clinical features most relevant to sample classification during the feature extraction process of the medical clinical dataset; and using a K-nearest neighbor classifier to classify the subset of clinical features most relevant to sample classification, thereby obtaining the optimal subset of clinical features for the medical clinical data. This invention overcomes the deficiency of the moss optimization algorithm in its insufficient ability to exploit local optimal solutions, effectively improving the algorithm's convergence accuracy.

The invention discloses an anti-cancer peptide recognition method and system based on multi-feature fusion and double-layer integrated learning. The method comprises the steps of S1, data preprocessing; s2, feature extraction and feature fusion based on a protein language model; s3, feature extraction; s4, performing dimension reduction processing on the high-dimensional features; s5, inputting each feature vector in the multi-source feature set into a corresponding XGBoost classifier for training and prediction, wherein each classifier outputs a peptide sequence as a preliminary prediction probability of the anti-cancer peptide; combining the preliminary prediction probabilities output by all classifiers into a probability feature vector; s6, inputting the probability feature vectors obtained from the upper layer into a K-nearest neighbor classifier and a soft voting classifier at the same time; the KNN outputs a first prediction probability, and the soft voting integrator outputs a second prediction probability; s7, calculating an arithmetic mean value of the first prediction probability and the second prediction probability as a prediction probability; and comparing the peptide sequence with a preset threshold value, if the peptide sequence is greater than or equal to the threshold value, determining that the peptide sequence is an anti-cancer peptide, otherwise, determining that the peptide sequence is a non-anti-cancer peptide.

The present invention discloses a method and device for voice forgery identification based on dynamic multi-view fusion. The method includes the following steps: obtaining multi-view dynamic feature data according to the original voice signal to construct a multi-view data set; clustering the data in the multi-view data set by using a dynamic weight clustering ensemble algorithm to generate pseudo-labels; integrating the pseudo-labels with the multi-view dynamic feature data to obtain a first enhanced data set and training a K-nearest neighbor classifier; using the trained K-nearest neighbor classifier to predict unseen data points other than the original voice signal in the voice signal to be identified. The present invention uses unsupervised learning to mine multi-perspective voice information, solving the deficiencies of existing methods in generalization, robustness, and feature utilization. The present invention has excellent performance on multiple data sets, with significant improvement in key indicators, strong generalization and robustness against different forged data points, providing a new path for voice deep forgery detection.

The invention belongs to the technical field of behavior recognition. The invention provides a human body action recognition method for mobile equipment based on a pulse neural network. According to the embodiment of the invention, the built-in sensor on the mobile equipment is used for collecting the time sequence data of the user, and the time sequence data is normalized to a fixed range suitable for feature extraction. The time slicing unit divides time series sensor data into small segments or time windows to capture time patterns, and the position coding unit indicates the sequence or position of data in each time window to ensure that the data is fully prepared for feature extraction. The attention mechanism coordinating unit generates attention weight of position sensing by decomposing global pooling into feature codes of time dimension and sensor dimension. The ANN network extracts relevant features from the input data, and the ANN-SNN mapping unit maps the features into the SNN network to extract time features. And a one-dimensional K-nearest neighbor classifier is used for classification, so that identification of user activities is realized.

A KNN-based lateral stability control method for heavy-duty AGVs is disclosed, characterized by the following steps: 1. Classifying AGV operating conditions into five categories based on different loads, and collecting motion state data of the AGV under different operating conditions using simulation software; 2. Establishing a K-nearest neighbor (KNN) classifier, and training and validating it using the collected dataset; 3. Designing a set of nonlinear sub-controllers based on fuzzy PID, calculating the required yaw moment, and distributing the torque to the four drive wheels according to the torque distribution rule; 4. Introducing an error judgment strategy, activating the controller based on the centroid sideslip angle error to control its lateral stability. This invention establishes a load KNN classifier for heavy-duty AGVs, which can monitor the centroid sideslip angle error in real time. If the error exceeds a threshold, the controller is activated, the classifier's classification result is matched to the corresponding sub-controller, and an additional yaw moment is calculated and applied to the AGV.

The invention discloses a sparse null space and norm-based Chinese wolfberry origin identification model construction method and application, Chinese wolfberry of the same origin is regarded as a category, and near infrared spectrum data of Chinese wolfberry samples of a plurality of origins are collected; preprocessing the acquired near infrared spectrum data to eliminate noise and scattering influence in the spectrum data; carrying out dimension reduction processing on the preprocessed near infrared spectrum data, and compressing the dimension of the spectrum data; performing multi-time projection processing on the spectral data in a sparse null space and a norm; and training a K-nearest neighbor classifier by using a sample obtained after projection, constructing a Chinese wolfberry origin identification model, and performing testing. The wolfberry samples can be classified by using the constructed wolfberry production place identification model, and a production place identification result is output; and the identification stability, accuracy and efficiency are improved.

The present invention discloses a network intrusion detection system that introduces sample geometry and multi-view information, including a process of constructing a highly separable metric space on a training set and classifying the network intrusion data to be detected using a K-nearest neighbor classifier. First, the collected intrusion data is subjected to a large margin criterion framework at each viewpoint; second, the density between heterogeneous sample clusters in the training set at each viewpoint is minimized; then, an adjacency matrix with label information is introduced for each viewpoint; and then, an adaptive maximization viewpoint cross-correlation method is adopted. The above steps are substituted into a unified learning framework, the optimal solution is calculated and the final metric space is obtained; finally, the network intrusion data to be detected is sent into the metric space and classified using a K-nearest neighbor classifier. The present invention makes up for the defect that the existing system cannot mine sample and viewpoint data information, and can make full use of sample geometry information and viewpoint complementary information to guide multi-viewpoint network intrusion detection tasks.

The invention discloses a dynamic multi-view fusion voice forgery identification method and a dynamic multi-view fusion voice forgery identification device. The method comprises the following steps: acquiring multi-view dynamic feature data according to an original voice signal, and constructing a multi-view data set; clustering the data in the multi-view data set by adopting a dynamic weight clustering integration algorithm to generate a pseudo tag; integrating the pseudo tag with the multi-view dynamic feature data to obtain a first enhanced data set, and training a K-nearest neighbor classifier; and using the trained K neighbor classifier to predict unseen data points except the original voice signals in the voice signals needing to be identified. According to the method, the multi-view voice information is mined through unsupervised learning, and the defects of an existing method in generalization, robustness and feature utilization are overcome. The method is excellent in performance on multiple data sets, key indexes are remarkably improved, generalization and robustness are high for different counterfeit data points, and a new path is provided for voice deep counterfeit detection.

The invention discloses a method and device for classifying carbon emission data of a thermal power plant and electronic equipment, and the method comprises the steps: obtaining a plurality of weighted Euclidean distances, and screening out candidate Euclidean distances meeting a preset distance condition from the weighted Euclidean distances; performing trend weighting processing on the candidate Euclidean distance to generate a corresponding trend weighted Euclidean distance; performing matching processing on the candidate Euclidean distance and the trend weighted Euclidean distance, and determining a target Euclidean distance corresponding to the first time sequence and the second time sequence; and based on the target Euclidean distance, using a nearest neighbor classifier to divide the subsequence categories of the thermal power plant carbon emission data, realizing classification of the thermal power plant carbon emission data, and based on a classification result, analyzing and controlling the production data related to the thermal power plant carbon emission. The device comprises a processor and a memory. The electronic equipment comprises a processor, a memory, a network interface, a display screen and an input device which are connected through a system bus. According to the method, different types of thermal power plant carbon emission data can be classified, and the safety and reliability of industrial production are improved.

The invention relates to a zero-sample speech synthesis method based on denoising non-parametric speaker adaptation, and belongs to the technical field of natural language processing. The method is used for a high-fidelity zero-sample speaker to adapt to speech without a speaker, and zero-sample speech synthesis faces the problems that a model is usually trained on a small data set, and the generalization ability is poor; when the condition depends on noise reference, unseen speaker identities are difficult to model; according to the zero-sample speech synthesis method based on denoising non-parametric speaker adaptation, speaker embedding is predicted by using a nearest neighbor classifier on a database of a large number of cache examples, similarity search is carried out by using representation of a speaker encoder, and the method does not need extra training and is high in robustness. Therefore, a highly robust model is formed, and the performance can be improved when the method depends on noise reference.

The present invention discloses a memory-enhanced action recognition method and system on a Riemannian manifold, and a storage medium. The method comprises the following steps: step 1: collecting human motion data and expressing it as a third-order tensor; step 2: expanding along three modes to obtain three corresponding matrices; step 3: calculating a memory-enhanced weight matrix using the human short-term memory mechanism; step 4: decomposing the weight matrix using the principal component analysis method to obtain weighted basis vectors; step 5: reorganizing and parallelizing the normalized mapping to a unit hypersphere, preserving the angular relationship; step 6: learning the modal weight parameters using the Monte Carlo Markov algorithm and calculating the geometric differences between points on the hypersphere; and step 7: using a K-nearest neighbor classifier to classify human motions. The present invention effectively solves the problem of temporal information loss in complex scenarios and is suitable for a variety of application scenarios such as medical health, virtual reality, and sports training.

The present invention discloses a small-sample HSI classification method based on prototype mask comparison and cross-stage feature refinement. The method comprises: performing meta-task training in a cross-source domain and a target domain, unifying the spectral dimensions of samples of a support set and a query set for each meta-task, and then performing multi-dimensional feature fusion that is resistant to neighboring pixel dependence to obtain support features and query features; constructing positive and negative sample pairs representing prototypes under each meta-task and performing prototype mask comparison through graph convolution; performing feature refinement to eliminate redundancy based on emphasized mapping of a cross-domain shared multi-layer perceptron, coordinating cross-stage feature refinement of support features and query features with the multi-layer perceptron and directional calibration to obtain refined support features and refined query features, and optimizing different losses in the two training stages; and using a small sample from the target domain in the testing stage, sequentially performing dimension unification mapping, a feature extraction module, and features extracted after the multi-layer perceptron to train a K-nearest neighbor classifier for classification prediction.