298results about How to "Improve fitting ability" patented technology

The invention relates to an external skeleton robot, in particular to a human-simulated external skeleton robot assisting the lower limbs. The human-simulated external skeleton robot assisting the lower limbs aims to solve the problems that an existing external skeleton robot is low in coupling degree of motion space and poor in wearing comfort, reliability and adaptation, and power needed by a motor is large. The human-simulated external skeleton robot assisting the lower limbs comprises an upper body back part, a left leg and a right leg. The left leg and the right leg respectively comprise a hip drive system, a knee drive system and a foot wearing system. A rear side connection board of the waist is in rotating connection with a load installation board. Each hip joint supporting board is provided with a first motor and a first reducer, wherein the first motor is provided with an encoder, and the output end of the first motor provided with the encoder is connected with the input end of the first reducer. Each hip joint connecting board can rotate in the vertical plane. Each thigh stretching board is in detachable connection with the corresponding hip joint connecting board. The output end of a main drive mechanism is connected with each crus connecting board. The lower surfaces of elastic boards are bonded with the upper surfaces of the rubber soles of the feet. The human-simulated external skeleton robot assisting the lower limbs can assist in walking.

The present invention relates to the upper U of an athletic shoe comprising a first opening 1 out of which the foot sticks upwards during wearing and a second opening 2 which is closed with a tongue. The two openings 1, 2 are continuous with each other, and the upper U comprises a medial side stretchable part 51 which covers a part of a medial side face of an arch of the foot, a lateral side stretchable part 50 which covers a part of a lateral side face of a ball of a little toe (a fifth toe), a front part 3 located forward of the two stretchable parts 50, 51 and a rear part 4 located backward of the two stretchable parts. Each of the stretchable parts 50, 51 is provided so as to essentially cross one of side faces of the upper U from a top surface of a sole S to the second opening 2. Young's modulus along the lengthwise direction L of the foot of each of the stretchable parts 50, 51 is set smaller than that of the front part 3 and the rear part 4.

The invention provides a machine learning-based stereoscopic image quality objective assessment method, which comprises the following steps of: extracting parameters based on which the image quality objective assessment of stereoscopic images is performed; performing machine learning by using the parameters extracted from standard image sequences in an image library; performing fitting between the stereoscopic image qualitative assessment results and the extracted parameters by using the learning results; and applying the fitting results to an image to be assessed and comparing the results with subjective assessment marks. The machine learning-based stereoscopic image quality objective assessment method provided by the invention has the beneficial effect that based on visual characteristics of human eyes, improvements on the form and weighted values of the formula for solving PSNR (Peak Signal to Noise Ratio) and SSIM (Structural Similarity ) are achieved, and a computation method of visual comfort and an assessment method using fused image parameters are set forth, and as joint fitting is performed by using a plurality of characteristic parameters, the fitting effect is improved.

[Problems] To prevent a hip cover portion of a back-side outer sheet from swelling and curling[Means for Solving Problems] In an underpants type disposable diaper, a ventral-side outer sheet 12F and a back-side outer sheet 12B are not connected but separated at a crotch portion, the back-side outer sheet 12F has a main unit section 13 that corresponds to joined sections 12A in an up-down direction and a back-side extension section 14 that extends below the back-side main unit section 13, the ventral-side outer sheet 12F is composed of only a ventral-side main unit section that corresponds to the joined sections 12A in the up-down direction, the back-side extension section 14 has a central portion 14M in the width direction overlapping the absorber 20 and hip cover portions 14C extending on both sides of the central portion 14M, in the back-side main unit section 13, first elongated resilient and elastic members 15 are fixed in a state of being extended in the width direction at a predetermined extension ratio; second elongated resilient and elastic members 16 are fixed to the hip cover portions 14C in a state of being extended in the width direction at a predetermined extension ratio, fourth elongated resilient and elastic members having high contracting forces are fixed to the lower end portion of the ventral-side outer sheet 12F in a state of being extended in the width direction so that the contraction forces are balanced between the ventral side and back side at the lower ends of the joined section 12A.

The invention discloses a deep learning control planning method of movement routes of a robot in an intelligent environment. The method comprises steps of step 1, constructing a global map three-dimensional coordinate system for a carrying region of a carrying robot, and acquiring a walkable region coordinate under the global map three-dimensional coordinate system; step 2, acquiring a training sample set; step 3, constructing a global static route planning model of the carrying robot; and step 4, inputting a start point and destination coordinate in a transmission task into the global static route planning model based on a fuzzy neural network and acquiring an optimal planed route corresponding to the carrying robot. According to the invention, by establishing the global static route planning module and a global dynamic obstacle avoiding planning model, and using quite strong non-linear fitting characteristics of the deep learning, the global optimal path can be quickly found and a problem is solve that the normal route planning often falls into the local optimization.

A disposable diaper having a neat appearance. The diaper 101 includes: a liquid-permeable top sheet 111 which covers a use surface side; a leakage preventing sheet 112 which covers a non-use surface side; an absorbent body 113 interposed between the top sheet and the leakage preventing sheet; and an external sheet 120 disposed on an outer surface side of the leakage preventing sheet, wherein three-dimensional gathers BS are formed along leg surrounding portions, and leg cut-out portions of the external sheet which form leg openings are located in a portion of the minimum width of a crotch portion, at a position nearer to a central side than places outward by 5 mm from rising start points of the three-dimensional gathers BS.

An earpiece including a tubular core portion attached to an end of an electro-acoustic transducer and extending in an acoustic wave output direction, and an external auditory canal fitting portion disposed outside the core portion to cover an end surface of the core portion and having lower hardness than the core portion is provided. This maintains appropriate flexibility of the external auditory canal fitting portion covering the end surface of the core portion and forms the core portion with appropriate hardness. This eliminates sound quality degradation due to the core portion deformation, earpiece detachment and contact of the core portion with the external auditory canal. Consequently, this improves fitting properties with the external auditory canal and attachment properties with the electro-acoustic transducer. The earpiece which improves fitting properties with the external auditory canal and attachment properties with the electro-acoustic transducer, and the electro-acoustic transducer including the same are thereby provided.

The present invention relates to the upper U of an athletic shoe comprising a first opening 1 out of which the foot sticks upwards during wearing and a second opening 2 which is closed with a tongue. The two openings 1, 2 are continuous with each other, and the upper U comprises a medial side stretchable part 51 which covers a part of a medial side face of an arch of the foot, a lateral side stretchable part 50 which covers a part of a lateral side face of a ball of a little toe (a fifth toe), a front part 3 located forward of the two stretchable parts 50, 51 and a rear part 4 located backward of the two stretchable parts. Each of the stretchable parts 50, 51 is provided so as to essentially cross one of side faces of the upper U from a top surface of a sole S to the second opening 2. Young's modulus along the lengthwise direction L of the foot of each of the stretchable parts 50, 51 is set smaller than that of the front part 3 and the rear part 4.

The invention discloses a weather phenomenon identification method based on a lightweight convolutional neural network, and belongs to the technical field of image identification. The method comprisesthe following steps: constructing a lightweight weather identification network; training a weather recognition network model; obtaining a weather picture to be identified and carrying out standardization processing; and inputting the processed data into the trained weather identification network and outputting the category of the data. The method makes full use of the advantages of the convolutional neural network in the field of large-scale image recognition, combines the ideas of deep separable convolution, attention mechanism, residual connection, transfer learning and the like, effectively reduces the calculation complexity of the model under the condition of not reducing the recognition precision, and provides possibility for the deployment of the model on small equipment.

The invention relates to an important point-based time sequence fixed segmentation algorithm. According to the main technical characteristics, the algorithm comprises the steps of performing normalization processing on time sequence data; calculating segmentation fitting errors and performing sorting according to priorities of the segmentation fitting errors from high to low; extracting first k segments with maximum fitting errors and performing sorting from long to short according to segment lengths of a time sequence; extracting first two segments sorted according to the segment lengths, and performing fitting preprocessing; and comparing symbol and value relationships of maximum values and minimum values of the fitting errors in segmentation, determining important points to perform segmentation, progressively increasing the number of the important points, and performing loop iteration until a fixed segment number is reached. The algorithm is reasonable in design, greatly reduces the fitting errors, improves the segmentation efficiency, has very good fitting effect and adaptability, and can be widely used in the fields of time sequence prediction, clustering, anomaly detection and the like.

A sports shoe has a sole S absorbing a shock of landing, an upper U covering an instep of a foot, and a fastening member 3 fitting the upper U to the instep of the foot. The upper U has a first opening 1 from which a leg extends upwardly during wearing and a second opening 2 which is closed with a tongue, and the first and second openings 1, 2 are continuous with each other in a front-rear direction, the upper comprises a medial side portion 7 covering a medial side surface of the foot, a lateral side portion 8 covering a lateral side surface of the foot, and stretchable portions 71, 81 forming a part of the medial side portion 7 and / or the lateral side portion 8, and the second opening 2 inclines toward the medial side of the foot along a ridge line of the foot as the second opening extends toward a front of the foot.

The invention discloses a lightweight fine-grained image recognition method for cross-layer feature interaction in a weak supervision scene, and the method comprises the steps: constructing a novel residual module through employing multi-layer aggregation grouping convolution to replace conventional convolution, and enabling the novel residual module to be directly embedded into a deep residual network frame, thereby achieving the lightweight of a basic network; then, performing modeling on the interaction between the features by calculating efficient low-rank approximate polynomial kernel pooling, compressing the feature description vector dimension, reducing the storage occupation and calculation cost of a classification full-connection layer, meanwhile, the pooling scheme enables the linear classifier to have the discrimination capability equivalent to that of a high-order polynomial kernel classifier, and the recognition precision is remarkably improved; and finally, using a cross-layer feature interaction network framework to combine the feature diversity, the feature learning and expression ability is enhanced, and the overfitting risk is reduced. The comprehensive performance of the lightweight fine-grained image recognition method based on cross-layer feature interaction in the weak supervision scene in the three aspects of recognition accuracy, calculation complexity and technical feasibility is at the current leading level.

The invention provides a behavior recognition system and method based on a multi-convolution kernel residual error network, a medium and equipment, and the method comprises the steps: collecting the motion physiological data or environment state data of a user, and sequentially carrying out the transmission, preprocessing and marking of the collected data to obtain a training set and a test set; constructing a behavior recognition model based on the multi-convolution kernel residual error network; classifying the training samples according to the prediction error of the training set and carrying out self-adaptive weight distribution on each class; calculating an overall loss function and partial derivatives of each parameter of the model, iteratively updating the model parameters to obtaina trained recognition model, testing by using the test set, storing the model parameters reaching the test standard, and generating a final deep learning model; inputting to-be-recognized data into the deep learning model to obtain a user behavior recognition result. According to the invention, multi-scale data features can be effectively utilized.

The invention discloses a tread contour fitting method capable of automatically extracting segmentation points, which comprises the steps of installing laser displacement sensors inside and outside a track according to a mirror symmetry mode; acquiring coordinate data of tread detection points, and converting the coordinate data of each laser displacement sensor into a coordinate system of a vertical plane parallel to the track direction; integrating the converted coordinate data corresponding to the two laser displacement sensors into the same coordinate; performing feature point extraction; determining initial segmentation intervals according to extracted feature points; performing curve fitting according to the initial segmentation intervals, and solving a fitting determination coefficient; comparing the fitting determination coefficient with a preset curve fitting determination coefficient threshold, and determining accurate segmentation points; and determining fitting intervals according to the accurate segmentation points, and performing curve fitting on each interval respectively so as to acquire a complete tread contour. The tread contour fitting method has the characteristics of automatic extraction, high fitting accuracy and high fitting speed.

The invention relates to an image prediction method based on a multi-layer convolution long-short-term memory neural network, and the method comprises the steps: extracting the change features of an inputted sequence image through a multi-layer convolution long-short-term memory neural network module, and extracting more abstract features through the deepening of single-layer convolution, therebyimproving the expression capability. In the prediction process, the extracted change features are utilized to predict a next image. The predicted image is input back to the prediction module again, sothat a next image is obtained. A single-layer convolution long-short-term memory neural network of an original method is poor in capability of extracting change features of sequence images, and showsthat structural details of the images output during prediction are fuzzy. According to the method, by deepening the convolution layer, more accurate change features are extracted, and the problem ofstructural detail blurring of the predicted image is obviously improved.

A cylindrical sliding bearing having a cylindrical steel back metal, and a bearing alloy layer, serving as a sliding surface, on the inside of the steel back metal is provided. The cylindrical sliding bearing further have a coat layer of Bi or a Bi-based alloy formed on the outer back surface of the steel back metal. Preferably, the coat layer is made from the Bi-based alloy of 1 to 30 mass % of one or more kinds of Sn, Pb, In, Ag and Cu; and the balance being Bi and inevitable impurities.

The invention discloses a method for removing station captions and subtitles in an image based on a deep neural network, and relates to the technical field of image restoration, and the method comprises the following steps: S1, building an image restoration model; S2, preprocessing images of the training set; S3, processing training data: taking the training image as a real image Pt; Setting a pixel point RGB value in a Mask1 region in the training image as 0, and taking the pixel point RGB value as a training image P1; Setting a pixel point RGB value in a Mask2 region in the training image as0, and taking the pixel point RGB value as a training image P2; S4, training the image restoration model to obtain a trained image restoration model; S5, image restoration; The method comprises the following steps of: preprocessing an image or a video needing to remove station captions and subtitles; According to the image restoration method, based on the deep learning idea, station captions andsubtitles in the image are automatically and rapidly removed, the processing process is clear and clear, restoration real-time performance is high, and the application range is wide.

The invention provides a compressed multi-scale feature fusion network-based image super-resolution reconstruction method. The invention aims to solve a technical problem that a reconstructed high resolution image has a low peak signal to noise ratio and low structural similarity in the prior art. The implementation process of the invention includes the following steps that: a training sample setcomposed of high- and low-resolution image pairs is obtained; a multi-scale feature fusion network is constructed; the multi-scale feature fusion network is trained; a compressed multi-scale feature fusion network is obtained; and the compressed multi-scale feature fusion network is adopted to perform super-resolution reconstruction on an RGB image to be reconstructed. According to the compressedmulti-scale feature fusion network-based image super-resolution reconstruction method of the invention, a plurality of multi-scale feature fusion layers which are connected with one another sequentially in a stacked manner in the multi-scale feature fusion network are adopted to extract the multi-scale features of low-resolution images, and nonlinear mapping is performed on the multi-scale features of the low-resolution images; and therefore, the improvement of the low peak signal to noise ratio and low structural similarity of the reconstructed high-resolution image can be benefitted. The method can be applied to fields such as remote sensing imaging, public safety, medical diagnosis.

The invention provides a simulation generation method for wind speed data of multiple wind power farms. The method includes the steps that according to historical data of wind speeds of the multiple wind power farms, delay time among the wind speeds is determined; a model for the wind speeds of the multiple wind power farms is constructed through a Copula function theory, and the constructing step includes the substeps that marginal distribution of all the wind speeds is determined, the suitable Copula function model is selected, model parameters are estimated, and the model is evaluated; the wind speeds of the multiple wind power farms are generated according to the constructed Copula function model; the wind speeds generated in a simulation mode are reconstructed and adjusted in the aspect of the time sequence through a stochastic differential equation and a delay relation. The simulation generation method for the wind speed data of the multiple wind power farms can meet fluctuation characteristics of the respective wind speeds of the wind power farms and the cross correlation among the wind speeds of the multiple wind power farms.

The invention provides a CNN-based power equipment fault judgment and early warning method, a terminal and a readable storage medium. The method comprises the steps of obtaining test data; preprocessing the data; processing the data by using an offline model; and performing fault prediction of data. According to the method, the coal mill data is modeled through a deep learning method, fault prediction is achieved, mass historical data of coal mill equipment are fully mined through an existing data mining and machine learning modeling method, and an efficient and practical model is establishedto conduct detection and early warning on the real-time state of the coal mill. Knowledge and experience of experts and operating personnel are combined with data mining and machine learning methods and complemented with each other. The data can be automatically analyzed and modeled according to the data characteristics, and the threshold of operating personnel is lowered. The fault prediction model of the coal mill established by the invention can contain more complex causal relationships implicit among the indexes, so that the possibility of loss of a large amount of effective information isavoided, and the result is relatively reasonable and accurate.

The invention discloses a wind speed prediction method and prediction system. The prediction method comprises the following steps: acquiring an original wind speed sequence; performing empirical modedecomposition on the wind speed sequence, and obtaining a plurality of inherent modal functions and residual items; classifying all inherent modal functions according to the instantaneous frequency mean value of each inherent modal function, obtaining a plurality of high-frequency modal functions and a plurality of low-frequency modal functions; training the least square support vector machine byadopting the training sample data of each high-frequency modal function to obtain a high-frequency prediction model; training a BP neural network through the training sample data of each low-frequencymode function to obtain a low-frequency prediction model; training the BP neural network through the training sample data of the residual items to obtain a residual prediction model; predicting the wind speed by utilizing all the high-frequency prediction models, the low-frequency prediction models and the residual prediction models. A prediction model is built on the basis of fluctuation characteristics of different components, the random fluctuation of the wind speed sequence can be effectively weakened, and the wind speed can be accurately predicted.

A sports shoe has a sole S absorbing a shock of landing, an upper U covering an instep of a foot, and a fastening member 3 fitting the upper U to the instep of the foot. The upper U has a first opening 1 from which a leg extends upwardly during wearing and a second opening 2 which is closed with a tongue, and the first and second openings 1, 2 are continuous with each other in a front-rear direction, the upper comprises a medial side portion 7 covering a medial side surface of the foot, a lateral side portion 8 covering a lateral side surface of the foot, and stretchable portions 71, 81 forming a part of the medial side portion 7 and / or the lateral side portion 8, and the second opening 2 inclines toward the medial side of the foot along a ridge line of the foot as the second opening extends toward a front of the foot.

The invention discloses a pulmonary nodule automatic detection method and system based on a pulmonary CT sequence, belongs to the field of medical image segmentation, and aims to improve the network fitting capability and the candidate nodule detection rate. The method comprises steps of S1, data preprocessing; S2, screening candidate nodules by using a full convolutional network; S3, using an image processing method to change the probability graph into nodule center point coordinates and radiuses; And S4, using multi-model fused 3D convolutional network detection to obtain a finally determined nodule coordinate and a corresponding radius set, better utilizing the spatial information of the lung CT, and introducing a residual structure and a feature mapping structure to enhance the fittingcapability of the 3D full convolutional network.

The invention relates to the technical field of risk control of the Internet financial industry, in particular to a credit evaluation method for optimizing a generalized regression neural network based on a grey wolf algorithm. The method comprises six steps, and compared with common BP and RBF neural networks, the method has the advantages that GRNN selected by the method is strong in nonlinear mapping capability, good in approximation performance and suitable for processing unstable data. The method has the advantages of being good in generalization ability, high in fitting ability, high intraining speed, convenient in parameter adjustment and the like, and compared with common optimization algorithms such as genetic algorithms and particle swarms, the grey wolf algorithm is few in parameter and simple in programming, and has the advantages of being high in convergence speed, high in global optimization ability, potential in parallelism, easy to implement and the like. The grey wolfalgorithm is adopted to optimize the GRNN network model, the prediction precision and stability are high, the defects that the GRNN prediction result is unstable and is very likely to fall into the local minimum value are effectively avoided, and rapid and accurate online real-time prediction of the credit score of the application user is achieved.

The invention relates to a support vector machine inverse-based bearingless asynchronous motor control system. According to the system, two Park<-1> conversions, two Clark <-1> conversions, two current hysteresis PWM (Pulse Width Modulation) inverters and a bearingless asynchronous motor form a composite controlled object; and the support vector machine inverse of the composite controlled object is created by the support vector machines and integrators through least squares offline learning; the support vector machine inverse and the composite controlled object are connected in series to form a pseudo-linear system which is formed by two linear radial displacement subsystems, a speed subsystem and a magnetic linkage subsystem and is subjected to closed-loop compound control in a multi-internal-model switching manner. The control system provided by the invention can identify real-time equivalent models of the controlled object at improved dynamic response speed and steady state control accuracy, has high robustness against external disturbance and motor parameter variation, and can realize the high-performance dynamic decoupling control of all controlled variables of the bearingless asynchronous motor.

The embodiment of the invention provides a data processing method, device and equipment. The method is used for generating a polynomial capable of fitting an activation function, wherein the polynomial is used for replacing the activation function to construct a machine learning model. The method comprises the following steps: determining the number of times of a polynomial and a fitting intervalaccording to feature information of service data; wherein the fitting interval comprises a plurality of sub-intervals; each subinterval corresponds to a weight coefficient; constructing a metric function according to the subintervals and the weight coefficients; the metric function is used for representing the difference degree of the polynomial and the activation function on the fitting interval;calculating a value of an undetermined coefficient in the polynomial according to the metric function; and generating an expression of the polynomial according to the value of the undetermined coefficient.