34results about How to "Improve gesture recognition rate" patented technology

The invention discloses a gesture image segmentation and recognition method based on an improved capsule network and algorithm, and belongs to the technical field of computer vision and artificial intelligence, The method comprises steps of removing the background by using a proposed U-shaped residual capsule network under a complex background; segmenting gesture image, using an image processing method to remove noises and positioning the gesture position of a binarized image of the image; removing the background of an original image with the positioned gesture area as a mask, only reserving the gesture image; finally inputting the gesture image into an improved matrix capsule network, and conducting recognition through an improved algorithm. Compared with U-Net algorithm, the improved algorithm greatly reduces the parameter quantity and improves the segmentation performance of the gesture image, thereby improving the recognition rate of the gesture image.

The embodiment of the invention, which relates to the field of the data processing technology, discloses a gesture recognition method and apparatus, and an electronic device, so that a problem of low gesture recognition efficiency in the prior art can be solved. The gesture recognition method comprises: a panoramic image that includes a gesture operation and is shot by a single camera from a panoramic assembly is obtained; distortion correction is carried out on the panoramic image by using a distortion correction algorithm corresponding to the panoramic assembly to obtain a correction image; blocking operation is carried out on the correction image to obtain a block image; and a gesture feature value of the block image is extracted and a command of the gesture operation is determined based on the gesture feature value. In addition, the embodiment of the invention also discloses a gesture recognition apparatus and an electronic device. Therefore, the gesture recognition efficiency can be improved.

The invention discloses a gesture recognition method fusing myoelectricity and micro-inertial measurement unit multi-modal signals, which comprises the following steps: acquiring myoelectricity data and motion data by using a myoelectricity electrode and a micro-inertial measurement unit, performing synchronous processing on the myoelectricity data and the motion data, and dividing a training set and a test set; dividing each signal segment into a plurality of sub-signal segments with fixed lengths by using a sliding window, and respectively extracting time domain and frequency domain features from the myoelectricity data and the motion data of each sub-signal segment; and respectively extracting shallow and deep features of the myoelectricity features and the motion features by using a convolutional neural network, respectively fusing the shallow and deep features, inputting the fused features into a classification network, finally fusing and outputting the probability of each gesture category in a decision-making layer, training a recognition model, and testing to obtain a gesture recognition rate. According to the gesture recognition method fusing the myoelectricity and the multi-modal signals of the micro-inertial measurement unit, respective advantages of the myoelectricity and the motion data can be fully utilized, so that various different gestures of the same subject can be recognized more accurately.

The invention discloses a hand motion recognition method based on a depth image and a color image. The method comprises the following steps: taking 36 types of gestures of an ASL sign language libraryas templates, obtaining gesture data through a Kinect sensor, and building a gesture database under the depth and color backgrounds; a regression-based target detection algorithm SSD is used as a research basis; under a Tensorflow deep learning framework, transfer learning is carried out on a selected target detection model by utilizing a gesture database self-built based on color and depth backgrounds respectively to obtain two types of network models capable of carrying out recognition detection on hand movement under the depth and color backgrounds. A hand motion recognition detection network framework with detection results fused under depth and color backgrounds is utilized, a non-maximum suppression algorithm is improved, and finally the effectiveness of hand motion recognition detection of the proposed network framework is obtained. According to the invention, the problems of missing detection and false detection of the target are avoided, the gesture recognition rate is improved, and single-hand recognition and double-hand recognition can be realized.

The invention relates to a cultural tourism service based holographic VR helmet, and belongs to the technical field of VR. On the basis of a cultural tourism service, a user can make interaction withthe VR helmet conveniently via gesture recognition, and control is convenient; a correction module corrects distortion of a collected image and then identifies and confirms the corrected image, so that the gesture recognition rate is improved; transverse and vertical fixed bands are arranged to provide convenience for mounting, dismounting and tightness adjustment, so that the helmet can be suitable for more people and people can use the helmet more comfortably; a driving mechanism drives VR glasses to rotate around a rotating shaft, the VR glasses are overturned up and down, the user is prevented from the trouble in wearing and taking off the glasses, a more comfortable user experience is provided, and operation is convenient; and a helmet body of a double-layer structure includes a cavity, a cooling medium is arranged in the cavity for heat radiation, so that the helmet is protected, and the user wears the helmet comfortably and is prevented from damage caused by high temperature when wearing the helmet for a long time.

The invention discloses a wrist type intelligent device. The wrist type intelligent device comprises a belt part and a wrist type intelligent terminal, the wrist type intelligent terminal comprises an input device, an output device, a computing device, a storage device and a power supply battery, the input device carries out infrared image acquisition on a hand wearing the device, and the computing device processes images, so that gesture operation information is obtained, different gesture operation information corresponds to different control commands, therefore the control over the wrist type intelligent device or other equipment connected with the wrist type intelligent device is completed with one hand, and the operation convenience and controllability are greatly improved.

The invention relates to a method for constructing a multi-channel myoelectricity feature image data set. The method comprises the following steps: firstly, combining a threshold comparison method with a Butterworth filter to complete preprocessing of an original electromyographic signal; then, performing screening twice from a common time domain and frequency domain features, and selecting four non-redundant myoelectricity features for feature extraction; secondly, generating a myoelectricity image based on a mapping relation between a one-dimensional signal and a two-dimensional image; and finally, completing construction of a multi-channel myoelectricity feature image according to an image splicing mode. A myoelectricity image data set is trained through the deep learning network, so that the gesture recognition rate can be effectively improved. The multi-channel myoelectricity feature image has richer feature information, complementation among information can be completed through multiple features, and the final recognition rate is 8%-9% higher than that of a single-channel myoelectricity feature image.

The present invention discloses a control method and an electronic device. The control method includes: determining a first operation area where an operating body performs gesture operation based on the sensing information acquired by the at least two sensing units, wherein, based on the at least The two sensing units can acquire sensing information of different spatial regions, the first operating region includes at least one of the different spatial regions; at least based on the configuration relationship between the operating region and the instruction, generate and execute the A first control instruction corresponding to the first operation area, wherein the first control instruction is different from a second control instruction generated when the operating body performs a gesture operation in the second operation area.

The invention discloses a gesture image segmentation and recognition method based on an improved capsule network and algorithm, which belongs to the technical field of computer vision and artificial intelligence. Under complex backgrounds, the proposed U-shaped residual capsule network is used to remove the background and segment gesture images , and then use the image processing method to remove the noise and locate the gesture position of the binarized image. Third, use the located gesture area as a mask to remove the background of the original image, and only keep the gesture image. Finally, the gesture image Input to the improved matrix capsule network and use the improved algorithm for identification. Compared with the U-Net algorithm, the improved algorithm greatly reduces the amount of parameters, improves the segmentation performance of gesture images, and thus improves the recognition rate of gesture images.

The invention discloses an isotropic three-dimensional gesture recognition method based on feature selection. The existing 3D gesture recognition algorithm does not consider the contribution of the extracted gesture-related features to the classification, and the redundant features affect the recognition rate. The present invention extracts 24 features from the collected three-dimensional coordinate data of gestures and inputs them into the random forest model, arranges the importance scores of each feature obtained by the training model from large to small, and selects among the 24 features arranged in k groups of each gesture The first n features of each group are combined into a combination feature. Based on the ten-fold cross-validation method and the Gaussian Naive Bayesian recognition model, the recognition rate of the Gaussian Naive Bayesian recognition model under 24 groups of combined features is obtained; The recognition rate of the naive Bayesian recognition model determines the selection of the combined features composed of the first few features for the final recognition model. The invention not only reduces the amount of feature-related data collection, simplifies the model calculation, but also improves the recognition rate.

The embodiment of the invention discloses a black screen gesture recognition method, device, storage medium and mobile terminal. The method includes monitoring the switching state of each black screen gesture in the application layer, and updating the black screen gesture switch in the driver layer according to the switching state; matching the turned on black screen gesture with the black screen gesture in the preset confusing gesture group, Judging according to the matching result whether to change the recognition condition of the opened black screen gesture; if so, then adopting the preset second recognition condition to replace the first recognition condition of the opened black screen gesture in the driver layer, and according to the second recognition The condition recognizes the gesture type of the blank screen gesture entered by the user. The above technical solution responds to the enabled black screen gestures according to the user's configuration of the black screen gestures, and makes targeted adjustments to the recognition conditions of the enabled black screen gestures, which can effectively avoid the occurrence of the low recognition rate of the current black screen gestures. Improved gesture recognition rate.

The embodiment of the invention discloses a gesture recognition method, device and electronic device, which relate to the technical field of data processing and can solve the problem of low gesture recognition efficiency in the prior art. The gesture recognition method in the embodiment of the present invention includes: obtaining a panoramic image including gesture operations captured by a single camera from a panoramic component; using a distortion correction algorithm corresponding to the panoramic component to perform a distortion correction operation on the panoramic image to obtain the correction performing a block operation on the corrected image to obtain a block image; extracting a gesture feature value of the block image, and determining a gesture operation command based on the gesture feature value. In addition, the embodiment of the invention also discloses a gesture recognition device and electronic equipment. Through the solutions of the embodiments of the present invention, the efficiency of gesture recognition can be effectively improved.

The invention relates to a monocular gesture recognition method under complex background and illumination. The monocular gesture recognition method comprises the following steps: S1, firstly performing skin color recognition on an obtained image so as to obtain a preliminary hand image; S2, then performing mobile object detection on the treated preliminary hand image by using an inter-frame difference method so as to obtain a more complete hand outline image; S3, establishing an outline convex hull model of the more complete hand outline image which is obtained in the step S2, judging gestures, and further completing hand recognition under the complex background so as to obtain a hand model; S4, finally calculating the center of the gravity of the hand model extracted in the step S3 so as to obtain a kinematic trajectory, and completing dynamic gesture recognition under the complex background. According to the monocular gesture recognition method disclosed by the invention, influence of illumination can be reduced, the rate of judging, reading and recognizing the gestures can be increased, and the gesture recognition can be performed under the complex background.

The invention discloses a hand gesture recognition method based on multi-feature fusion and fingertip detecting. The method comprises a training process and a recognition process. In the training process, for a complex hand gesture, reasonable hand gesture features are selected, a multi-feature fusion feature extracting algorithm is used, the hand gesture is subjected to support vector machine training, and a training model is formed. In the recognition process, for an input video image sequence, hand gesture detecting is carried out first, then multi-feature extracting and fusion are carried out, and multiple features are input into the support vector machine to obtain a recognition results. Meanwhile, the hand gesture is subjected to fingertip detecting based on defects, through a defect screener, the positions of fingertips of fingers are located, then two-time recognition and detecting results are subjected to synthesized, and the final hand gesture recognition results are obtained. The problem that in a complex scene, the hand gesture recognition rate is not high can be effectively solved, the requirement of real-time performance is met, and the method can be well used in human-machine interaction.

The invention provides a gesture control device and a gesture recognition method. The gesture control device comprises a computation terminal and a plurality of sensors, wherein the computation terminal is used for carrying out offline natural gesture modeling and online gesture recognition; the sensors are respectively arranged at the small arm, big arm and trunk of a tested object so as to correspondingly acquire the gesture coordinates of the small arm, big arm and trunk; and the sensors are connected with the computation terminal in a communication manner. In the online gesture recognition process, the computation terminal carries out computation processing on the basis of data acquired by the sensors, and the actual state, obtained through the computation processing, of the tested object is automatically compared with a gesture model obtained in the offline natural gesture modeling process of the computation terminal to complete the gesture recognition of the tested object. In the gesture control device, the computation terminal can process the repeated actions, so that the offline natural gesture modeling process and the online gesture recognition process are simplified, the gesture recognition rate is improved, and the control of natural gestures close to the human habits becomes possible.

The invention provides a gesture image segmentation and recognition method and device based on deep learning. According to the method, firstly, a gesture image is preprocessed, so that the size of the image is fixed; secondly, in a complex background, a dense segmentation network is used for densely connecting hole convolution with different hole rates to obtain gesture multi-scale information on different visual fields, so that the accuracy of feature expression is improved; besides, in order to fuse details and spatial position information on different levels and improve the segmentation performance of the whole network, the dense segmentation network adopts an encoder-decoder structure, redundant background information is removed, and accurate segmentation of the gesture image is realized; and finally, the mask image which only retains the gesture image is inputted into a gesture recognition network, and recognition is conducted by adopting an improved algorithm. The segmentation performance of the gesture image can be improved, so that the recognition rate of the gesture image is improved.

An interactive architectural design platform based on hand motion capture, including a hand information capture system, an information processing conversion system, a processing and processing output module, and an architectural design platform. The architectural design platform communicates with the processing and processing output module, and includes a variety of In construction software, the processing output module includes a learning module and a motion compensation module. The design platform is equipped with a hand information capture system of infrared sensors and 3D scanners to capture hand movements and gestures, and uses the information processing and conversion system to perform data recognition and conversion, and the processing and output module according to the set gestures and commands Group, control the architectural design platform to make corresponding actions, so as to realize rapid drawing and modeling, and improve efficiency. By setting the command input action, the bionic robot finger can be fine-tuned, the gesture range of the bionic robot hand can be expanded, and the parameters can be adjusted. All captures are performed to improve the gesture recognition rate.