TWI678661B - Palm print recognition apparatus and method having data extension - Google Patents

Abstract

The invention provides a palmprint identification device and method with data expandability. The device includes a plurality of capacitance sensing units, an identification unit and an energy storage unit. These capacitive sensing units sense the depth of the palm print and transmit them to the identification unit. The identification unit receives capacitance changes of a plurality of capacitance sensing units for training and identification functions, and amplifies the sample data amount through a data expansion algorithm. The energy storage unit provides device power. Thereby, an identification device with high accuracy and low cost can be provided.

Description

Palm print identification device and method with data expandability

The invention relates to a palmprint recognition device and method with non-contact capacitive real-time sensing and identification technology, which has particularly expandable data.

With the advancement of information and communication technology, identification devices have been developed with low cost, convenient installation and simple use.

Generally known identity identification devices usually use multi-factor identity identification, and choose one of the knowledge, possessions, and personal characteristics of the user as the basis for identification. Common identification methods include memory passwords, keys, and fingerprints. However, the shortcomings of these identification methods are that the application of multi-factor identification technology may cause the product to be expensive and more complicated to use than single-factor identification. Therefore, the development of identity recognition devices is bound to require more novel designs.

According to the Republic of China Invention Patent No. I562010 "Sight-track authentication system and method combined with hand authentication, computer-readable recording media and computer program products", It discloses a gaze trajectory authentication system, which includes an image capture device, a display device, and a control unit. This image capture device is used to capture user images. The display device provides a display interface. The control unit includes a hand feature recognition module, an eye movement analysis module, a sight track unlocking module, and an autonomous will confirmation module. This gaze track authentication system confirms the user's identity through hand characteristics, and detects the user's gaze direction and eye movements. On the one hand, the gaze track of the user is recorded according to the gaze direction of the user to confirm the identity of the user. On the other hand, the eye movements of the user are analyzed to confirm whether the user is operating under his own will by analyzing the eye movements of the user. The disadvantage of this patent is that the use of image capture devices can not reduce the cost of the identification device. Furthermore, this patent is unlocked by controlling the recognition of hand characteristics, eye movement analysis, gaze trajectory and voluntary will, which may cause the complexity of general user identification and use. improve.

According to the Republic of China Invention Patent No. I536272 "Biometric Device and Method", this patent discloses a biometric device and method using an infrared vertical cavity surface-emitting laser (VCSEL) as an alternative light source for a biometric device. It can provide clearer images in the recognition of venous images, and uses a unique optical light guide mechanism to achieve the effect of miniaturization. In addition, this biometric device and method also uses a single image sensor unit to simultaneously capture two images of finger veins and fingerprints in the same sensing signal to perform subsequent analysis and comparison of vein characteristics and fingerprint characteristics, making it harder Relatively lower body cost, simpler circuit design, and higher processing efficiency. The disadvantage of this patent is that the vertical cavity surface-emitting laser sensor using infrared rays Testing device, you need to put your finger in a specific recognition device, which limits the use. Furthermore, the image processing unit used in this biometric identification device is expensive, which may increase the cost of the identification device.

It can be seen that there are still many shortcomings in the above-mentioned customary methods. It is not a good design, and it needs to be improved.

In view of this, the present invention provides a palmprint identification device and method with data scalability, which uses a capacitive sensing method to sense palmprint information and uses a data scalability algorithm to reduce the cost of the device.

The palmprint recognition device with data expandability according to the embodiment of the present invention includes a plurality of capacitive sensing units and a recognition unit. Capacitive sensing units are used to collect the surrounding capacitance changes. The identification unit analyzes the collected capacitance changes to increase the amount of data to obtain an extended feature vector, and uses the extended feature vector to identify the identity through an identification algorithm. This identification algorithm uses an identification model trained by machine learning techniques.

The palmprint identification method with data extensibility according to an embodiment of the present invention includes the following steps. A plurality of capacitance changes are collected, and the data analysis of these capacitance changes is amplified to obtain an extended feature vector. This extended feature vector is used to identify the identity through the identification algorithm. The identification algorithm uses an identification model trained by machine learning techniques.

One of the objectives of the present invention is to perform biometric identification on the capacitive sensing side, Through data expansion, the number of nodes in the capacitive sensing unit of the hardware device is reduced. At the same time, the identification calculation is completed based on an expanded sample of existing data. Compared with the conventional technology using a high-precision sensing unit as a basis for biometric identification, the embodiments of the present invention can effectively reduce costs. Furthermore, because the palm print has a large depth, the accuracy of the capacitance sensing unit can effectively reduce and reduce the cost of the device.

One of the objectives of the present invention is that, because each user's palm print is unique and the orientation of the user's palm is different, the device is based on a multi-factor identity recognition method, so as to improve user intuition and improve the system Security purpose. Furthermore, the long-distance capacitance sensing method according to the embodiment of the present invention can detect the identification basis of the distance of the object, and solve the application of different recognition angles of the palm in different situations. In addition, the combination of data scalability algorithms can reduce the construction cost of the capacitive sensing unit, achieve the advantages of low cost, and increase the willingness to install. Cheap price and other advantages.

In order to make the above features and advantages of the present invention more comprehensible, embodiments are hereinafter described in detail with reference to the accompanying drawings.

1‧‧‧ Palmprint recognition device with data expandability

1a‧‧‧Capacitive sensing unit

1b‧‧‧Identification unit

1c‧‧‧energy storage unit

S1 ~ S7‧‧‧step

H‧‧‧ Palm

FIG. 1 is a block diagram of components of a palmprint recognition device with data expandability according to an embodiment of the present invention.

FIG. 2 is a schematic diagram of a palmprint recognition device with data expandability according to an embodiment of the present invention.

FIG. 3 is a flowchart of a palm print identification work with data extensibility according to an embodiment of the present invention.

In the general identity recognition device, three points based on multi-factor identity recognition are used as the basis for identification, which are knowledge they know, something they have, and something they are. If the identification system has more of the above conditions, the system security can be improved.

FIG. 1 is a block diagram of components of a palmprint recognition device with data expandability. Please refer to FIG. 1. The palmprint identification device 1 with data expandability includes a plurality of capacitive sensing units 1a, an identification unit 1b, and an energy storage unit 1c. The plurality of capacitance sensing units 1a measure a peripheral capacitance change amount, and transmit the result (that is, the capacitance change amount) to the identification unit 1b. The capacitance sensing unit 1a can transmit the amount of capacitance change through a wired or wireless transmission interface (for example, circuit wiring, Wi-Fi, Bluetooth, etc.). In addition, the capacitive sensing unit 1a can be arranged on a rigid circuit board or a flexible circuit board, and can form a node with metal properties. The identification unit 1b has at least a microprocessor, a chip capable of operating the operating system, or other computing circuits. The energy storage unit 1c is used to provide power to the palmprint identification device 1.

FIG. 2 is a schematic diagram of a palmprint recognition device with data expandability. Referring to FIG. 2, this device 1 includes a long-term board (or a circuit board) in which capacitor nodes (that is, the capacitive sensing units 1 a) are arranged in a 3 × 6 (18). Due to the large area of the palm print of the H, a large area capacitance value needs to be captured to produce a good identification effect. It should be noted that the area and shape of this panel can be adjusted according to actual needs. Capacitance sensing unit 1a can sense Measure an object within a specific distance (for example, 1 to 5 cm) to generate a change in capacitance and convert it into a voltage value to sense the palmprint condition of the palm in real time. The amount of capacitance change can also be used to detect the distance of the object to identify the basis, to solve the application of different recognition angles of the palm H in different occasions.

In order to facilitate the understanding of the operation flow of the embodiments of the present invention, a number of embodiments will be described in detail below to describe the process of identity recognition in the embodiments of the present invention. Hereinafter, the method described in the embodiment of the present invention will be described with each element and module in the palmprint recognition device 1. Each process of the method can be adjusted according to the implementation situation, and is not limited to this.

FIG. 3 shows a flowchart of a palmprint recognition method with data scalability. Referring to FIG. 3, the identification unit 1b first collects the feedback capacitance change amount (formed by the detected capacitance sensing value) returned by each capacitance sensing unit 1a (step S1). For example, the recognition unit 1b to the collected data samples (i.e., amount of capacitance variation) expressed in a matrix as: τ (t) = [C 1 C 2 ... C n] ... (1) where n is the capacitive sensing The number of measurement units 1a, C is the capacitance t sensed by the capacitance sensing unit 1a is the sampling time, and τ ( t ) is the capacitance matrix.

The identification unit 1b expands these capacitance change amounts (step S2). For example, the identification unit 1b uses an amplification matrix (composed of cos and sin) for the capacitance matrix τ ( t ) to amplify the amount of data.

其中τ ₀(t)為第t個取樣時間點內編號為0之電容感測單元1a所感測之電容值。τ ₁(t)為第t個取樣時間點內編號為1之電容感測單元1a所感測之電容值。由公式(2)得知，其電容感測樣本從原本兩個提昇為三個(即，擴展特徵向量[ρ ₁ ρ ₂ ρ ₃])。換言之，本發明實施例可節省y=(2/3)x個電容感測單元1a(x為擴展特徵向量中的元素個數)，從而降低裝置成本，進而提升使用者安裝意願。需說明的是，此範例以兩個電容變化量為例，而隨此數量改變，擴增矩陣內容亦隨之改變。

Τ ₀ ( t ) is the capacitance value sensed by the capacitance sensing unit 1 a with the number 0 in the t- th sampling time point. τ ₁ ( t ) is the capacitance value sensed by the capacitance sensing unit 1 a with the number 1 in the t- th sampling time point. It is known from formula (2) that the capacitance sensing sample is increased from two to three (that is, the extended eigenvector [ ρ ₁ ρ ₂ ρ ₃ ]). In other words, the embodiment of the present invention can save y = (2/3) x capacitive sensing units 1a (x is the number of elements in the extended feature vector), thereby reducing the cost of the device and further increasing the user's willingness to install. It should be noted that in this example, two capacitance changes are taken as an example, and the content of the amplification matrix changes as the quantity changes.

Next, the recognition unit 1b enters a recursive Chebyshev Neural Network (RCNN) recognition calculation (step S3). At this time, if in the training mode (step S4), the recognition unit 1b uses the recognition algorithm to perform weight training on the recognition model of the RCNN (step S5). For example, the identification unit 1b uses the calculation sample (ie, extended feature vector) obtained by the formula (2), and uses a recursive Chebyshev-type neural network as a training algorithm. The Chebyshev activation function can be expressed as U _n = 2 ρU _{n -1} ( ρ ) -U _{n -2} ( ρ ) ... (3) where U ₁ ( ρ ) = 1.

其中σ _j 為神經網路偏移值，j個RCNN節點，

為上個Φ _j 的狀態，λ _j 為遞回式權重。 Therefore, the activation function output equation can be obtained

Where σ _j is the neural network offset value, j RCNN nodes,

Is the last Φ _j state, and λ _j is the recursive weight.

其中μ _i 為輸出權重，ξ為推論結果。 Finally, the recognition unit 1b can obtain the output of the RCNN (ie, a recognition model based on machine learning technology) as:

Where μ _i is the output weight and ξ is the inference result.

其中e=x _d -ξ，x _d 為目標辨識結果。 It is worth noting that RCNN uses the energy function as the basis for training and convergence of samples, and its energy function can be defined as

Where e = x _d - ξ and x _d is the target recognition result.

Therefore, the identification unit 1b can obtain the updated weight equation as:

其中η ₁,η ₂,η ₃>0。

Where η ₁ , η ₂ , η ₃ > 0.

Finally, if the difference (e) between the inference (ξ) output from the identification model and the given target ( x _d ) is less than the expected value, the training mode can be ended. Different users can sequentially train different recognition models through the training mode of the palmprint recognition device 1. It should be noted that in this embodiment, the RCNN algorithm is used for training, but in other embodiments, algorithms of other machine learning techniques can also be applied.

其中x _d 為訓練目標。 If in the recognition mode (step S6), the recognition unit 1b inputs the extended feature vector after the capacitance change amount is expanded to the trained recognition model to obtain the identification result of the identity through the performance index through RCNN calculation ( That is, it is judged whether or not it conforms to the trained palm print) (step S7), thereby achieving the effect of identifying security. For example, when the identification unit 1b is set to the identification mode, it uses the performance index calculation to obtain the identification conclusion, as shown in formula (10):

Where x _d is the training target.

After the identification unit 1b completes the identification, the identification result can be sent to the background device in a wireless or wired manner to achieve different value-added service purposes. The palmprint recognition device 1 according to the embodiment of the present invention can be applied to any application environment that requires identity recognition, such as an access control system, device permissions, etc., and it can be adjusted according to the needs of the user.

The features and effects of the embodiments of the present invention are as follows: The previous traditional multi-factor identification technology is mainly aimed at the identity authentication technology, and proposes to use biometrics such as hand fingerprints and vein images as the identification basis. However, the general biometric method causes the sensing device to be expensive and the calculation is relatively complicated, and it is more likely to have limitations such as greatly reducing battery life and difficulty in installation. Furthermore, if the sensing accuracy is increased, the number of sensing units needs to be increased, and the cost of the device cannot be reduced. For the average user, if the price of the device drops and the identity recognition function can be made in real time, it is bound to make the identity verification security device more popular. In contrast, in order to achieve simple measurement and low price, it is more important to consider circuit functions and use scenarios. When compared with other conventional technologies, the palmprint recognition device and method provided by the embodiments of the present invention have the following advantages:

1. The embodiments of the present invention can reduce the number of sensing by the capacitive sensing unit.

2. The embodiments of the present invention can be measured without using a high-precision capacitive sensing unit. Palm print information.

3. The embodiment of the present invention uses a data diversity algorithm to achieve the identification effect.

4. The multi-factor identity recognition method according to the embodiment of the present invention improves the user system security purpose.

5. The embodiments of the present invention are not limited to the factors that change the local climate and the user's physiology, so that the discrimination performance is different.

Although the present invention has been disclosed as above with the examples, it is not intended to limit the present invention. Any person with ordinary knowledge in the technical field can make some modifications and retouching without departing from the spirit and scope of the present invention. The protection scope of the present invention shall be determined by the scope of the attached patent application.

Claims (10)

A palmprint identification device with data expandability includes: a plurality of capacitive sensing units to collect surrounding capacitance changes; and an identification unit that analyzes the collected capacitance changes to amplify the amount of data to obtain expansion The feature vector, and the extended feature vector is identified by an identification algorithm, wherein the identification algorithm uses an identification model trained by machine learning technology. The palmprint identification device with data expandability as described in item 1 of the scope of patent application, wherein the identification unit amplifies the amount of data through an expansion matrix for the collected capacitance changes, and if the capacitance changes Is 2, the expansion matrix is: The palmprint identification device with data expandability described in item 1 of the scope of patent application, wherein in the training mode, the identification unit trains the identification model through the machine learning technology based on the capacitance changes to update the identification model the weight of. As described in item 1 of the patent application scope, the palmprint recognition device with data extensibility, wherein the machine learning technology is related to a recursive Chebyshev Neural Network (RCNN). The palmprint recognition device with data extensibility described in item 1 of the patent application scope, wherein in the training mode, if the difference between the inference result output by the recognition model and a given target is less than the expected value, the recognition unit ends The training mode of the recognition model. A palmprint identification method with data expansibility includes: collecting a plurality of capacitance changes; expanding the data amount analysis of the capacitance changes to obtain an extended feature vector; and identifying the identity by using the extended feature vector through a recognition algorithm Where the identification algorithm utilizes an identification model trained by machine learning techniques. The palmprint identification method with data expansion as described in item 6 of the scope of patent application, wherein the step of amplifying the data amount of the capacitance changes to obtain the extended feature vector includes: collecting the capacitance change amounts The amount of data is amplified through an expansion matrix, and if the number of the capacitance changes is 2, the expansion matrix is: According to the method of palmprint identification with data expansion described in item 6 of the scope of patent application, before the step of amplifying the data amount of the capacitance changes to obtain the extended feature vector, the method further includes: in the training mode, Based on the capacitance changes, the identification model is trained through the machine learning technology to update the weight of the identification model. The palmprint identification method with data extensibility described in item 6 of the patent application scope, wherein the machine learning technology is related to a recursive Chebyshev-like neural network (RCNN). According to the palmprint identification method with data extensibility described in item 6 of the scope of the patent application, before the step of analyzing the capacitance changes by amplifying the data amount to obtain the extended feature vector, the method further includes: If the difference between the inference result output by the recognition model and the given target is less than the expected value, the training mode of the recognition model is ended.