Key triggering method based on multi-group optical induction and deep learning fusion

Abstract

The invention relates to an intelligent identification technology, and discloses a key triggering method based on multi-group optical induction and deep learning fusion, comprising the following steps: using an optical sensor of a key device to synchronously collect multi-dimensional optical data of a preset key action, the optical sensor being located below a key of the key device; performing convolution processing on the multi-dimensional optical data to obtain optical data features; and inputting the optical data features into a pre-constructed key trigger recognition model, and outputting a key trigger type by using the pre-constructed key trigger recognition model. The invention further provides a key triggering device based on multi-group optical induction and deep learning fusion. According to the invention, the mistaken touch rate of the key can be reduced and the key triggering precision can be improved.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent recognition technology, and in particular relates to a key triggering method based on the fusion of multiple groups of optical sensing and deep learning. Background Art

[0002] Currently, the triggering technology of optical axis keyboards mainly relies on a single set of infrared transmitter-receiver tubes, which trigger key signals by switching on and off the optical path. They are specifically divided into two categories:

[0003] One is the optical triggering technology, which takes the emission-blocking-receiving link of infrared light as the core and realizes the non-contact triggering of key signals by precisely controlling the state of the optical path. The specific process is as follows: First, a single group of infrared LEDs emits 850nm or 940nm near-infrared light, which is compressed into a half-angle by a micro-condensing lens.

[0004] After the collimated light beam with a beam angle of ≤30° is aligned, it is directed to the infrared photosensitive tubes with a spacing of 1-3mm along the coaxial arrangement. In terms of status determination, when no key is pressed, the black ABS grating driven by the handle completely blocks the light path. At this time, the light intensity received by the photosensitive tube is less than 0.1mW / cm2, which is in the cut-off region. The collector-emitter resistance is greater than 10MΩ, and the circuit outputs a high level (such as 5V). The system determines that it is "not triggered". When the key is pressed, the grating moves 1.5-2.0mm (occupying 60%-80% of the light path spacing) with the handle to move out of the light path, and the light intensity received by the photosensitive tube is greater than 5mW / cm 2 , entering the saturation region, the collector-emitter resistance is less than 100Ω, the circuit output is low (e.g., 0V), and the system determines it as "triggered." To combat interference, an 850±20nm bandpass filter at the front of the phototransistor attenuates ambient light interference, while a narrow slit optical path with a width of ≤0.8mm limits non-axial light, ensuring signal triggering accuracy and interference resistance.

[0005] Another approach involves optical coupling, which achieves reliable conversion of optical path states into electrical signals through precise coordination of optical components and optimized signal chains. Similarly, optical coupling relies on optical components, primarily infrared light-emitting diodes (LEDs), infrared phototransistors (phototransistors), handles, and gratings, for light sensing. It identifies open and closed circuit states through optical path state and signal mapping. In terms of anti-interference technology, a combination of "spectral filtering-spatial restriction-temperature compensation" technologies is adopted: first, a bandpass filter with a passband of 940±20nm is pasted on the front end of the photosensitive tube to effectively attenuate visible light such as indoor lighting, red light, and far-infrared light in the 10μm band radiated by the human body, with a filtration efficiency of over 95%; second, the optical path channel adopts a narrow slit design with a width of 0.8mm, isolated by black shading plates on both sides, allowing only axial light to pass through, suppressing side ambient light scattering such as desktop reflections; third, in response to the red shift phenomenon of the light-emitting tube wavelength of +0.3nm / ℃ as the temperature increases, the power supply voltage is dynamically adjusted by connecting a 10kΩ NTC thermistor in series to ensure that the response deviation of the photosensitive tube is less than 3% within the operating temperature range of -10℃ to 50℃, thereby improving the anti-interference ability and environmental adaptability of the optical axis technology in multiple dimensions.

[0006] Existing technologies have significant limitations in their sensing and anti-interference capabilities. For example, in terms of sensing capabilities, existing optical axis technology relies on the opposing layout of a single set of infrared light-emitting diodes and photosensitive tubes, and can only output binary signals (on / off) through the "light path on / off". It is unable to capture the continuously changing characteristics of optical signals, such as the change in pressing force reflected by the slope of the light intensity change during the key press, and the dynamic change in the occlusion area corresponding to the grating offset trajectory. The hardware lacks multiple sensor arrays, and it is impossible to obtain key information such as pressing force, speed, and spatial position through multi-angle and multi-wavelength light source configurations. The signal processing level uses Schmitt triggers to directly quantize analog signals into high and low levels, discarding transition state data and timing characteristics near the threshold. This can only achieve simple "triggered / not triggered" judgments, making it difficult to support refined interaction needs. For example, in games, it is impossible to distinguish between light and heavy taps, and cannot meet users' needs for high-precision interaction with input devices. In addition, in terms of anti-interference mechanism, existing technologies achieve anti-interference through fixed-parameter bandpass filters (such as a central wavelength of 940±20nm bandpass range) and narrow-slit mechanical structures. However, this static design cannot cope with ambient light wavelength drift (such as the introduction of unexpected infrared bands by new light sources), aging caused by long-term use of components (such as grating wear that reduces shading efficiency), and physical offsets caused by temperature changes (such as thermal expansion and contraction causing optical path alignment deviations > 0.2mm). The signal processing level relies on fixed thresholds (such as light intensity > 5mW / cm 2 Trigger) and RC filter circuits lack the ability to learn and distinguish dynamic interference (such as instantaneous strong light pulses and non-sustained mechanical vibrations), and the false trigger rate can be as high as 15% in complex environments. Existing trigger technologies have a high false trigger rate and low trigger accuracy. Summary of the Invention

[0007] The present invention provides a key triggering method based on the fusion of multiple groups of optical sensors and deep learning, which can reduce the false touch rate of keys and improve the key triggering accuracy.

[0008] To achieve the above objectives, the present invention provides a key triggering method based on the fusion of multiple optical sensors and deep learning, comprising:

[0009] Synchronously collecting multi-dimensional optical data of a preset key action using an optical sensor of a key device, wherein the optical sensor is located below a key of the key device;

[0010] Perform convolution processing on multi-dimensional optical data to obtain optical data features;

[0011] The optical data features are input into a pre-built key trigger recognition model, and the pre-built key trigger recognition model is used to output the key trigger type.

[0012] Optionally, the optical sensor includes multiple groups of optical transmitting-receiving pairs of transistors distributed under multiple keys of the key device, and a preset number of groups of optical transmitting-receiving pairs of transistors are arranged in a matrix distribution under each key.

[0013] Optionally, each set of optical transmitting-receiving pairs includes:

[0014] Infrared LED, used to support 850nm / 940nm dual wavelength optional, with a micro-condenser lens to compress the beam half angle to 25°, forming a highly collimated optical path;

[0015] Phototransistor, front-end integrated 850±20nm or 940±20nm bandpass filter;

[0016] The button is equipped with a handle with a trapezoidal grating. The edge of the grating is designed to have a 45° slope. When the button is pressed, the displacement of the grating along the light path changes linearly with the blocked area.

[0017] Optionally, the synchronously collecting multi-dimensional optical data of the preset key action using an optical sensor includes:

[0018] Initialization step: When the key is not pressed, the trapezoidal grating completely blocks the light paths of all optical transmitter-receiver pairs under the key, and the light receiving circuits of all optical transmitter-receiver pairs under the key output a high level;

[0019] Dynamic acquisition step: When the button is pressed, all optical transmitting-receiving tubes under the button synchronously collect multi-dimensional optical data.

[0020] Optionally, all optical transmitting-receiving pairs under the button synchronously collect multi-dimensional optical data, including: collecting light intensity timing curves to record the continuous change process of light intensity in a preset light intensity range; collecting the percentage of blocked area, calculating the real-time blocked area change through the geometric relationship between the grating displacement and the optical path spacing, and calculating the displacement trajectory of the button being pressed; collecting wavelength offset: by switching infrared LEDs of different wavelengths, it is used to monitor the impact of ambient light wavelength drift on the response of the optical transmitting-receiving pairs.

[0021] Optionally, the key trigger recognition model includes a cascaded input layer, a convolutional layer, a pooling layer, and a fully connected layer;

[0022] Among them, the data dimension of the input layer is the number of sensors × the number of sampling points;

[0023] The convolution layer is a stack of 3 layers of 3×1 convolution kernels, which extracts features through a local sliding window;

[0024] The pooling layer uses the maximum pooling strategy to reduce the dimension of the convolutional layer output, retaining the peak features in each segment of data; the fully connected layer converts the feature vector into a key action probability value through multi-layer neuron mapping and outputs the key trigger type.

[0025] In order to solve the above problems, the present invention also provides a key triggering device based on the fusion of multiple sets of optical sensing and deep learning, the device comprising: a controller, and a key device connected to the controller;

[0026] The controller executes the steps of the key triggering method based on the fusion of multiple sets of optical sensing and deep learning.

[0027] Optionally, the spacing between two adjacent groups of optical transmitting-receiving tube pairs is 1-3 mm, and each group of optical transmitting-receiving tube pairs is configured with a bandpass filter and a micro-condensing lens, wherein the bandpass filter allows light wavelengths in the range of 850±20 nm or 940±20 nm to pass, and the micro-condensing lens compresses the light beam half angle ≤30°.

[0028] Optionally, the button includes a handle connected to the trapezoidal grating.

[0029] The present invention utilizes an optical sensor to synchronously collect multi-dimensional optical data of preset key actions, thereby obtaining continuous and multi-dimensional optical data, and further extracting the continuously changing characteristics of the optical data. In addition, the multi-dimensional optical data is subjected to convolution processing to obtain optical data features, which can efficiently extract key features from the optical data. In addition, the optical data features are input into a pre-built key trigger recognition model, and the pre-built key trigger recognition model is used to output the key trigger type, which can reduce the false touch rate of the key and improve the key trigger accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 A flowchart of a key triggering method based on the fusion of multiple optical sensors and deep learning provided by one embodiment of the present invention.

[0031] Figure 2 This is a state diagram of the optical path of a traditional optical axis single-group sensor of a key triggering method based on the fusion of multiple optical sensing groups and deep learning provided by one embodiment of the present invention.

[0032] Figure 3 This is a diagram of the optical path disconnection state of a traditional optical axis single-group sensor in a key triggering method based on the fusion of multiple optical sensors and deep learning provided by one embodiment of the present invention.

[0033] Figure 4 This is a mechanical layout diagram of a traditional optical axis transmitting tube according to a key triggering method based on the fusion of multiple sets of optical sensing and deep learning provided by one embodiment of the present invention.

[0034] Figure 5 This is a mechanical layout diagram of a traditional optical axis receiving tube for a key triggering method based on the fusion of multiple sets of optical sensing and deep learning provided by one embodiment of the present invention.

[0035] Figure 6 This is a mechanical layout diagram of a traditional optical axis shielding plate according to a key triggering method based on the fusion of multiple sets of optical sensors and deep learning provided by one embodiment of the present invention.

[0036] Figure 7 A side view of a conventional optical coupling technology for a key triggering method based on the fusion of multiple optical sensors and deep learning, provided in one embodiment of the present invention.

[0037] Figure 8 Diagram of optical path separation of traditional optical coupling technology for a key triggering method based on the fusion of multiple optical sensors and deep learning, provided in one embodiment of the present invention.

[0038] Figure 9 A diagram showing the optical path disconnection of a traditional optical coupling technology for a key triggering method based on the fusion of multiple optical sensors and deep learning, provided in one embodiment of the present invention.

[0039] Figure 10 This is a light path diagram of a traditional optical coupling technology for a key triggering method based on the fusion of multiple sets of optical sensors and deep learning provided by one embodiment of the present invention.

[0040] Figure 11 This is a layout diagram of multiple sets of side-mounted tube pairs using traditional optical coupling technology based on a key triggering method that integrates multiple sets of optical sensing and deep learning, provided in one embodiment of the present invention.

[0041] Figure 12A diagram of optical data collected by multiple sets of sensors in a key triggering method based on the fusion of multiple sets of optical sensing and deep learning provided by one embodiment of the present invention.

[0042] Figure 13 A diagram showing the sensor array distribution for a key triggering method based on the fusion of multiple optical sensors and deep learning, provided in one embodiment of the present invention.

[0043] Figure 14 This is a model structure design diagram of a key triggering method based on the fusion of multiple sets of optical sensors and deep learning provided by one embodiment of the present invention.

[0044] Figure 15 This is a flowchart of the triggering principle of a key triggering method based on the fusion of multiple sets of optical sensors and deep learning provided by one embodiment of the present invention.

[0045] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0046] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0047] The embodiment of the present application provides a key triggering method based on the fusion of multiple sets of optical sensing and deep learning. The execution subject of the key triggering method based on the fusion of multiple sets of optical sensing and deep learning includes but is not limited to at least one of the electronic devices such as the server, the terminal, etc. that can be configured to execute the method provided by the embodiment of the present application. In other words, the key triggering method based on the fusion of multiple sets of optical sensing and deep learning can be executed by software or hardware installed on the terminal device or the server device, and the software can be a blockchain platform. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc. The server can be an independent server, or it can be a cloud server that provides cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDN), and basic cloud computing services such as big data and artificial intelligence platforms.

[0048] Reference Figure 1 FIG2 is a flow chart of a key triggering method based on the fusion of multiple optical sensors and deep learning according to an embodiment of the present invention. In this embodiment, the key triggering method based on the fusion of multiple optical sensors and deep learning includes:

[0049] S1. Use an optical sensor to synchronously collect multi-dimensional optical data of preset key actions.

[0050] In an embodiment of the present invention, an optical sensor refers to a device that uses optical principles to detect changes in light signals (such as light intensity, wavelength, phase, etc.) and converts them into electrical signals or other processable forms. Its core is to realize the conversion of optical signals into electrical signals, thereby sensing, measuring or controlling physical quantities, environmental parameters or target states.

[0051] As an embodiment of the present invention, the use of an optical sensor to synchronously collect multi-dimensional optical data of preset key actions includes: the optical sensor includes multiple groups of optical transmitting-receiving pairs of tubes dispersed under multiple keys of a key device, and a preset number of optical transmitting-receiving pairs of tubes are arranged in a matrix distribution under a single key.

[0052] In the embodiment of the present invention, the preset number of groups can be set to 2-4 groups, such as one group each for the upper and lower and left and right, with a group spacing of 2 mm.

[0053] Furthermore, each set of optical transmitting-receiving tubes includes:

[0054] Infrared LED, used to support 850nm / 940nm dual wavelength optional, with a micro-condenser lens to compress the beam half angle to 25°, forming a highly collimated optical path;

[0055] Phototransistor, front-end integrated 850±20nm or 940±20nm bandpass filter;

[0056] The button is equipped with a handle with a trapezoidal grating. The edge of the grating is designed to have a 45° slope. When the button is pressed, the displacement of the grating along the light path changes linearly with the blocked area.

[0057] For example, the arrangement of a preset number of optical transmitting-receiving pairs distributed in a matrix under a single key may be implemented as follows:

[0058] ① Infrared LED: Use model 1R12-21C / TR8, which supports dual wavelengths of 850nm / 940nm. Combined with a micro-condenser lens, the beam half angle is compressed to 25°, forming a highly collimated optical path and improving signal stability.

[0059] ② Phototransistor: Model T12-21C / TR8, with an integrated 850±20nm or 940±20nm bandpass filter at the front end, with a filtration efficiency of >95%, effectively attenuating ambient light interference;

[0060] ③Button: Equipped with a button handle with a trapezoidal grating. The grating edge is designed with a 45° slope to ensure that when the button is pressed, the displacement of the grating along the light path and the blocked area are linearly related, which facilitates accurate analysis of the pressing force and speed.

[0061] Furthermore, the synchronous collection of multi-dimensional optical data of the preset key action by the optical sensor includes:

[0062] Initialization step: When the key is not pressed, the trapezoidal grating completely blocks the light paths of all optical transmitter-receiver pairs under the key, and the light receiving circuits of all optical transmitter-receiver pairs under the key output a high level;

[0063] Dynamic acquisition step: When the button is pressed, all optical transmitting-receiving tubes under the button synchronously collect multi-dimensional optical data.

[0064] Furthermore, all optical transmitter-receiver pairs under the button synchronously collect multi-dimensional optical data, including: collecting light intensity timing curves to record the continuous change process of light intensity in a preset light intensity range; collecting the percentage of blocked area, calculating the real-time blocked area change through the geometric relationship between grating displacement and optical path spacing, and calculating the displacement trajectory of the button press; collecting wavelength offset: by switching infrared LEDs of different wavelengths, it is used to monitor the impact of ambient light wavelength drift on the response of the optical transmitter-receiver pairs.

[0065] Exemplarily, the synchronous collection of multi-dimensional optical data of a preset key action by an optical sensor may be implemented by the following steps:

[0066] Initialization steps: When no key is pressed, the trapezoidal grating completely blocks all sensor light paths. The phototransistor is in the cutoff region because the received light intensity is less than 0.1mW / cm2, and the resistance is greater than 10MΩ. The corresponding circuit outputs a high level (open circuit state).

[0067] Dynamic acquisition steps: When the button is pressed, multiple sets of sensors synchronously collect multi-dimensional optical data at a sampling frequency of 1000Hz. The multi-dimensional optical data includes: light intensity time series curve: recording light intensity from <0.1mW / cm 2 to>5mW / cm 2 The continuous change process of the grating is used to extract the slope feature to characterize the pressing force; the percentage of the blocked area: the geometric relationship between the grating displacement and the optical path spacing is used to calculate the real-time blockage area change and deduce the displacement trajectory of the button press; the wavelength offset: by switching different wavelength LEDs (850nm / 940nm), the influence of the ambient light wavelength drift on the sensor response is monitored to achieve multi-spectral anti-interference.

[0068] S2. Perform convolution processing on the multi-dimensional optical data to obtain optical data features.

[0069] Extracting optical data features from multidimensional optical data using convolutional neural networks.

[0070] In an embodiment of the present invention, the convolution layer is a stack of three layers of 3×1 convolution kernels, which extracts features such as light intensity mutation points and slope changes through a local sliding window, for example, identifying the light intensity jump rate at the moment a key is pressed.

[0071] S3. Input the optical data features into a pre-built key trigger recognition model, and use the pre-built key trigger recognition model to output the key trigger type.

[0072] As an embodiment of the present invention, the key trigger recognition model includes a cascaded input layer, a convolutional layer, a pooling layer, and a fully connected layer;

[0073] Among them, the data dimension of the input layer is the number of sensors × the number of sampling points;

[0074] The convolution layer is a stack of 3 layers of 3×1 convolution kernels, which extracts features through a local sliding window;

[0075] The pooling layer uses the maximum pooling strategy to reduce the dimension of the convolution layer output and retain the peak features in each segment of data;

[0076] The fully connected layer converts the feature vector into a key action probability value through multi-layer neuron mapping and outputs the key trigger type.

[0077] Exemplarily, the outputting of the key trigger type by using the pre-built key trigger recognition model may be implemented by the following steps:

[0078] ① Input layer: Receives normalized optical data from multiple sets of sensors. The data dimension is "number of sensors × number of sampling points". For example, the input dimension for four sets of sensors at a sampling frequency of 1000 Hz is 4 × 1000. This ensures that the spatial distribution information and temporal variation characteristics of each sensor are preserved.

[0079] ② Convolutional layer (conv1D-1 / 2 / 3): Stacks three layers of 3×1 convolution kernels and uses a local sliding window to extract temporal features such as light intensity mutation points and slope changes. For example, it can identify the rate of light intensity increase at the moment a key is pressed.

[0080] ③ Pooling layer (MaxPooling1D): Use the maximum pooling strategy to reduce the dimension of the convolution layer output, retain the peak features in each segment of data, compress the data dimension while avoiding the loss of key information;

[0081] ④ Fully connected layer: Through multi-layer neuron mapping, the feature vector is converted into a key action probability value to obtain the key trigger type. For example, the output "light tap" probability is 0.92 and "heavy tap" probability is 0.08, realizing multi-category classification.

[0082] Furthermore, the method of outputting the key trigger type using the pre-built key trigger recognition model further includes: training the pre-built key trigger recognition model.

[0083] For example, the following steps may be used to train the pre-built key trigger recognition model:

[0084] Large-scale dataset construction: Over 100,000 samples were collected, covering multiple scenarios such as 0-100g pressing force (resolution 5g), 10-50mm / s pressing speed (resolution 5mm / s), and key center / edge pressing, simulating real user input habits.

[0085] The labeling system includes three types of tags:

[0086] ①Key ID: accurate to a single key (such as A key, Ctrl key) and key combination (such as Ctrl+Alt+Delete);

[0087] ②Force level: divided into three levels: light (0-40g), medium (40-70g), and heavy (70-100g);

[0088] ③Trigger type: Distinguish between single-key trigger and combination-key trigger, and support complex type recognition.

[0089] For example, the pre-built key trigger recognition model training strategy is shown in the following steps:

[0090] ① Loss function: Use the cross entropy loss function to optimize the matching degree between the probability output of the multi-classification problem and the true label;

[0091] ②Optimizer: Select Adam optimizer, set the initial learning rate to 0.001, and accelerate convergence by dynamically adjusting the first-order moment and second-order moment estimates;

[0092] ③ Training goal: Through 100 rounds of iterative training, the test set classification accuracy is greater than 99% and the false trigger rate is less than 0.5%, ensuring the robustness of the pre-built key trigger recognition model in complex scenarios.

[0093] Furthermore, the key trigger type is transmitted to the computer device via the USB protocol or the PS / 2 protocol, supporting an intelligent recognition function that adapts to the key habits of different users.

[0094] The present invention utilizes an optical sensor to synchronously collect multi-dimensional optical data of preset key actions, thereby obtaining continuous and multi-dimensional optical data, and further extracting the continuously changing characteristics of the optical data. In addition, the multi-dimensional optical data is subjected to convolution processing to obtain optical data features, which can efficiently extract key features from the optical data. In addition, the optical data features are input into a pre-built key trigger recognition model, and the pre-built key trigger recognition model is used to output the key trigger type, which can reduce the false touch rate of the key and improve the key trigger accuracy.

[0095] like Figure 2 As shown, it is a state diagram of the optical path of a traditional optical axis single-group sensor of a key triggering method based on the fusion of multiple groups of optical sensing and deep learning provided by an embodiment of the present invention.

[0096] like Figure 3 As shown, it is a diagram of the optical path disconnection state of a traditional optical axis single-group sensor of a key triggering method based on the fusion of multiple groups of optical sensing and deep learning provided by an embodiment of the present invention.

[0097] like Figure 4 As shown, it is a mechanical layout diagram of a traditional optical axis transmitting tube according to a key triggering method based on the fusion of multiple sets of optical sensing and deep learning provided by one embodiment of the present invention.

[0098] like Figure 5 As shown, it is a mechanical layout diagram of a traditional optical axis receiving tube according to a key triggering method based on the fusion of multiple sets of optical sensing and deep learning provided by one embodiment of the present invention.

[0099] like Figure 6 As shown, it is a mechanical layout diagram of a traditional optical axis shielding plate according to a key triggering method based on the fusion of multiple sets of optical sensing and deep learning provided by an embodiment of the present invention.

[0100] like Figure 7 , which is a side view of a traditional optical coupling technology for a key triggering method based on the fusion of multiple sets of optical sensing and deep learning provided by an embodiment of the present invention.

[0101] like Figure 8 As shown, it is a light path separation diagram of a traditional optical coupling technology of a key triggering method based on the fusion of multiple sets of optical sensing and deep learning provided by an embodiment of the present invention.

[0102] like Figure 9 As shown, it is a light path disconnection diagram of a traditional optical coupling technology of a key triggering method based on the fusion of multiple sets of optical sensing and deep learning provided by an embodiment of the present invention.

[0103] like Figure 10 As shown, it is a light path conduction diagram of a traditional optical coupling technology of a key triggering method based on the fusion of multiple sets of optical sensing and deep learning provided by an embodiment of the present invention.

[0104] like Figure 11 As shown, a layout diagram of multiple sets of side-mounted tubes of a traditional optical coupling technology based on a key triggering method fused with multiple sets of optical sensing and deep learning is provided in one embodiment of the present invention.

[0105] like Figure 12 As shown, a diagram of optical data collected by multiple sets of sensors in a key triggering method based on the fusion of multiple sets of optical sensing and deep learning provided by an embodiment of the present invention.

[0106] like Figure 13 As shown, it is a sensor array distribution diagram of a key triggering method based on the fusion of multiple sets of optical sensing and deep learning provided by an embodiment of the present invention.

[0107] like Figure 14 As shown, it is a model structure design diagram of a key triggering method based on the fusion of multiple sets of optical sensing and deep learning provided by one embodiment of the present invention.

[0108] like Figure 15 As shown, it is a trigger principle flow chart of a key triggering method based on the fusion of multiple sets of optical sensing and deep learning provided by one embodiment of the present invention.

[0109] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0110] Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims rather than the foregoing description, and all changes that come within the meaning and range of equivalents of the claims are intended to be embraced therein. Any reference to a figure in a claim should not be construed as limiting the claim to which it relates.

[0111] Blockchain, as used in this article, refers to a novel application model for computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and encryption algorithms. Blockchain is essentially a decentralized database, a series of data blocks generated using cryptographic methods. Each block contains information about a batch of online transactions, used to verify the validity of this information (to prevent counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, the platform product service layer, and the application service layer.

[0112] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Artificial Intelligence (AI) is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use knowledge to achieve optimal results.

[0113] Furthermore, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices recited in a system claim may also be implemented by a single unit or device through software or hardware. Terms such as "first" and "second" are used to indicate names and do not imply any particular order.

[0114] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A key triggering method based on the fusion of multiple optical sensors and deep learning, characterized in that: The method comprises: Synchronously collecting multi-dimensional optical data of a preset key action using an optical sensor of a key device, wherein the optical sensor is located below a key of the key device; Perform convolution processing on multi-dimensional optical data to obtain optical data features; The optical data features are input into a pre-built key trigger recognition model, and the pre-built key trigger recognition model is used to output the key trigger type.

2. The key triggering method based on the fusion of multiple optical sensors and deep learning as claimed in claim 1, characterized in that: The optical sensor includes multiple groups of optical transmitting-receiving pairs of tubes distributed under multiple keys of the key device, and a preset number of optical transmitting-receiving pairs of tubes are arranged in a matrix distribution under each key.

3. The key triggering method based on the fusion of multiple optical sensors and deep learning as claimed in claim 2, characterized in that: Each optical transmitter-receiver pair includes: Infrared LED, supporting dual wavelengths of 850nm / 940nm, and working with a micro-condenser lens to compress the beam half-angle to 25°, forming a highly collimated optical path; Phototransistor, front-end integrated 850±20nm or 940±20nm bandpass filter; The button is equipped with a handle with a trapezoidal grating. The edge of the grating is designed to have a 45° slope. When the button is pressed, the displacement of the grating along the light path changes linearly with the blocked area.

4. The key triggering method based on the fusion of multiple optical sensors and deep learning as claimed in claim 1, characterized in that: The method of synchronously collecting multi-dimensional optical data of preset key actions using an optical sensor includes: Initialization step: When the key is not pressed, the trapezoidal grating completely blocks the light paths of all optical transmitter-receiver pairs under the key, and the light receiving circuits of all optical transmitter-receiver pairs under the key output a high level; Dynamic acquisition step: When the button is pressed, all optical transmitting-receiving tubes under the button synchronously collect multi-dimensional optical data.

5. The key triggering method based on the fusion of multiple optical sensors and deep learning as claimed in claim 4, characterized in that: All optical transmitter-receiver pairs under the button synchronously collect multi-dimensional optical data, including: collecting light intensity timing curves to record the continuous change process of light intensity in a preset light intensity range; collecting the percentage of blocked area, calculating the real-time blocked area change and the displacement trajectory of the button press through the geometric relationship between the grating displacement and the optical path spacing; collecting wavelength offset: by switching infrared LEDs of different wavelengths, it is used to monitor the impact of ambient light wavelength drift on the response of the optical transmitter-receiver pairs.

6. The key triggering method based on the fusion of multiple optical sensors and deep learning as claimed in claim 1, characterized in that: The key trigger recognition model includes a cascaded input layer, a convolutional layer, a pooling layer, and a fully connected layer; Among them, the data dimension of the input layer is the number of sensors × the number of sampling points; The convolution layer is a stack of 3 layers of 3×1 convolution kernels, which extracts features through a local sliding window; The pooling layer uses the maximum pooling strategy to reduce the dimension of the convolution layer output and retain the peak features in each segment of data; The fully connected layer converts the feature vector into a key action probability value through multi-layer neuron mapping and outputs the key trigger type.

7. A key trigger device based on the fusion of multiple optical sensors and deep learning, characterized in that: Including a controller, and a button device connected to the controller; The controller executes the steps of the key triggering method based on the fusion of multiple sets of optical sensing and deep learning as described in any one of claims 1 to 6.

8. The key triggering device based on the fusion of multiple optical sensors and deep learning as claimed in claim 7, characterized in that: At least two sets of optical transmitting-receiving tubes are arranged under each button of the button device, which are used to collect optical data of the button action from different angles and positions. The optical data includes at least two items of light intensity change, dynamic change of shading area, and wavelength shift.

9. A key triggering device based on the fusion of multiple optical sensors and deep learning as described in claim 7 or 8, characterized in that: The spacing between two adjacent sets of optical transmitting-receiving tube pairs is 1-3 mm, and each set of optical transmitting-receiving tube pairs is equipped with a bandpass filter and a micro-condensing lens, wherein the bandpass filter allows light wavelengths to pass within the range of 850±20 nm or 940±20 nm, and the micro-condensing lens compresses the light beam half angle ≤30°.

10. A key triggering device based on the fusion of multiple optical sensors and deep learning according to claim 7 or 8, characterized in that: The key includes a handle connected to a trapezoidal grating.

Images