Non-contact handwriting device based on ultraviolet specific image sensor and preparation method

Abstract

The invention discloses a non-contact handwriting device based on an ultraviolet specific image sensor and a preparation method, and relates to the technical field of image sensing and human-computer interaction. Comprising a glass substrate 1, a transparent thin film transistor array 2 formed on the substrate, a photosensitive material layer 3 formed on the thin film transistor array, and an ultraviolet curing adhesive packaging layer 4 covering the photosensitive material layer. A light-sensitive material is a Bi2S3 material with a wavelength selective light current memory characteristic, generates long-time light current maintenance based on a photo-thermal electric effect in an ultraviolet band, rapidly attenuates in visible to near-infrared bands, and has ultraviolet input selectivity and ambient light anti-interference capability. According to the system, an ultraviolet laser pen is used as an input tool, handwriting patterns which can be kept for a long time are formed on an image sensor array, and image erasing based on a background normalization mechanism is supported. The image sensor array is composed of synaptic-like devices, has a long-term optical memory characteristic, and realizes integration of perception and storage.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of image sensing and human-computer interaction, in particular to a non-contact handwriting device based on an ultraviolet specific image sensor and a preparation method. BACKGROUND

[0002] With the development of human-computer interaction technology, non-contact gesture recognition and handwriting input gradually become an important input method in the field of intelligent terminals, intelligent wearable devices, public information terminals and the like. The existing non-contact handwriting input system is mostly dependent on a high frame rate camera based on visible light or near-infrared imaging technology, and realizes hand trajectory recognition through image processing algorithm. However, such system usually has the problems of high system complexity, serious environmental light interference, great dependence on computing resources and the like.

[0003] In commercial fields, non-contact interaction interfaces such as Microsoft Kinect and Leap Motion controller commonly adopt near-infrared camera technology. The time-of-flight ranging technology adopted by Microsoft Kinect needs to be calibrated to ensure positioning accuracy, and its performance is directly affected by light source power and target object reflectivity (see Liu, H, Wang, L, Gesture recognition for human-robot collaboration: a review, Int. J. Ind. Ergon, 2018, 68, 355–367.). Leap Motion controller as a gesture interaction interface can capture gesture actions with sub-millimeter accuracy, but the interaction space is limited to the inverted pyramid area above the device (see Aditya, K, et al, Recent trends in HCI: a survey on data glove, LEAP motion and Microsoft Kinect, In 2018 IEEE International Conference on System, Computation, Automation and Networking, (ICSCA) 1–5 (IEEE, 2018).). Another innovative technology, ShadowSense, realizes touch control by detecting the shadow generated when an object enters, hovers or leaves the sensor field of view. This system is based on the cooperative work of multiple light sources and multiple sensors, and by calculating the proportional change of light intensity between the “full lighting state” and the “shadow state” of the object in the touch plane, the accurate position of the object is determined in real time. However, this shadow sensing technology has technical limitations such as greater environmental light interference and difficulty in sensing depth information. In recent years, transparent near-infrared photodetector arrays have been widely used to construct non-contact handwriting or gesture interaction interfaces, with advantages such as visual transparency, compact structure, and strong integrability. Representative schemes include a non-contact interaction interface based on a large-area solution-processed near-infrared sensitive organic photodetector array, and user input recognition is achieved through near-infrared laser reflection signals (see Takeshi Kamijo, et al, A touchless user interface based on a near-infrared-sensitive transparent optical imager, Nat. Electron, 2023, 6, 451-461.). This interface can be placed directly in front of a conventional display screen, thereby avoiding the viewing angle limitation problem in traditional visual systems.However, it relies on near-infrared reflection signals for input recognition, is easily disturbed by infrared background light (such as sunlight, heat sources) in the environment, leading to recognition errors or false touch phenomena, limiting its stable operation under complex lighting conditions. In addition, the existing interactive system based on near-infrared organic photodetector array only has instantaneous light response capability and does not have a synaptic-like long-term memory function that can sustain pattern information, so the architecture still relies on the separation design of image sensing and data storage, making it difficult to realize integrated handwriting image acquisition, resulting in a large amount of redundant data storage and frequent image transmission during system operation, not only increasing storage bandwidth and computational burden, but also increasing overall power consumption and structural complexity. In terms of materials, organic photodetector materials are sensitive to environmental factors such as humidity, oxygen, and ultraviolet light, and the device may experience performance degradation under long-term working or exposure conditions, affecting system lifespan and reliability.

[0004] With the development of new optoelectronic functional materials, new image sensors with spectral selective response and optoelectronic memory effect have gradually become an important research direction for non-contact input systems. In particular, the ultraviolet band has very low background interference in natural environments, and if the image sensor has specific response capability to ultraviolet light, it can achieve high signal-to-noise ratio and strong anti-interference input recognition function. In addition, if the device can maintain a long-term photocurrent output after the ultraviolet excitation is removed, i.e., it has a memory-like characteristic, it does not need to rely on high-frame-rate image acquisition and high-speed data processing to achieve the preservation and reading of handwritten trajectories. This "optical writing - electrical retention" mechanism significantly simplifies the input system structure, improves input efficiency, and is conducive to the construction of low-power, low-latency non-contact human-computer interaction interfaces.

[0005] However, there is currently a lack of an image sensor array structure that combines ultraviolet response specificity, memory retention capability, low-temperature patterning compatibility, and high visible light transmittance for constructing high-performance, integrable, low-power non-contact handwriting input devices. SUMMARY

[0006] Therefore, the present application provides a non-contact handwriting device based on an ultraviolet-specific image sensor and a preparation method, which can use a photosensitive material with selective photocurrent retention capability in the ultraviolet band, can achieve a retention time of pattern information on the device up to hundreds of seconds, and has an image sensor array structure that combines ultraviolet response specificity, memory retention capability, low-temperature patterning compatibility, and high visible light transmittance.

[0007] To achieve the above-mentioned purpose, one aspect of an embodiment of the present application provides a non-contact handwriting device based on an ultraviolet-specific image sensor, comprising: a substrate, a thin film transistor array formed on the substrate, a photosensitive material formed on the thin film transistor array, and a packaging layer covering the photosensitive material.

[0008] The ultraviolet laser pen is used for handwriting input operation on the surface of the encapsulation layer, and the thin film transistor array is loaded with a rear-end array reading system.

[0009] Preferably, the substrate is a transparent glass substrate, and the transparent ITO is an electrode.

[0010] Further, the thin film transistor array is used for image reading, and the surface array is 16*16 to 256*256, and the size of each pixel is 100*100 microns to 1*1 millimeter. Preferably, the photosensitive material is selected from Bi2S3 material; the photosensitive material is patterned and grown on the substrate by a silver seed layer induction method; and the pattern of the photosensitive material on the substrate is a line array with fixed line width and line spacing, and the line width and line spacing range is 5-20 microns.

[0011] Further, the encapsulation layer covering the photosensitive material is an ultraviolet curing adhesive.

[0012] Further, the rear-end array reading system identifies and analyzes the collected handwriting image; specifically, a convolutional neural network model is used to automatically extract and train the digital features contained in the image, so as to realize the identification and classification of different digital inputs.

[0013] Further, each imaging includes a calibration process, and the calibration process is specifically: using a light source to fully illuminate the image sensor to calibrate as bright, removing the light source to fully calibrate as dark, and then imaging; each calibration process is equivalent to assigning each pixel a maximum and minimum threshold value and normalizing, and imaging is performed on this basis, so that repeated erasing and writing imaging can be realized in any current state by reassigning the threshold value and normalizing.

[0014] Another aspect of the embodiment of the present application also provides a preparation method of a non-contact handwriting device based on a ultraviolet-specific image sensor, and the non-contact handwriting device based on the ultraviolet-specific image sensor is prepared by the following steps: S1: providing a thin film transistor array.

[0015] S2: cleaning the thin film transistor array.

[0016] S3: preparing a pixel array pattern on the thin film transistor array by a photolithography process; S4: placing the line array pattern of the photosensitive material into a thermal evaporation film coating machine to evaporate silver, and obtaining a silver material pixel array after peeling off; S5: placing the silver material pixel array into a plasma cleaning machine to bombard and oxidize into an oxidized silver seed layer; S6: spin-coat the Bi2S3 precursor solution onto the silver oxide seed layer; S7: heat the chip on which the Bi2S3 precursor solution is spin-coated in a glove box.

[0017] S8: encapsulate the image sensor on which the Bi2S3 precursor solution is spin-coated with ultraviolet curing glue.

[0018] Further, the step S2 of cleaning the thin film transistor chip comprises: using ethanol to ultrasonically clean the thin film transistor chip for 5-10 minutes; The step S3 of the photoetching process comprises: uniform coating, pre-baking, exposure, and development; The step S3 of the pixel array pattern comprises: 16x16 to 256x256 pixel patterns, each pixel pattern is composed of a plurality of fixed line width and line spacing line arrays, and the line width and line spacing range is 5-20 microns.

[0019] Further, the method for preparing the non-contact handwriting device based on the ultraviolet-specific image sensor, wherein the thickness of the silver material in the step S4 is 5-30 nanometers. The purpose of this step is to pattern the silver seed layer pattern on the thin film transistor array, so as to facilitate the subsequent patterned growth of the Bi2S3 photosensitive material; The step S4 of the stripping process comprises: soaking in acetone for 5-20 minutes; In the step S5, the chip is placed in a plasma cleaning machine, and the oxidation parameters are: 2-10 sccm oxygen flow rate to bombard the chip on which the silver material is evaporated for 10-30 minutes, so as to oxidize the evaporated silver material into silver oxide; The Bi2S3 precursor solution configuration method in the step S6 is: weighing BiCl3 and thiourea, dissolving them in N,N-dimethylformamide solution, so that the molar ratio of Bi to S is 0-1:4, and stirring to mix them uniformly; The spin-coating process in the step S6 is: placing the chip in a spin coater and rotating at a speed of not less than 2000 rpm per minute for 10-60 seconds, and dropping 15-30 microliters of the mixed precursor solution during the rotation process; The heating parameters in the step S7 are: placing the chip on a hot plate at 150-200 degrees Celsius and heating for 3-10 minutes; The chip preparation process needs to repeat the steps S6 and S7, i.e. the spin-coating and heating process 1-5 times, to form a uniform and dense Bi2S3 material; The image sensor packaging method in step S8 includes: first, uniformly coating the ultraviolet curing glue on the entire surface of the image sensor chip to form a protective layer. In the gluing process, the electrode pin area is reserved to remain exposed to ensure the feasibility of subsequent circuit connection. After the gluing is completed, the ultraviolet light source is used to irradiate the curing glue layer for 5 minutes to achieve sufficient curing of the curing glue, forming a stable and uniform packaging cover layer; The patterning preparation method of the Bi2S3 material in this step is realized by using a silver seed layer stimulation method, that is, the final Bi2S3 material is selectively grown only on the place where the silver material is evaporated on the chip in step S4, and the place where the silver material is not evaporated will not grow, thereby forming a 16*16 to 256*256 pixel array pattern on the chip.

[0020] Beneficial effects: 1. The non-contact handwriting device based on the ultraviolet-specific image sensor provided by the application adopts a photoelectric detector array with memory characteristics to replace the traditional handwriting input image sensing and storage separate architecture, so that the sensing and storage are integrated to realize low-frequency handwriting image acquisition, and the storage and transmission data redundancy caused by the traditional high-frame-rate camera multi-image acquisition and processing is greatly reduced, which is beneficial to reduce power consumption and simplify the architecture. 2. The non-contact handwriting device based on the ultraviolet-specific image sensor provided by the application selects a Bi2S3 photosensitive material with wavelength-specific photoelectric current memory characteristics. The material can realize a photoelectric response holding time of hundreds of seconds in the ultraviolet band, and the photoelectric response holding time is significantly shortened to only a few seconds in the visible to near-infrared band. The response behavior is mainly due to the ultraviolet light memory mechanism based on the photothermal effect, which gives the device a selective long-term holding capability for ultraviolet input signals. By combining an ultraviolet laser pen as an input tool, the application can effectively identify the handwriting trajectory while significantly suppressing the background noise interference from the ambient visible light and infrared light, thereby improving the input recognition reliability and stability of the system in complex lighting conditions. Further, the application also supports an image resetting function based on a background normalization mechanism, thereby realizing active cleaning and repeated erasing of the writing pattern, and meeting the application requirements of multiple rounds of dynamic handwriting input. 3. The image sensor array adopted by the application has excellent optical transmittance (up to 80%) and large-area patterning capability prepared at low temperature, so that it can be directly integrated into a transparent display panel to realize the integration of display and input functions, and meet the needs of thinning and transparentization of new generation intelligent terminals. BRIEF DESCRIPTION OF DRAWINGS

[0021] Figure 1 It is a structure schematic diagram of the ultraviolet-specific image sensor in the embodiments of the application.

[0022] Figure 2A schematic diagram of a single-pixel structure of the ultraviolet-specific image sensor in the embodiment of the present application.

[0023] Figure 3 A plot of the photocurrent of a single pixel in the ultraviolet-specific image sensor in the embodiment of the present application as a function of time after irradiation with a laser at wavelengths of 320 nm, 520 nm, 1060 nm and 1342 nm for 10 seconds.

[0024] Figure 4 A plot of the photocurrent of a single pixel in the ultraviolet-specific image sensor in the embodiment of the present application as a function of time after irradiation with a laser at a wavelength of 320 nm for 1 second, 2 seconds, 4 seconds, 10 seconds, 20 seconds and 40 seconds.

[0025] Figure 5 A plot of the photocurrent of a single pixel in the ultraviolet-specific image sensor in the embodiment of the present application as a function of time after irradiation with a laser at a wavelength of 320 nm for 10 seconds at different laser powers.

[0026] Figure 6 A mechanism diagram of the ultraviolet-specific image sensor in the embodiment of the present application based on the photothermal effect to achieve long-term retention of light response.

[0027] Figure 7 A transmittance test diagram of the ultraviolet-specific image sensor in the embodiment of the present application at different wavelengths.

[0028] Figure 8A A sequence of process images of the "cherry" pattern collected by the ultraviolet-specific image sensor in the embodiment of the present application after irradiation with a laser at a wavelength of 320 nm and decayed over time.

[0029] Figure 8B A sequence of process images of the "lion" pattern collected by the ultraviolet-specific image sensor in the embodiment of the present application after irradiation with a laser at a wavelength of 320 nm and decayed over time.

[0030] Figure 9A A collected image of a handwritten number, English word and Chinese character pattern on the image sensor using an ultraviolet laser pen at a wavelength of 320 nm in the embodiment of the present application.

[0031] Figure 9B A response image of a handwritten number "2" pattern on the image sensor using laser pens at wavelengths of 320 nm, 520 nm, 808 nm and 1342 nm in the embodiment of the present application.

[0032] Figure 10 A handwritten number collection by the image sensor under 320 nm ultraviolet laser input and recognition and output through a convolutional neural network model in the embodiment of the present application. Figure 11This is the original image of the number 2 and the feature extraction process. Detailed Implementation

[0033] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0034] This invention provides a non-contact handwriting input device based on an ultraviolet-specific image sensor and its fabrication method. The device employs a photosensitive material with selective photocurrent retention capability in the ultraviolet band, enabling the retention time of pattern information on the device to reach the order of hundreds of seconds. Figure 1 As shown, the device includes: a substrate 1, a thin-film transistor array 2 formed on the substrate, a photosensitive material 3 formed on the thin-film transistor array, and an encapsulation layer 4 covering the photosensitive material. Wherein: The image sensor chip uses a transparent glass substrate.

[0035] The thin-film transistors in the thin-film transistor array use transparent ITO as electrodes. The thin-film transistor array is used for image readout, with an area of ​​16×16 to 256×256 and each pixel size ranging from 100 μm×100 μm to 1 mm×1 mm.

[0036] The thin-film transistor chip is equipped with a back-end array reading system. This system can read the current signal collected on the thin-film transistor chip and finally generate an image on the terminal APP through wireless transmission.

[0037] The photosensitive material is Bi2S3.

[0038] Photosensitive materials are patterned and grown on a substrate using a silver seed layer induction method. (This method is currently patent pending, 202410093772.8) The pattern of the photosensitive material on the substrate is a linear array with fixed line width and line spacing, ranging from 5 micrometers to 20 micrometers.

[0039] The encapsulation layer covering the photosensitive material is a UV-curable adhesive.

[0040] This invention also provides a method for fabricating a non-contact handwriting input device based on an ultraviolet-specific image sensor, the method comprising the following steps: S1: Provides a transparent thin-film transistor array; S2: Clean the thin-film transistor array; S3: Pixel array patterns are fabricated on a thin-film transistor array using photolithography. S4: Place the pixel array pattern (the linear array pattern corresponding to the photosensitive material) into a thermal evaporation coating machine to evaporate silver, and peel it off to obtain the silver material pixel array; S5: Put the thin film transistor chip with silver material into the plasma cleaning machine to bombard and oxidize into silver oxide; S6: Spin-coat the Bi2S3 precursor solution on the silver oxide seed layer.

[0041] S7: Put the chip with spin-coated Bi2S3 precursor solution into the glove box for heating.

[0042] S8: Package the image sensor with spin-coated Bi2S3 photosensitive material with ultraviolet curing glue.

[0043] In the embodiment of the present application, the Bi2S3 photosensitive material can be monolithically integrated on the thin film transistor array. (The principle and method are in the application patent 202510184631.1) In the embodiment of the present application, the step S2 of cleaning the thin film transistor array includes: using ethanol to ultrasonically clean the thin film transistor array for 5-10 minutes.

[0044] In the embodiment of the present application, the step S3 of photoetching process includes: uniform coating, pre-baking, exposure, and development.

[0045] In the embodiment of the present application, the step S3 of pixel array pattern includes 16x16 to 256x256 pixel patterns, each pixel pattern is composed of a plurality of fixed line width and line spacing line array, and the line width and line spacing range is 5-20 microns.

[0046] In the embodiment of the present application, the thickness of the silver material in step S4 is 5-30 nanometers. The purpose of this step is to pattern the silver seed layer pattern on the thin film transistor array, so as to facilitate the subsequent pattern growth of the Bi2S3 photosensitive material.

[0047] In the embodiment of the present application, the stripping process in step S4 includes: soaking in acetone for 5-20 minutes.

[0048] In the embodiment of the present application, the chip is placed in the plasma cleaning machine in step S5, and the oxidation parameters are: 2-10 sccm oxygen flow rate, and the chip with evaporated silver material is bombarded for 10-30 minutes to oxidize the evaporated silver material into silver oxide.

[0049] In the embodiment of the present application, the Bi2S3 precursor solution configuration method in step S6 is: weighing BiCl3 and thiourea, dissolving them in N, N-dimethylformamide solution, making the molar ratio of Bi to S 0-1:4, and stirring to mix them uniformly.

[0050] In the embodiment of the present application, the spin-coating process in step S6 is: placing the chip in the spin coater and rotating at a speed not less than 2000 rpm per minute for 10-60 seconds, and dropping 15-30 microliters of the mixed precursor solution during the rotation.

[0051] In this embodiment of the invention, the heating parameters in step S7 are: placing the chip on a hot plate at 150~200 degrees Celsius and heating for 3~10 minutes.

[0052] In this embodiment of the invention, the chip fabrication process requires repeating steps S6 and S7 (i.e. spin coating-heating) 1 to 5 times to form a uniform and dense Bi2S3 material.

[0053] In this embodiment of the invention, the patterning preparation method of the Bi2S3 material mentioned in this step is achieved by using a silver seed layer stimulation method. That is, the final Bi2S3 material is selectively grown only in the areas on the chip where silver material is deposited in step S4, and will not grow in the areas where silver material is not deposited, thereby forming a 16×16 to 256×256 pixel array pattern on the chip.

[0054] In this embodiment of the invention, the image sensor packaging method in step S8 includes: firstly, uniformly coating the entire surface of the image sensor chip with ultraviolet-curable adhesive to form a protective layer. During the adhesive coating process, the electrode pin area is left exposed to ensure the feasibility of subsequent electrical connections. After the adhesive coating is completed, the cured adhesive layer is irradiated with an ultraviolet light source (e.g., an ultraviolet flashlight) for 5 minutes to ensure full curing of the adhesive and the formation of a stable and uniform encapsulation cover layer.

[0055] like Figure 2 As shown Figure 1 A schematic diagram of the structure of a single pixel of the ultraviolet-specific image sensor prepared in this paper. Each pixel is composed of a linear array pattern of several lines with fixed line width and line spacing.

[0056] The performance of the ultraviolet-specific photoelectric sensor array prepared by the method described above according to the embodiments of the present invention will be verified below through actual data examples.

[0057] First, basic photoelectric performance tests were performed on the single pixel of the ultraviolet-specific image sensor.

[0058] like Figure 3As shown in the figure, when the bias voltage is 0.1 V, the photoelectric detector is irradiated with laser with a wavelength of 320 nm, 520 nm, 1060 nm and 1342 nm respectively, and the irradiation time is 10 seconds. The experimental results show that, after the laser irradiation is removed, under the condition of keeping the same photocurrent amplitude, the device shows a significantly prolonged photoelectric relaxation process after being excited by 320 nm wavelength ultraviolet laser. Further comparison of the device response under different wavelength laser excitation shows that the light response holding time of the device gradually shortens with the increase of the excitation light wavelength, showing a clear ultraviolet specific response behavior. The results show that the photoelectric detector has a longer response holding effect in the ultraviolet band, which can be used to realize the memory storage function of the ultraviolet input signal, thereby meeting the requirements of high signal-to-noise ratio and maintainability for non-contact handwriting input.

[0059] As shown in the figure, Figure 4 When the bias voltage is 0.1 V, the photoelectric detector is irradiated with ultraviolet laser with a wavelength of 320 nm and a power of 9.1 mW / cm 2 , and the irradiation time is 1 second, 2 seconds, 4 seconds, 10 seconds, 20 seconds and 40 seconds respectively. After the light is removed, the holding time of the photocurrent is tested. The experimental results show that the photocurrent relaxation time of the device significantly increases with the extension of the irradiation time, that is, the memory effect under ultraviolet excitation shows a positive correlation accumulation characteristic with the irradiation time. This phenomenon shows that the photoelectric detector has good time-adjustable ultraviolet light memory ability, and can realize the memory reinforcement of the input information according to the irradiation duration, providing programmable response characteristics with controllable time length for non-contact handwriting input.

[0060] As shown in the figure, Figure 5 When the bias voltage is 0.1 V, the photoelectric detector is irradiated with ultraviolet laser with a wavelength of 320 nm and a power of 9.1 mW / cm 2 , and the irradiation time is 1 second, 2 seconds, 4 seconds, 10 seconds, 20 seconds and 40 seconds respectively. After the light is removed, the holding time of the photocurrent is tested. The experimental results show that the photocurrent relaxation time of the device significantly increases with the extension of the irradiation time, that is, the memory effect under ultraviolet excitation shows a positive correlation accumulation characteristic with the irradiation time. This phenomenon shows that the photoelectric detector has good time-adjustable ultraviolet light memory ability, and can realize the memory reinforcement of the input information according to the irradiation duration, providing programmable response characteristics with controllable time length for non-contact handwriting input.

[0061] Figure 6This diagram illustrates the long-term photoresponse retention mechanism of the ultraviolet-specific image sensor in this embodiment of the invention. The mechanism is based on the photothermoelectric effect induced by ultraviolet light. As the temperature increases, the dark current of the device significantly increases, indicating that charge carriers are difficult to recombine rapidly under thermal excitation, leading to the accumulation of photogenerated charge carriers in a quasi-stable state within the device, thus achieving sustained photocurrent retention. Because ultraviolet light has higher photon energy, its irradiation significantly increases the device temperature, resulting in an increase in dark current with rising temperature. This process promotes the thermal excitation and redistribution of charge carriers and inhibits the rapid recombination of photogenerated charge carriers, thereby achieving long-term retention of the input optical signal. Compared to visible and infrared light, ultraviolet light more readily excites a sustained response in this structure, indicating that the long-term retention behavior of the device is mainly dominated by an ultraviolet-induced photothermal driving mechanism, exhibiting ultraviolet-selective response characteristics similar to synaptic memory.

[0062] Figure 7 The image shows the transmittance test results of the ultraviolet-specific image sensor in this embodiment of the invention at different wavelengths. The experimental results show that the image sensor has good optical transmittance performance in the visible light band, with a maximum transmittance of about 80%, indicating that it has the potential to be integrated with display modules.

[0063] Secondly, the memory imaging performance of the 64×64 pixel array ultraviolet specific image sensor chip of the present invention was verified.

[0064] like Figure 8A As shown, under ultraviolet laser irradiation with a wavelength of 320 nanometers, the image sensor successfully captured and formed a "cherry" pattern. After the laser was removed, the pattern gradually decayed over time and could be retained on the device for up to about 90 seconds, demonstrating a remarkable image memory retention capability. Figure 8B The image shows a "lion" pattern recorded under the same test conditions. It exhibits similar retention behavior after the illumination stops, showing a strong time-dependent visual persistence effect.

[0065] It is worth noting that, Figure 8A and Figure 8B The high pixel density created by the 64×64 array structure results in clear and vivid pattern details, reflecting the image sensor chip's high-resolution image reproduction capability and excellent ultraviolet-specific optical memory response performance in ultraviolet imaging scenarios. This characteristic verifies the practical value of the image sensor of this invention in applications such as non-contact handwriting input and visual ultraviolet information storage.

[0066] Furthermore, the present invention also supports an image reset function based on a background normalization mechanism, thereby enabling active clearing and repeated erasing of the writing pattern, meeting the application requirements of multi-round dynamic handwriting input.

[0067] like Figure 9AAs shown, the ultraviolet laser pen with a wavelength of 320 nanometers is used for handwriting input operation on the surface of the ultraviolet-specific image sensor, and the numbers "0-9", English words and commonly used Chinese characters are written in turn. The experimental results show that the written handwriting patterns can be clearly captured by the image sensor, and remain significantly identifiable on the device after the light is removed, showing good pattern retention ability and stable ultraviolet memory response characteristics. The results further verify the applicability and practical performance of the application in the multi-language handwriting input scene.

[0068] As shown in Figure 9B The laser pens with wavelengths of 320 nanometers, 520 nanometers, 808 nanometers and 1342 nanometers are used for handwriting input test of the number "2" on the image sensor. The experimental results show that only under the excitation condition of 320 nanometer ultraviolet band, the image sensor can clearly record and retain the written number "2" pattern; while under the excitation condition of 520 nanometers to 1342 nanometers, the image sensor fails to produce effective pattern response, and the pattern is invisible. This phenomenon fully shows that the image sensor has a highly selective response characteristic to ultraviolet light, and has excellent ultraviolet-specific memory ability, which is beneficial to effectively suppress the environmental visible light and infrared interference in practical application.

[0069] As shown in Figure 10 In combination with machine learning technology, the handwritten number patterns collected by the image sensor are recognized and analyzed. Specifically, the convolutional neural network model is used to automatically extract and train the number features contained in the image, so as to realize high-precision recognition and classification of different number inputs. This method effectively improves the processing ability of the system for handwriting input data, has good generalization performance and application expansibility, and can be applied to real-time input recognition requirements in various intelligent interaction scenes.

[0070] As shown in Figure 10 In combination with machine learning technology, the handwritten number patterns collected by the image sensor are recognized and analyzed. Specifically, the convolutional neural network model is used to automatically extract and train the number features contained in the image, so as to realize high-precision recognition and classification of different number inputs.

[0071] In the embodiment of the present application, the convolutional neural network model comprises an input part, a feature extraction part and a classification part. The input part obtains the response value of the image sensor and converts it into a 256-order gray value as a single-channel input of the convolutional neural network model. The feature extraction part comprises three convolutional layers and three pooling layers. The first convolutional layer uses 16 3*3 convolutional kernels and applies a ReLU activation function, followed by a 2*2 maximum value pooling layer. The second convolutional layer uses 32 3*3 convolutional kernels and applies a ReLU activation function, followed by a 2*2 maximum value pooling layer. The third convolutional layer uses 64 3*3 convolutional kernels and applies a ReLU activation function, followed by a 2*2 maximum value pooling layer. The classification part flattens the extracted feature image into a one-dimensional vector, followed by a fully connected layer comprising 128 neurons, which applies a ReLU activation function. Finally, 10 output neurons corresponding to 0-9 digit classification are connected, which adopts a Softmax activation function to output a probability distribution, and the corresponding prediction result is output according to the probability distribution. As shown in Figure 11 The original image of the number 2 and the feature extraction process are shown in FIG. 2.

[0072] In the embodiment of the present application, the digital feature refers to the feature map formed after the convolutional layer and the pooling layer after the handwritten digital image input into the convolutional neural network model. As shown in Figure 11 The original image and the feature extraction process are shown in FIG. 2. In the first convolutional kernel, the feature map can obtain shallow simple local edges. The feature map of the second convolutional kernel can obtain specific arc features. The third convolutional kernel obtains deep stroke segment information. The neurons of the hidden layer can obtain the handwritten digital prediction result by analyzing different deep features contained in different numbers.

[0073] The convolutional neural network model is obtained by pre-training. In the pre-training process, the handwritten digital data is collected by the image sensor to obtain multiple frames of handwritten digital images. After the images are labeled, they are combined with the MNIST handwritten digital library to construct a training set, which is input into the convolutional neural network model. The model will output a prediction result. The loss is calculated according to the prediction result and the actual label. The parameters of the target model are continuously adjusted until the test result accuracy is high, and the model adjustment is stopped. The weight data of the target model is obtained.

[0074] The method of automatically extracting and training the digital features contained in the image by using the convolutional neural network model effectively improves the processing capability of the system for handwritten input data, has good generalization performance and application expansion, and can be applied to real-time input recognition requirements in various intelligent interaction scenes.

[0075] In summary, the above is only a preferred embodiment of the present application, and is not used to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A non-contact handwriting device based on an ultraviolet-specific image sensor, characterized in that, It comprises: a substrate (1), a thin film transistor array (2) formed on the substrate, a photosensitive material (3) formed on the thin film transistor array (2), and an encapsulation layer (4) covering the photosensitive material (3); An ultraviolet laser pen is used for handwriting input operation on the surface of the encapsulation layer (4), and the thin film transistor array (2) is loaded with a back-end array reading system which reads the current signal collected on the thin film transistor array (2) to generate a handwriting image which is displayed on a terminal APP through wireless transmission.

2. The ultraviolet-specific image sensor-based contactless handwriting device of claim 1, wherein, The substrate (1) is a transparent glass substrate, and the transparent ITO is an electrode.

3. The ultraviolet-specific image sensor-based contactless handwriting device of claim 1, wherein, The thin film transistor array (2) is used for image readout, and the surface array is 16x16 to 256x256, and the size of each pixel is 100x100 microns to 1mmx1mm.

4. The ultraviolet-specific image sensor-based contactless handwriting device of claim 1, wherein, The photosensitive material (3) is selected from Bi2S3 material. The photosensitive material (3) is patterned and grown on the substrate by a silver seed layer induction method. The pattern of the photosensitive material (3) on the substrate is a line array with fixed line width and line spacing, and the line width and line spacing range is 5 microns-20 microns.

5. The ultraviolet-specific image sensor-based contactless handwriting device according to claim 1, wherein, The encapsulation layer (4) covering the photosensitive material (3) is an ultraviolet curing adhesive.

6. The ultraviolet-specific image sensor-based contactless handwriting device according to claim 1, wherein, The back-end array reading system identifies and analyzes the collected handwriting image; specifically, a convolutional neural network model is used to automatically extract and train the digital features contained in the image, thereby realizing the identification and classification of different digital inputs.

7. The ultraviolet-specific image sensor-based contactless handwriting device according to claim 1, wherein, It also includes a calibration process before each imaging, which is specifically: using a light source to fully illuminate the image sensor to calibrate as bright, removing the light source to fully calibrate as dark, and then imaging; each calibration process is equivalent to assigning each pixel a maximum and minimum threshold and normalizing, and then imaging, so that in any current state, repeated erasing and imaging can be realized by reassigning the threshold and normalizing.

8. A method of manufacturing a non-contact handwriting device based on an ultraviolet-specific image sensor, characterized by, For the non-contact handwriting device based on the ultraviolet specific image sensor as claimed in claims 1-7, the following steps are adopted for preparation: S1: providing a thin film transistor array (2); S2: cleaning the thin film transistor array (2); S3: preparing a pixel array pattern on the thin film transistor array (2) by a photolithography process; S4: placing the corresponding line array pattern of the photosensitive material (3) into a thermal evaporation film deposition machine to evaporate silver, and peeling off to obtain a silver material pixel array; S5: placing the silver material pixel array into a plasma cleaning machine to bombard and oxidize into an oxidized silver seed layer; S6: spin coating Bi2S3 precursor solution onto the oxidized silver seed layer; S7: placing the image sensor with spin-coated Bi2S3 precursor solution into a glove box for heating. S8: encapsulating the image sensor with spin-coated Bi2S3 precursor solution with ultraviolet curing adhesive.

9. The method of claim 8, wherein the method further comprises: forming a plurality of pixels on the substrate; and forming a plurality of color filters on the substrate, wherein each color filter is associated with a corresponding pixel. The cleaning of the thin film transistor chip in step S2 includes: ultrasonic cleaning the thin film transistor chip with ethanol for 5-10 minutes; The photolithography process in step S3 includes: uniform coating, pre-baking, exposure, and development; The pixel array pattern of step S3 includes 16x16 to 256x256 pixel patterns, each pixel pattern being composed of a plurality of fixed line width and line spacing line arrays, and the line width and line spacing range is 5-20 microns.

10. The method of claim 8, wherein the ultraviolet-specific image sensor-based non-contact handwriting device is prepared by the method, and the method comprises the following steps: S1, preparing a chip; S2, preparing a pixel array pattern; S3, preparing a silver material; S4, preparing a silver seed layer pattern; S5, preparing a silver oxide layer; S6, preparing a Bi2S3 precursor solution; S7, preparing a Bi2S3 material; S8, packaging the image sensor; and S9, packaging the image sensor. The thickness of the silver material evaporated in step S4 is 5-30 nanometers. The purpose of this step is to pattern the silver seed layer pattern on the thin film transistor array, so as to facilitate the subsequent patterned growth of the Bi2S3 photosensitive material; The stripping process in step S4 includes: soaking in acetone for 5-20 minutes; In step S5, the chip is placed in a plasma cleaning machine, and the oxidation parameters are: 2-10 sccm oxygen flow rate, and the chip evaporated with silver material is bombarded for 10-30 minutes to oxidize the evaporated silver material into silver oxide; In step S6, the Bi2S3 precursor solution is prepared by: weighing BiCl3 and thiourea, dissolving them in N,N-dimethylformamide solution, making the molar ratio of Bi to S 0-1:4, and stirring to make them uniformly mixed; In step S6, the spin coating process is: placing the chip in a spin coater and rotating at a speed of not less than 2000 rpm per minute for 10-60 seconds, and dropping 15-30 microliters of the uniformly mixed precursor solution during the rotation; In step S7, the heating parameters are: placing the chip on a 150-200 degree Celsius hot plate and heating for 3-10 minutes; In the chip preparation process, steps S6 and S7, i.e. the spin coating-heating process, need to be repeated 1-5 times to form a uniform and dense Bi2S3 material; In step S8, the image sensor packaging method includes: first, uniformly coating ultraviolet curing glue on the entire surface of the image sensor chip to form a protective layer. During the glue coating process, the electrode pin area is reserved to remain exposed to ensure the feasibility of subsequent circuit connection. After the glue coating is completed, the curing glue layer is irradiated using a UV light source for 5 minutes to achieve sufficient curing of the curing glue and form a stable and uniform packaging cover layer; The patterned preparation method of the Bi2S3 material in this step is achieved by using the silver seed layer stimulation method, i.e. the final Bi2S3 material only selectively grows in the places where silver material is evaporated on the chip in step S4, and no growth occurs in places where silver material is not evaporated, thereby forming a 16x16 to 256x256 pixel array pattern on the chip.

Images