Desktop type intelligent digital assistant terminal

Abstract

A desktop type intelligent digital assistant terminal comprises a four-layer integrated framework of an end-side intelligent core, a multi-mode interaction layer, a modular hardware layer and an office ecological cooperation layer, deep cooperation of all layers and bidirectional circulation of data and instructions are achieved, a closed-loop intelligent service system is formed, the end-side autonomous operation capacity is guaranteed, and the service efficiency is improved. Scene flexible adaptation and office ecology deep fusion are considered; meanwhile, high-frequency pain points of an office scene are focused on function design, multiple practical innovation functions are added, core breakthrough is achieved on the algorithm level, and the intelligence and differentiation advantages of the terminal are further enhanced.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary technical field of artificial intelligence, human-computer interaction and office Internet of Things, and specifically relates to a desktop intelligent digital assistant terminal. Background Technology

[0002] With the widespread adoption of artificial intelligence, the Internet of Things, and edge computing technologies in office settings, desktop smart assistant terminals, with their convenient interactive features, are gradually becoming core devices for enterprises to improve office efficiency and optimize collaboration experiences. They are widely used in diverse scenarios such as open-plan offices, meeting rooms, and confidential offices. However, existing desktop smart assistant terminals and related technical solutions still reveal many significant technical pain points in practical applications, failing to fully meet users' core needs for convenient interaction, data security, and scenario adaptability. In terms of interaction methods, most existing terminals adopt a single voice interaction or a simple modal overlay design of "voice + gesture," without considering the dynamic changes in office scenarios. For example, in densely populated open-plan office areas with high ambient noise, single voice commands are easily interfered with, leading to a significant drop in recognition accuracy. In dimly lit environments during late-night overtime work, interaction methods relying on screen visual feedback can easily cause eye fatigue, and traditional wake-up-based voice interaction is cumbersome. In multi-person collaborative meeting room scenarios, terminals struggle to distinguish the commands of different users, easily resulting in erratic responses, a fragmented overall interaction experience, and low efficiency.

[0003] In terms of data processing and privacy protection, current mainstream technical solutions generally rely on cloud computing power to support core functions, such as file retrieval, schedule planning, and document generation, all of which require internet access to access cloud server resources. This results in the terminal completely losing its core service capabilities in an offline environment, making it unsuitable for offline office scenarios. Furthermore, existing solutions often upload sensitive data such as user facial features, office document content, and interaction command records to cloud storage, posing a risk of leakage during data transmission and storage, and failing to meet the stringent data security requirements of confidential office scenarios such as state-owned enterprises, research institutions, and the financial industry. In addition, some offline terminals only integrate fixed functional modules, lacking the ability to learn autonomously and adapt to user habits, and thus unable to provide customized services based on different users' office preferences.

[0004] In terms of hardware architecture and ecosystem collaboration, existing terminals mostly adopt a fixed, integrated design with extremely poor functional scalability. The office needs of users in different positions vary significantly. Administrative staff require identity authentication functions, designers need projection display functions, and those who work long hours require health monitoring functions, etc. Existing integrated terminals cannot flexibly adapt to these personalized needs, requiring users to purchase dedicated equipment, which increases enterprise procurement costs and occupies limited desktop space. At the same time, existing terminals lack sufficient connectivity with office equipment such as computers, printers, and scanners, as well as with enterprise OA systems and office software, lacking a unified collaborative interface. This leads to a break in the closed loop of office workflow—from command initiation to device execution to result output—making it difficult to truly achieve efficient collaborative office processes.

[0005] In summary, existing desktop smart assistant terminals have significant shortcomings in key dimensions such as interactive adaptability, data security, scenario adaptability, and ecosystem collaboration, and can no longer meet the diversified and demanding usage needs in complex office scenarios. There is an urgent need for a new desktop smart digital assistant terminal technology solution that can specifically address the above-mentioned technical pain points. Summary of the Invention

[0006] The core objective of this invention is to completely resolve the technical shortcomings of existing desktop smart assistant terminals in terms of interactive adaptability, data security, scenario adaptability, and ecosystem collaboration, and to provide a desktop smart digital assistant terminal that is both efficient and convenient as well as safe and reliable. This terminal can deeply adapt to the diverse needs of complex office scenarios, becoming an intelligent collaboration partner for users in the office process, rather than a simple functional tool, ultimately achieving a significant improvement in office efficiency and comprehensive protection of data security.

[0007] To achieve the above objectives, this invention innovatively proposes a four-layer integrated architecture consisting of an edge-side intelligent core, a multimodal interaction layer, a modular hardware layer, and an office ecosystem collaboration layer. Each layer collaborates deeply, with bidirectional data and instruction flow, forming a closed-loop intelligent service system. This ensures both the edge's autonomous operation capabilities and flexible adaptation to various scenarios, while also achieving deep integration with the office ecosystem. Furthermore, the functional design focuses on high-frequency pain points in office scenarios, adding several practical and innovative functions, and achieving core breakthroughs at the algorithm level, further enhancing the terminal's intelligence and differentiated advantages.

[0008] The edge-side intelligent core serves as the decision-making and computing hub of the terminal. Its core is the deep integration of a lightweight hybrid intelligent engine, a local privacy sandbox, and intelligent algorithms specifically designed for office scenarios. The lightweight hybrid intelligent engine, based on a pruned and optimized large model specifically for office scenarios, compresses model parameters to within 5GB, which the edge can handle. Combined with a built-in NPU chip, it enables offline inference, completing high-frequency office tasks such as schedule planning, local file retrieval, and simple document generation without relying on cloud computing power, with response latency controlled within 0.5 seconds. Addressing core pain points in office scenarios, the engine innovatively integrates a precise office intent recognition algorithm and a multi-task intelligent scheduling algorithm—the precise office intent recognition algorithm… Based on the improved BERT-BiLSTM hybrid model, the algorithm accurately identifies "implicit office needs" by deeply analyzing the semantics of user commands and combining them with historical office behavior data. For example, when a user issues the command "organize last week's meeting minutes", the algorithm can automatically associate meeting-related documents, key points of participants' speeches, and to-do items to generate a structured integrated document, rather than simply performing a single file organization operation. The multi-task intelligent scheduling algorithm adopts a dynamic priority queue mechanism, which intelligently sorts and allocates resources for multiple concurrent office tasks based on the user's current office status and system resource usage, ensuring that high-priority tasks respond first and avoiding operational conflicts and lag.

[0009] The local privacy sandbox employs AES-256 encryption technology and differential privacy enhancement algorithms to provide dual protection for sensitive user data, covering facial features, voiceprint information, office document content, interaction records, etc. All data is stored on the device side and synchronized across devices only when the user actively authorizes it, through end-to-end encryption technology. The differential privacy enhancement algorithm adds minimal noise to the original data, effectively resisting data inference attacks without affecting the data's usability, further ensuring the security of confidential office data. In addition, the device-side intelligent core also integrates an environmental and user status perception unit, which collects environmental data such as noise intensity, light brightness, and personnel distribution in the office environment in real time, as well as user data such as facial status and voice characteristics, providing decision-making basis for the multimodal interaction layer.

[0010] The multimodal interaction layer integrates four major interaction modules: voiceprint recognition, face recognition, gesture recognition, and lip reading. It innovatively adopts a multimodal feature fusion algorithm and a scene-adaptive decision-making algorithm to achieve intelligent optimization and dynamic switching of interaction modes. The multimodal feature fusion algorithm uses an attention mechanism to weight and fuse feature data collected from each modality, rather than simply adding them together. This effectively improves the accuracy of command recognition in complex environments. For example, in high-noise scenarios, it automatically increases the weight of lip reading and gesture features and reduces the interference of voice features to ensure that the recognition accuracy is not less than 95%. The scene-adaptive decision-making algorithm is based on a reinforcement learning framework. By continuously learning the user's interaction preferences in different scenarios, it dynamically adjusts the switching strategy of interaction modes to achieve a personalized interaction experience.

[0011] Meanwhile, the microphone array in this interaction layer features a 3-meter long-distance sound pickup and adaptive noise suppression algorithm, which can accurately filter environmental noise such as keyboard typing and other people's conversations, and extract clear user voice signals; the camera supports clear face capture in low-light environments and integrates a real-time face posture correction algorithm, so that even if the user turns their face or looks down, they can still accurately complete identity authentication; for multi-person collaboration scenarios, the algorithm can also realize parallel parsing of multi-user commands, and respond to the differentiated needs of different users at the same time, avoiding command conflicts.

[0012] The modular hardware layer adopts an architecture design combining a core base and detachable expansion modules. The core base integrates the main control chip, high-definition touch display, basic sensors, and communication modules, providing general interaction and computing functions. The detachable expansion modules support hot-swappable connections and include three core expansion units: an identity authentication module, a projection module, and a health monitoring module. The identity authentication module integrates an ID card and IC card reader and innovatively incorporates a liveness detection + multi-factor authentication algorithm. Through dual verification of facial liveness detection and card information verification, it prevents identity theft and is suitable for office scenarios requiring strict identity verification, such as government and finance. The projection module has a built-in desktop projection distortion correction algorithm that can automatically adapt to different desktop materials (wood, wood, etc.). The projection effect of the glass desktop ensures a clear and distortion-free projection interface, replacing the screen display, saving desktop space, and facilitating simultaneous viewing and operation by multiple people; the health monitoring module integrates a heart rate sensor and fatigue detection unit, and adopts a "multi-dimensional physiological feature fusion evaluation algorithm" to comprehensively judge the user's fatigue level through multi-dimensional data such as heart rate, eye state, and operation frequency. When it detects that the user has been sitting for more than 1 hour or the fatigue level exceeds the standard, it will issue a reminder through voice or vibration, and link with the edge intelligent core to recommend appropriate rest programs to help users develop healthy office habits; all expansion modules use a unified interface protocol, which can be quickly adapted to the core base without additional driver configuration, realizing flexible expansion of functions.

[0013] The office ecosystem collaboration layer constructs a unified cross-device collaboration interface and software linkage protocol, and innovatively integrates an intelligent workflow connection algorithm and a seamless cross-device data migration algorithm to break down the barriers to collaboration between office equipment, software, and systems. The intelligent workflow connection algorithm can identify users' continuous office behaviors and automatically connect related operations. For example, after a user initiates a command through a terminal, the algorithm can automatically call the Office software on the computer, synchronize the user's commonly used contract templates, and automatically prompt whether to print and send it to the OA system for approval after the document is written, forming a closed-loop process of command initiation - template call - document generation - subsequent execution, without requiring users to manually switch between multiple software and devices. The seamless cross-device data migration algorithm supports real-time synchronization of file and task status between terminals and devices such as computers and mobile phones. For example, a file retrieval task that a user has not completed on the terminal can be directly migrated to the computer to continue the operation. The file modification records, search keywords, and other statuses are completely consistent, achieving seamless switching across devices.

[0014] Meanwhile, this layer deeply integrates mainstream office software and enterprise OA systems, supporting document exchange between the terminal and software such as Office and WPS, as well as functions such as to-do synchronization and workflow initiation with the enterprise OA system; the built-in device status monitoring unit adopts the "device failure prediction algorithm" to predict potential equipment failures in advance by analyzing the operating parameters of office equipment, and promptly reminds users to perform maintenance through the terminal to avoid interruption of office workflow; for printing needs, it also supports the "intelligent print queue optimization algorithm", which automatically sorts documents according to their urgency and the number of pages to be printed, rationally allocates printer resources, and reduces waiting time.

[0015] The beneficial effects of this invention are: 1. This invention achieves autonomous service, personalized learning, and high security in offline mode through an office scenario-specific algorithm and dual privacy protection mechanism of the edge-side intelligent core, while accurately adapting to the dual needs of offline environment and confidential scenario; 2. The fusion algorithm and adaptive decision-making mechanism of the multimodal interaction layer of this invention solve the problems of low interaction accuracy and fragmented experience in complex office environments, and realize personalized and high-precision human-computer interaction; 3. The modular hardware architecture extension modules and supporting algorithms of this invention further enhance the flexibility of scenario adaptation and the practicality of functions; the process connection and cross-device migration algorithms of the office ecosystem collaboration layer break down the barriers of the entire office chain, build an efficient closed-loop office process, and significantly improve office efficiency. Attached Figure Description

[0016] Figure 1 This is a diagram of the four-layer integrated architecture of the desktop intelligent digital assistant terminal of the present invention.

[0017] Figure 2This is a breakdown diagram of the innovative edge-side intelligent core algorithm of this invention.

[0018] Figure 3 This is a diagram illustrating the adaptive switching and algorithm fusion of the multimodal interaction layer in this invention.

[0019] Figure 4 This is a diagram showing the modular hardware layer extension module and algorithm adaptation of the present invention.

[0020] Figure 5 This is a diagram showing the closed-loop process and algorithm mapping of the office ecosystem collaboration layer of this invention.

[0021] Figure 6 This is a diagram of the dual protection architecture of the local privacy sandbox of the present invention.

[0022] Figure 7 This is a schematic diagram of the terminal of the present invention. Detailed Implementation

[0023] The technical solution of the present invention will be further described in detail below through embodiments and in conjunction with the accompanying drawings. Many specific details are set forth in the following description to provide a full understanding of the present invention. However, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar modifications without departing from the spirit of the invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0024] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Preferred embodiments of the invention are shown in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a thorough and complete understanding of the disclosure of the invention.

[0025] Implementation, for example Figures 1-6 As shown, in order to achieve the above objectives, this invention proposes a groundbreaking four-layer integrated architecture: edge-side intelligent core, multimodal interaction layer, modular hardware layer, and office ecosystem collaboration layer. Combining the actual application needs of office scenarios, it provides detailed explanations of the terminal's hardware selection, software deployment, algorithm implementation, and scenario adaptation, ensuring deep collaboration among all layers and feasible functional implementation.

[0026] The terminal hardware is based on a core base and features a heterogeneous multi-core processor and a built-in NPU chip with a computing power of no less than 8 TOPS. It is equipped with 16GB of LPDDR5 high-speed memory and 512GB of UFS 3.1 high-speed storage, providing ample computing and storage resources for the lightweight hybrid intelligent engine of the edge-side intelligent core, ensuring low-latency operation of the model's offline inference. The core base integrates a 10.1-inch 1920×1200 resolution high-definition touchscreen display, supporting multi-touch operation; a 6-microphone circular array enables sound pickup from a distance of 3 meters, combined with an 8-megapixel wide-angle camera, high-fidelity speakers, and a vibration feedback module, forming the basic interactive hardware support; it also incorporates an ambient light sensor, noise sensor, infrared human body sensor, and hardware status monitoring module to collect real-time office environment data and terminal operating status data. The modular hardware layer features detachable expansion modules that connect to the core base via a unified Type-C hot-swappable interface. The identity authentication module integrates an ID card reader chip, an IC card reader module, and a liveness detection camera, achieving secure identity authentication through dual verification of facial liveness detection and card information verification. The projection module uses a DLP projection chip, supporting 1280×720 resolution and projection sizes from 30 to 100 inches, and incorporates a desktop projection distortion correction algorithm to adapt to different desktop materials. The health monitoring module incorporates a heart rate sensor and fatigue detection unit, collecting and comprehensively evaluating user physiological data in real time. For communication, the core base integrates a Wi-Fi 6 and Bluetooth 5.2 dual-mode communication module, with an optional 5G module to meet connectivity needs in various office environments.

[0027] The software and algorithm deployment focuses on autonomous operation and security on the edge side. The lightweight hybrid intelligent engine, the core of the edge intelligence, is based on a large, tailored and optimized model specifically for office scenarios, compressing the model parameters to within 5GB. Relying on the NPU chip, it achieves offline inference, ensuring that the response latency of high-frequency office functions such as schedule planning, local file retrieval, and simple document generation is controlled within 0.5 seconds. The engine's built-in precise office intent recognition algorithm is based on an improved BERT-BiLSTM hybrid model. Through deep semantic parsing and association with users' historical office behavior data, it accurately identifies implicit office needs. For example, when a user commands "organize last week's meeting minutes," it automatically associates relevant documents, key points of participants' speeches, and to-do items to generate a structured and integrated document. The multi-task intelligent scheduling algorithm adopts a dynamic priority queue mechanism, combining the user's office status and system resource usage to intelligently sort and allocate resources for concurrent tasks, avoiding operational conflicts and lag. The local privacy sandbox employs AES-256 encryption technology and differential privacy enhancement algorithms to provide dual protection for sensitive data such as user facial features, voiceprint information, and office document content. All data is stored on the device side and is synchronized across devices only through end-to-end encryption technology when the user actively authorizes it. The differential privacy enhancement algorithm resists inference attacks by adding minimal noise without affecting the usability of the data.

[0028] The multimodal interaction layer integrates four modules: voiceprint recognition, face recognition, gesture recognition, and lip reading. It optimizes interaction through multimodal feature fusion algorithms and scene-adaptive decision-making algorithms. The multimodal feature fusion algorithm uses an attention mechanism to weight and fuse feature data from each modality. In high-noise scenarios, it automatically increases the weight of lip reading and gesture features, reducing speech interference and ensuring a command recognition accuracy of at least 95%. The scene-adaptive decision-making algorithm, based on a reinforcement learning framework, continuously learns user interaction preferences in different scenarios and dynamically adjusts switching strategies to achieve a personalized interactive experience. The microphone array features an adaptive noise suppression algorithm to accurately filter environmental noise; the camera supports low-light face capture and real-time face pose correction algorithms, adapting to identity authentication in scenarios such as side-facing and head-down gestures; for multi-user collaborative scenarios, the algorithm supports parallel parsing of multi-user commands to avoid erratic responses.

[0029] The office ecosystem collaboration layer deeply integrates with mainstream office software such as Office and WPS, as well as enterprise OA systems, through a unified cross-device collaboration interface and software linkage protocol, enabling document interoperability, to-do synchronization, and workflow initiation. The intelligent workflow concatenation algorithm identifies continuous user work behavior and automatically links related operations. The seamless cross-device data migration algorithm supports real-time synchronization of files and task status between terminals and devices such as computers and mobile phones, achieving seamless switching across devices. The built-in device status monitoring unit employs a device fault prediction algorithm to predict potential faults and remind users to maintain the system by analyzing office equipment operating parameters. The intelligent print queue optimization algorithm automatically sorts documents based on their urgency and the number of pages to be printed, rationally allocating printer resources.

[0030] In practical applications, the terminal can be flexibly adapted to different office scenarios. In high-noise environments like open-plan offices, the multimodal interaction layer automatically enhances the weight of lip reading and gesture recognition to ensure accurate command recognition. In confidential office scenarios, a local privacy sandbox protects data security, while offline functionality meets the needs of working without internet access. Administrative staff can add an identity authentication module, designers can pair it with a projection module, and those who work long hours can enable a health monitoring module, with functionality rapidly expandable through hot-swapping. Simultaneously, the terminal connects with various office devices and systems through an office ecosystem collaboration layer, building an efficient closed-loop office workflow. Combined with the personalized service capabilities of the edge-side intelligent core, it comprehensively improves office efficiency and user experience, adapting to the diverse needs of open-plan offices, meeting rooms, and confidential offices.

[0031] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that variations may be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A desktop intelligent digital assistant terminal, characterized in that, This system features a four-layer integrated architecture: edge-side intelligent core, multimodal interaction layer, modular hardware layer, and office ecosystem collaboration layer. Deep collaboration across all layers and bidirectional data and command flow create a closed-loop intelligent service system. The edge-side intelligent core serves as the decision-making and computing hub of the terminal. Its core is the deep integration of a lightweight hybrid intelligent engine, a local privacy sandbox, and intelligent algorithms specific to office scenarios. The lightweight hybrid intelligent engine is based on a large, tailored, and optimized model specifically for office scenarios, compressing the model parameters to within 5GB, which the edge can handle. It works with the built-in NPU chip to achieve offline inference, with response latency controlled within 0.5 seconds. The engine innovatively integrates an accurate office intent recognition algorithm and a multi-task intelligent scheduling algorithm. The accurate office intent recognition algorithm is based on an improved BERT-BiLSTM hybrid model, which accurately identifies office needs through deep semantic analysis of user commands and combined with the user's historical office behavior data. The multi-task intelligent scheduling algorithm adopts a dynamic priority queue mechanism, which intelligently sorts and allocates resources for multiple concurrent office tasks based on the user's current office status and system resource usage, ensuring that high-priority tasks respond first and avoiding operational conflicts and lag.

2. The desktop intelligent digital assistant terminal according to claim 1, characterized in that, The local privacy sandbox uses AES-256 encryption technology and differential privacy enhancement algorithm to dual protect sensitive user data, covering facial features, voiceprint information, office document content, and interaction records. All data is stored on the device side and is synchronized across devices only when the user actively authorizes it through end-to-end encryption technology. Differential privacy enhancement algorithms effectively resist data inference attacks by adding minimal noise to the original data without affecting the data's usability, further ensuring the security of confidential office data. In addition, the edge-side intelligent core also integrates an environment and user status perception unit, which collects environmental data such as noise intensity, light brightness, and personnel distribution in the office environment in real time, as well as user data such as facial status and voice features, providing decision-making basis for the multimodal interaction layer.

3. A desktop intelligent digital assistant terminal according to claim 1, characterized in that, The multimodal interaction layer integrates four major interaction modules: voiceprint recognition, face recognition, gesture recognition, and lip reading. It innovatively adopts a multimodal feature fusion algorithm and a scene-adaptive decision algorithm to achieve intelligent optimization and dynamic switching of interaction modes. The multimodal feature fusion algorithm uses an attention mechanism to weight and fuse feature data collected from each modality, rather than simply adding them together, effectively improving the accuracy of command recognition in complex environments. In high-noise scenarios, it automatically increases the weight of lip reading and gesture features, reduces interference from voice features, and ensures that the recognition accuracy is not less than 95%. The scene-adaptive decision algorithm is based on a reinforcement learning framework. By continuously learning the user's interaction preferences in different scenarios, it dynamically adjusts the switching strategy of interaction modes to achieve a personalized interaction experience. The microphone array in this interaction layer features a 3-meter long-distance sound pickup and adaptive noise suppression algorithm, which can accurately filter environmental noise such as keyboard typing and other people's conversations, and extract clear user voice signals; the camera supports clear face capture in low-light environments and integrates a real-time face pose correction algorithm to accurately complete identity authentication; for multi-user collaboration scenarios, the algorithm can also realize parallel parsing of multi-user commands, while responding to the differentiated needs of different users and avoiding command conflicts.

4. A desktop intelligent digital assistant terminal according to claim 1, characterized in that, The modular hardware layer adopts an architecture design that combines a core base and detachable expansion modules. The core base integrates a main control chip, a high-definition touch screen, basic sensors and communication modules, providing general interaction and computing functions. The detachable expansion module supports hot-swappable connection and includes three core expansion units: an identity authentication module, a projection module, and a health monitoring module. The identity authentication module integrates an ID card and IC card reader area and innovatively incorporates liveness detection and multi-factor authentication algorithms to perform dual verification of face liveness detection and card information verification. The projection module has a built-in desktop projection distortion correction algorithm that automatically adapts the projection effect to different desktop materials.

5. A desktop intelligent digital assistant terminal according to claim 4, characterized in that, The health monitoring module integrates a heart rate sensor and a fatigue detection unit, and adopts a "multi-dimensional physiological feature fusion evaluation algorithm" to comprehensively judge the user's fatigue level by using multi-dimensional data such as heart rate, eye status, and operation frequency. When the user's fatigue level is detected to be excessive, a reminder is issued through voice or vibration, and the intelligent core on the edge is linked to recommend an appropriate rest plan to help users develop healthy office habits. All expansion modules adopt a unified interface protocol, which can be quickly adapted to the core base without additional driver configuration, and realize flexible expansion of functions.

6. A desktop intelligent digital assistant terminal according to claim 1, characterized in that, The office ecosystem collaboration layer constructs a unified cross-device collaboration interface and software linkage protocol, and innovatively integrates an intelligent workflow connection algorithm and a seamless cross-device data migration algorithm to break down the barriers to collaboration between office equipment, software, and systems. The intelligent workflow connection algorithm can identify users' continuous office behaviors and automatically connect related operations. After a user initiates a command through the terminal, the algorithm can automatically call the Office software on the computer, synchronize the user's commonly used contract templates, and automatically prompt whether to print and send it to the OA system for approval after the document is written, forming a closed-loop process of command initiation - template call - document generation - subsequent execution, without requiring users to manually switch between multiple software and devices. The seamless cross-device data migration algorithm supports real-time synchronization of file and task status between the terminal and devices such as computers and mobile phones. For example, a file retrieval task that a user has not completed on the terminal can be directly migrated to the computer to continue the operation. The file modification records, search keywords, and other statuses are completely consistent, realizing seamless switching across devices. This layer deeply integrates mainstream office software and enterprise OA systems, supporting document exchange between the terminal and software such as Office and WPS, as well as functions such as to-do synchronization and workflow initiation with the enterprise OA system; the built-in device status monitoring unit adopts the "device failure prediction algorithm" to predict potential equipment failures in advance by analyzing the operating parameters of office equipment, and promptly reminds users to perform maintenance through the terminal to avoid interruption of office processes; For printing needs, it also supports an "intelligent print queue optimization algorithm" that automatically sorts documents based on their urgency and the number of pages to be printed, and allocates printer resources accordingly.

7. A desktop intelligent digital assistant terminal according to claim 4, characterized in that, The core base is equipped with a heterogeneous multi-core processor and a built-in NPU chip with a computing power of no less than 8 TOPS, along with 16GB of LPDDR5 high-speed memory and 512GB of UFS 3.1 high-speed storage. This provides ample computing and storage resources for the lightweight hybrid intelligent engine of the edge-side intelligent core, ensuring low-latency operation of the model's offline inference. The core base integrates a 10.1-inch 1920×1200 resolution high-definition touch screen, supporting multi-touch operation; a 6-microphone circular array enables sound pickup from a distance of 3 meters, combined with an 8-megapixel wide-angle camera, high-fidelity speakers, and a vibration feedback module, forming the basic interactive hardware support; it also incorporates an ambient light sensor, a noise sensor, an infrared human body sensor, and a hardware status monitoring module to collect real-time office environment data and terminal operating status data.

Images