CN107491729A - The Handwritten Digit Recognition method of convolutional neural networks based on cosine similarity activation - Google Patents

The Handwritten Digit Recognition method of convolutional neural networks based on cosine similarity activation Download PDF

Info

Publication number
CN107491729A
CN107491729A CN201710566272.1A CN201710566272A CN107491729A CN 107491729 A CN107491729 A CN 107491729A CN 201710566272 A CN201710566272 A CN 201710566272A CN 107491729 A CN107491729 A CN 107491729A
Authority
CN
China
Prior art keywords
mrow
mtd
msubsup
mtr
munderover
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.)
Granted
Application number
CN201710566272.1A
Other languages
Chinese (zh)
Other versions
CN107491729B (en
Inventor
刘昱
穆翀
刘明
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tianjin University
Original Assignee
Tianjin University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Tianjin University filed Critical Tianjin University
Priority to CN201710566272.1A priority Critical patent/CN107491729B/en
Publication of CN107491729A publication Critical patent/CN107491729A/en
Application granted granted Critical
Publication of CN107491729B publication Critical patent/CN107491729B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
    • G06V30/10—Character recognition
    • G06V30/32—Digital ink
    • G06V30/333—Preprocessing; Feature extraction
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/30—Noise filtering
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/32—Normalisation of the pattern dimensions
    • 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
    • G06V10/56—Extraction of image or video features relating to colour
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
    • G06V30/10—Character recognition
    • G06V30/32—Digital ink
    • G06V30/36—Matching; Classification
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
    • G06V30/10—Character recognition

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Data Mining & Analysis (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Biomedical Technology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Health & Medical Sciences (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Evolutionary Biology (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses a kind of Handwritten Digit Recognition method of the convolutional neural networks based on cosine similarity activation, Step 1: being pre-processed first to target handwritten numeral image, normalized including at least denoising, enhancing, image correcting inclination and picture size；Step 2: using transition matrix, pretreated target handwritten numeral image is transformed into the orthogonal YC of color componentrCbColor space；Step 3: cosine similarity is incorporated into the activation primitive of convolutional neural networks convolutional layer and completes network training, then target image is input in the convolutional neural networks model trained after being changed in step 2, carries out digital sort differentiation.Compared with prior art, the present invention is applied to Handwritten Digit Recognition of the illumination condition with the condition of writing in the case of bad, has the characteristics of real-time is good, and recognition accuracy is high.

Description

The Handwritten Digit Recognition method of convolutional neural networks based on cosine similarity activation
Technical field
The present invention relates to the multiple technologies such as pattern-recognition, artificial intelligence, machine learning field, more particularly to one kind to adopt The Handwritten Digit Recognition method of extracting tool is characterized with the convolutional neural networks that cosine similarity is activation primitive.
Background technology
In the information age, artificial intelligence is the multi-field cross discipline gradually risen at nearly more than 20 years, is related to general The multi-door subjects such as rate opinion, statistics, Approximation Theory, convextiry analysis, algorithm complex theory.Deep learning is artificial intelligence field one New research direction, in recent years based on neutral net be based especially on deep learning neutral net speech recognition, machine vision, The progress of making a breakthrough property in the application of the multiclass such as commending system.Its motivation is the nerve connection for establishing modeling human brain Structure, when handling image, sound and these signals of text, data characteristics is described by the layering of multiple conversion stages, And then provide the explanation of data.The essence of deep learning is to form more abstract high-rise expression attribute by combining low-level feature Classification or feature, to find that the distributed nature of data represents.It is the core of artificial intelligence, computer is had intelligence Fundamental way, it is applied using the every field throughout artificial intelligence in the data processing and problem analysis of other field Extensively.
With the continuous development of information technology, Arabic numerals turn into one of important information interchange instrument, realize and calculate Machine accurately identifies the key point for being to speed up social informatization process to handwritten numeral.The data that people often use all are for we Uncle's numeral, so the Study of recognition of handwriting digital is always leading using Arabic numerals.Handwritten Digit Recognition is optical character One important branch of identification technology, it has a wide range of applications in postcode, financial statement, bank money etc., and And the always study hotspot of image procossing, area of pattern recognition.With the development of society, various countries' economic interaction is increasingly deepened, People will handle substantial amounts of bill daily, so handwritten numeral is essential in this field, such as people will be handled perhaps The data such as more checks, invoice, goods list, these largely will come into contacts with numeral.The digital morphological write by different people It is different, vary, or even write sometimes it is lack of standardization, even same person, due to extraneous and oneself factor influence, Also can make it is hand-written go out numeral there is very big otherness, so accurately identifying for handwritten numeral is extremely complex and difficult.
For popular, everyone hand-written Arabic numerals mode is the digital shape that different, different people is write State is different, varies, while the digital specification sex differernce that the difference for writing environment causes to write out greatly increases, and uses computer It is highly difficult the great handwritten numeral of this otherness to be identified classification, and recognition accuracy is not high, but with artificial The development of intelligence, be based particularly on that the convolutional neural networks of deep learning obtain in computer vision application it is breakthrough enter Exhibition, convolutional neural networks form more abstract high-rise expression attribute classification or feature by combining low-level feature, to find number According to distributed nature represent.Make it possible accurately and rapidly to identify handwritten numeral, and to different illumination conditions and difference The very big Handwritten Digit Recognition of property has very strong robustness.
The content of the invention
Based on above-mentioned technical problem, the present invention proposes a kind of hand of the convolutional neural networks based on cosine similarity activation Write digit recognition method, it is therefore an objective to improve recognition accuracy, and in the case of this method is bad to illumination condition and writing condition Handwritten Digit Recognition there is good robustness.
A kind of Handwritten Digit Recognition method of convolutional neural networks based on cosine similarity activation proposed by the present invention, should Method comprises the following steps：
Step 1: being pre-processed first to target handwritten numeral image, tilt and entangle including at least denoising, enhancing, image Just normalized with picture size；
Step 2: using transition matrix, pretreated target handwritten numeral image is transformed into the mutual not phase of color component The YC of passrCbColor space；Conversion formula is as follows：
Step 3:, cosine similarity is incorporated into the activation primitive of convolutional neural networks convolutional layer and complete network instruction Practice, then will in step 2 change after target image be input in the convolutional neural networks model trained, carry out digital sort sentence Not；
Wherein：Convolutional layer is as similarity function, its formula using cosine similarity in convolutional neural networks：
X=[x1,x2,....,xn]T
Wherein, CosSim (Wi (j), X) and it is vectorial Wi (j)With X cosine similarity；X is input layer input signal；Wi (j)Table Show the weights pattern of i-th of node of jth layer,Represent s-th of the input of jth layer and the connection weight of i-th of node.
Convolutional layer activation primitive is：
Wherein, T is Node B threshold；
The handwritten numeral image is Arabic numerals image, and image is unit numbers, and numeral includes 0 to 9 totally ten Numeral, discriminant classification result are one kind in 0 to 90 class numerals.
Compared with prior art, the present invention has following good effect：
The present invention is applied to Handwritten Digit Recognition of the illumination condition with the condition of writing in the case of bad, good with real-time, The characteristics of recognition accuracy is high.
Brief description of the drawings
Fig. 1 is the Handwritten Digit Recognition method flow of the convolutional neural networks based on cosine similarity activation of the present invention Figure.
Embodiment
Embodiments of the present invention are described in further detail below in conjunction with accompanying drawing.
The Handwritten Digit Recognition method of the convolutional neural networks based on cosine similarity activation of the present invention, this method are having When body is implemented, two kinds of flows are broadly divided into：First, the flow handled the target handwritten numeral image of input；Second, foundation The flow that the handwritten numeral image data collection pre-set is handled, they comprise the following steps respectively：
The present invention can be that the target handwritten numeral image of input is handled, and flow specifically includes：
Step 101, input target handwritten numeral image；
Step 102, the target handwritten numeral image to input carry out denoising, enhancing, image correcting inclination and picture size The image preprocessings such as normalization；
Step 103, using transition matrix target image is transformed into the orthogonal YC of color componentrCbColor space；
Conversion formula is as follows：
Step 104, target image is input to the convolutional neural networks trained, i.e., cosine similarity is incorporated into convolution In the activation primitive of neutral net convolutional layer, and complete network training；Convolutional layer using cosine similarity as similarity function, Its formula is：
X=[x1,x2,....,xn]T
Wherein：CosSim(Wi (j), X) and it is vectorial Wi (j)With X cosine similarity；X is input layer input signal；Wi (j)Table Show the weights pattern of i-th of node of jth layer,Represent s-th of the input of jth layer and the connection weight of i-th of node；
Convolutional layer activation primitive formula is：
Wherein, T is Node B threshold；
Target image is input in the convolutional neural networks model trained after being changed in step 103, carries out numeral point Class differentiates)；Convolutional neural networks are disaggregated models end to end, in the full articulamentum of last layer will before convolutional layer learn To result be mapped in 0 to 9 this 10 numerals, respectively correspond to 0 to 9 this 10 class handwritten numeral picture input network obtained by Classification results；
Step 105, output differentiate result；Wherein：Heretofore described handwritten numeral image is Arabic numerals image, and Image is unit numbers, and numeral includes 0 to 9, and totally ten numerals, discriminant classification result are one kind in 0 to 9 totally ten class numerals.
The present invention can also be handled according to the handwritten numeral image data collection pre-set, and flow includes following step Suddenly：
Step 201, the handwritten numeral image data collection marked is chosen from database；
Step 202, carry out data images pretreatment；
Step 203, triple channel decorrelation processing, data images conversion are carried out using the conversion formula in step 103 To YCrCbColor space；
Step 204, determine whether to train picturePurpose is that data set is divided into training sample set and test sample collection；, Determine whether to train picture approach using cross validation appraisal procedure；
If not training picture, step 205 is performed, then as test sample；
Step 206, the convolutional neural networks trained with test sample test；
Step 207, tested using test data set pair convolutional neural networks, judge to estimate whether accuracy rate is more than In threshold value；
Step 207, when accuracy rate is more than or equal to threshold value, then network performance is good, goes to step 104 re -training net Network；The method that accuracy rate judges is as follows：
Atrain≥Atrain_th
Wherein：Atrain_thIt is estimation accuracy rate threshold value, rule of thumb sets；
If training picture, performs step 208, as training sample；
Step 209, target image is input to the convolutional neural networks model trained after being changed in step 203 In；
Step 210, the convolutional neural networks trained；Step 206 is gone to, and then performs subsequent step 207.
Digital sort differentiation is carried out, output differentiates result.
The invention is not limited in foregoing specific steps.The present invention expands in any this specification the new feature disclosed Or any new combination, or the combination of new step.To sum up, this specification content should not be construed as limiting the invention.

Claims (1)

  1. A kind of 1. Handwritten Digit Recognition method of the convolutional neural networks based on cosine similarity activation, it is characterised in that the party Method comprises the following steps：
    Step 1: pre-processed first to target handwritten numeral image, including at least denoising, enhancing, image correcting inclination and Picture size normalizes；
    Step 2: using transition matrix, it is orthogonal that pretreated target handwritten numeral image is transformed into color component YCrCbColor space；Conversion formula is as follows：
    <mrow> <mfenced open = "[" close = "]"> <mtable> <mtr> <mtd> <mi>Y</mi> </mtd> </mtr> <mtr> <mtd> <msub> <mi>C</mi> <mi>r</mi> </msub> </mtd> </mtr> <mtr> <mtd> <mrow> <mi>C</mi> <mi>b</mi> </mrow> </mtd> </mtr> </mtable> </mfenced> <mo>=</mo> <mfenced open = "[" close = "]"> <mtable> <mtr> <mtd> <mn>16</mn> </mtd> </mtr> <mtr> <mtd> <mn>128</mn> </mtd> </mtr> <mtr> <mtd> <mn>128</mn> </mtd> </mtr> </mtable> </mfenced> <mo>+</mo> <mfrac> <mn>1</mn> <mn>255</mn> </mfrac> <mfenced open = "[" close = "]"> <mtable> <mtr> <mtd> <mn>65.481</mn> </mtd> <mtd> <mn>128.553</mn> </mtd> <mtd> <mn>24966</mn> </mtd> </mtr> <mtr> <mtd> <mrow> <mo>-</mo> <mn>37.797</mn> </mrow> </mtd> <mtd> <mrow> <mo>-</mo> <mn>74.203</mn> </mrow> </mtd> <mtd> <mn>112</mn> </mtd> </mtr> <mtr> <mtd> <mn>112</mn> </mtd> <mtd> <mrow> <mo>-</mo> <mn>93.786</mn> </mrow> </mtd> <mtd> <mrow> <mo>-</mo> <mn>18214</mn> </mrow> </mtd> </mtr> </mtable> </mfenced> <mo>&amp;times;</mo> <mfenced open = "[" close = "]"> <mtable> <mtr> <mtd> <mi>R</mi> </mtd> </mtr> <mtr> <mtd> <mi>G</mi> </mtd> </mtr> <mtr> <mtd> <mi>B</mi> </mtd> </mtr> </mtable> </mfenced> </mrow>
    Step 3:, cosine similarity is incorporated into the activation primitive of convolutional neural networks convolutional layer and completes network training, then Target image is input in the convolutional neural networks model trained after being changed in step 2, carries out digital sort differentiation；
    Wherein：Convolutional layer is as similarity function, its formula using cosine similarity in convolutional neural networks：
    <mrow> <mi>C</mi> <mi>o</mi> <mi>s</mi> <mi>S</mi> <mi>i</mi> <mi>m</mi> <mrow> <mo>(</mo> <mrow> <msubsup> <mi>W</mi> <mi>i</mi> <mrow> <mo>(</mo> <mi>j</mi> <mo>)</mo> </mrow> </msubsup> <mo>,</mo> <mi>X</mi> </mrow> <mo>)</mo> </mrow> <mo>=</mo> <mfrac> <mrow> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>n</mi> </munderover> <msubsup> <mi>w</mi> <mi>i</mi> <mrow> <mo>(</mo> <mi>j</mi> <mo>)</mo> </mrow> </msubsup> <msub> <mi>x</mi> <mi>i</mi> </msub> </mrow> <msqrt> <mrow> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>n</mi> </munderover> <msup> <mrow> <mo>(</mo> <msubsup> <mi>w</mi> <mi>i</mi> <mrow> <mo>(</mo> <mi>j</mi> <mo>)</mo> </mrow> </msubsup> <mo>)</mo> </mrow> <mn>2</mn> </msup> <msqrt> <mrow> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>n</mi> </munderover> <msubsup> <mi>x</mi> <mi>i</mi> <mn>2</mn> </msubsup> </mrow> </msqrt> </mrow> </msqrt> </mfrac> <mo>=</mo> <mfrac> <mrow> <mo>&lt;</mo> <msubsup> <mi>W</mi> <mi>i</mi> <mrow> <mo>(</mo> <mi>j</mi> <mo>)</mo> </mrow> </msubsup> <mo>,</mo> <mi>X</mi> <mo>&gt;</mo> </mrow> <mrow> <mo>|</mo> <mo>|</mo> <msubsup> <mi>W</mi> <mi>i</mi> <mrow> <mo>(</mo> <mi>j</mi> <mo>)</mo> </mrow> </msubsup> <mo>|</mo> <mo>|</mo> <mo>|</mo> <mo>|</mo> <mi>X</mi> <mo>|</mo> <mo>|</mo> </mrow> </mfrac> </mrow>
    X=[x1,x2,....,xn]T
    <mrow> <msubsup> <mi>W</mi> <mi>i</mi> <mrow> <mo>(</mo> <mi>j</mi> <mo>)</mo> </mrow> </msubsup> <mo>=</mo> <msup> <mrow> <mo>&amp;lsqb;</mo> <msubsup> <mi>w</mi> <msub> <mi>i</mi> <mn>1</mn> </msub> <mrow> <mo>(</mo> <mi>j</mi> <mo>)</mo> </mrow> </msubsup> <mo>,</mo> <msubsup> <mi>w</mi> <msub> <mi>i</mi> <mn>2</mn> </msub> <mrow> <mo>(</mo> <mi>j</mi> <mo>)</mo> </mrow> </msubsup> <mo>,</mo> <mn>....</mn> <mo>,</mo> <msubsup> <mi>w</mi> <msub> <mi>i</mi> <mi>n</mi> </msub> <mrow> <mo>(</mo> <mi>j</mi> <mo>)</mo> </mrow> </msubsup> <mo>&amp;rsqb;</mo> </mrow> <mi>T</mi> </msup> </mrow>
    Wherein, CosSim (Wi (j), X) and it is vectorial Wi (j)With X cosine similarity；X is input layer input signal；Wi (j)Represent jth The weights pattern of i-th of node of layer,Represent s-th of the input of jth layer and the connection weight of i-th of node.
    Convolutional layer activation primitive is：
    <mfenced open = "" close = ""> <mtable> <mtr> <mtd> <mrow> <mi>y</mi> <mo>=</mo> <mi>s</mi> <mi>i</mi> <mi>g</mi> <mi>n</mi> <mrow> <mo>&amp;lsqb;</mo> <mrow> <mi>C</mi> <mi>o</mi> <mi>s</mi> <mi>S</mi> <mi>i</mi> <mi>m</mi> <mrow> <mo>(</mo> <mrow> <msubsup> <mi>W</mi> <mi>i</mi> <mrow> <mo>(</mo> <mi>j</mi> <mo>)</mo> </mrow> </msubsup> <mo>,</mo> <mi>X</mi> </mrow> <mo>)</mo> </mrow> <mo>-</mo> <mi>T</mi> </mrow> <mo>&amp;rsqb;</mo> </mrow> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <mo>=</mo> <mi>s</mi> <mi>i</mi> <mi>g</mi> <mi>n</mi> <mrow> <mo>&amp;lsqb;</mo> <mrow> <mfrac> <mrow> <mo>&lt;</mo> <msubsup> <mi>W</mi> <mi>i</mi> <mrow> <mo>(</mo> <mi>j</mi> <mo>)</mo> </mrow> </msubsup> <mo>,</mo> <mi>X</mi> <mo>&gt;</mo> </mrow> <mrow> <mo>|</mo> <mo>|</mo> <msubsup> <mi>W</mi> <mi>i</mi> <mrow> <mo>(</mo> <mi>j</mi> <mo>)</mo> </mrow> </msubsup> <mo>|</mo> <mo>|</mo> <mo>|</mo> <mo>|</mo> <mi>X</mi> <mo>|</mo> <mo>|</mo> </mrow> </mfrac> <mo>-</mo> <mi>T</mi> </mrow> <mo>&amp;rsqb;</mo> </mrow> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <mo>=</mo> <mi>s</mi> <mi>i</mi> <mi>g</mi> <mi>n</mi> <mrow> <mo>&amp;lsqb;</mo> <mrow> <mfrac> <mrow> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>n</mi> </munderover> <msubsup> <mi>w</mi> <mi>i</mi> <mrow> <mo>(</mo> <mi>j</mi> <mo>)</mo> </mrow> </msubsup> <msub> <mi>x</mi> <mi>i</mi> </msub> </mrow> <msqrt> <mrow> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>n</mi> </munderover> <msup> <mrow> <mo>(</mo> <msubsup> <mi>w</mi> <mi>i</mi> <mrow> <mo>(</mo> <mi>j</mi> <mo>)</mo> </mrow> </msubsup> <mo>)</mo> </mrow> <mn>2</mn> </msup> <msqrt> <mrow> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>n</mi> </munderover> <msubsup> <mi>x</mi> <mi>i</mi> <mn>2</mn> </msubsup> </mrow> </msqrt> </mrow> </msqrt> </mfrac> <mo>-</mo> <mi>T</mi> </mrow> <mo>&amp;rsqb;</mo> </mrow> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <mo>=</mo> <mfenced open = "{" close = ""> <mtable> <mtr> <mtd> <mrow> <mn>1</mn> <mo>,</mo> <mi>C</mi> <mi>o</mi> <mi>s</mi> <mi>S</mi> <mi>i</mi> <mi>m</mi> <mrow> <mo>(</mo> <mrow> <msubsup> <mi>W</mi> <mi>i</mi> <mrow> <mo>(</mo> <mi>j</mi> <mo>)</mo> </mrow> </msubsup> <mo>,</mo> <mi>X</mi> </mrow> <mo>)</mo> </mrow> <mo>-</mo> <mi>T</mi> <mo>&amp;GreaterEqual;</mo> <mn>0</mn> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <mn>0</mn> <mo>,</mo> <mi>C</mi> <mi>o</mi> <mi>s</mi> <mi>S</mi> <mi>i</mi> <mi>m</mi> <mrow> <mo>(</mo> <mrow> <msubsup> <mi>W</mi> <mi>i</mi> <mrow> <mo>(</mo> <mi>j</mi> <mo>)</mo> </mrow> </msubsup> <mo>,</mo> <mi>X</mi> </mrow> <mo>)</mo> </mrow> <mo>-</mo> <mi>T</mi> <mo>&lt;</mo> <mn>0</mn> </mrow> </mtd> </mtr> </mtable> </mfenced> </mrow> </mtd> </mtr> </mtable> </mfenced>
    Wherein, T is Node B threshold；
    The handwritten numeral image is Arabic numerals image, and image is unit numbers, and numeral includes 0 to 9 totally ten numerals, Discriminant classification result is one kind in 0 to 90 class numerals.
CN201710566272.1A 2017-07-12 2017-07-12 Handwritten digit recognition method based on cosine similarity activated convolutional neural network Expired - Fee Related CN107491729B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710566272.1A CN107491729B (en) 2017-07-12 2017-07-12 Handwritten digit recognition method based on cosine similarity activated convolutional neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710566272.1A CN107491729B (en) 2017-07-12 2017-07-12 Handwritten digit recognition method based on cosine similarity activated convolutional neural network

Publications (2)

Publication Number Publication Date
CN107491729A true CN107491729A (en) 2017-12-19
CN107491729B CN107491729B (en) 2020-07-28

Family

ID=60643762

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710566272.1A Expired - Fee Related CN107491729B (en) 2017-07-12 2017-07-12 Handwritten digit recognition method based on cosine similarity activated convolutional neural network

Country Status (1)

Country Link
CN (1) CN107491729B (en)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108154136A (en) * 2018-01-15 2018-06-12 众安信息技术服务有限公司 For identifying the method, apparatus of writing and computer-readable medium
CN109670433A (en) * 2018-12-13 2019-04-23 南京工程学院 A kind of Handwritten Digit Recognition method based on convolution Yu included angle cosine Furthest Neighbor
CN110210410A (en) * 2019-06-04 2019-09-06 南京邮电大学 A kind of Handwritten Digit Recognition method based on characteristics of image
CN111563563A (en) * 2020-07-16 2020-08-21 南京华苏科技有限公司 Method for enhancing combined data of handwriting recognition
CN111652108A (en) * 2020-05-28 2020-09-11 中国人民解放军32802部队 Anti-interference signal identification method and device, computer equipment and storage medium
CN111967424A (en) * 2020-08-27 2020-11-20 西南大学 Buckwheat disease identification method based on convolutional neural network

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105320961A (en) * 2015-10-16 2016-02-10 重庆邮电大学 Handwriting numeral recognition method based on convolutional neural network and support vector machine
CN106599941A (en) * 2016-12-12 2017-04-26 西安电子科技大学 Method for identifying handwritten numbers based on convolutional neural network and support vector machine

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105320961A (en) * 2015-10-16 2016-02-10 重庆邮电大学 Handwriting numeral recognition method based on convolutional neural network and support vector machine
CN106599941A (en) * 2016-12-12 2017-04-26 西安电子科技大学 Method for identifying handwritten numbers based on convolutional neural network and support vector machine

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108154136A (en) * 2018-01-15 2018-06-12 众安信息技术服务有限公司 For identifying the method, apparatus of writing and computer-readable medium
CN108154136B (en) * 2018-01-15 2022-04-05 众安信息技术服务有限公司 Method, apparatus and computer readable medium for recognizing handwriting
CN109670433A (en) * 2018-12-13 2019-04-23 南京工程学院 A kind of Handwritten Digit Recognition method based on convolution Yu included angle cosine Furthest Neighbor
CN110210410A (en) * 2019-06-04 2019-09-06 南京邮电大学 A kind of Handwritten Digit Recognition method based on characteristics of image
CN110210410B (en) * 2019-06-04 2022-09-23 南京邮电大学 Handwritten number recognition method based on image characteristics
CN111652108A (en) * 2020-05-28 2020-09-11 中国人民解放军32802部队 Anti-interference signal identification method and device, computer equipment and storage medium
CN111563563A (en) * 2020-07-16 2020-08-21 南京华苏科技有限公司 Method for enhancing combined data of handwriting recognition
CN111563563B (en) * 2020-07-16 2020-11-03 南京华苏科技有限公司 Method for enhancing combined data of handwriting recognition
CN111967424A (en) * 2020-08-27 2020-11-20 西南大学 Buckwheat disease identification method based on convolutional neural network

Also Published As

Publication number Publication date
CN107491729B (en) 2020-07-28

Similar Documents

Publication Publication Date Title
CN109492101B (en) Text classification method, system and medium based on label information and text characteristics
CN107491729A (en) The Handwritten Digit Recognition method of convolutional neural networks based on cosine similarity activation
CN109460473B (en) Electronic medical record multi-label classification method based on symptom extraction and feature representation
Naz et al. Offline cursive Urdu-Nastaliq script recognition using multidimensional recurrent neural networks
Dekhtyar et al. Re data challenge: Requirements identification with word2vec and tensorflow
CN106557462A (en) Name entity recognition method and system
CN108804677A (en) In conjunction with the deep learning question classification method and system of multi-layer attention mechanism
CN106909946A (en) A kind of picking system of multi-modal fusion
CN112801146A (en) Target detection method and system
CN106529525A (en) Chinese and Japanese handwritten character recognition method
CN105205449A (en) Sign language recognition method based on deep learning
CN104156690B (en) A kind of gesture identification method based on image space pyramid feature bag
Li et al. Nuclear norm regularized convolutional Max Pos@ Top machine
CN110046264A (en) A kind of automatic classification method towards mobile phone document
Angona et al. Automated Bangla sign language translation system for alphabets by means of MobileNet
CN109255339B (en) Classification method based on self-adaptive deep forest human gait energy map
Das et al. Determining attention mechanism for visual sentiment analysis of an image using svm classifier in deep learning based architecture
CN111914553A (en) Financial information negative subject judgment method based on machine learning
CN116010874A (en) Emotion recognition method based on deep learning multi-mode deep scale emotion feature fusion
CN113987188B (en) Short text classification method and device and electronic equipment
CN107818173A (en) A kind of false comment filter method of Chinese based on vector space model
CN113420833A (en) Visual question-answering method and device based on question semantic mapping
Bintoro et al. Lampung script recognition using convolutional neural network
Fayyaz et al. CNN and traditional classifiers performance for sign language recognition
CN111242131B (en) Method, storage medium and device for identifying images in intelligent paper reading

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
GR01 Patent grant
GR01 Patent grant
CF01 Termination of patent right due to non-payment of annual fee
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20200728

Termination date: 20210712