A route guidance electronic station board based on a large language model
By integrating a sound pickup module, a voice playback module, and a local AI large voice model unit into a traditional electronic bus stop sign, natural language and dialect interaction are achieved, solving the problem of insufficient intelligence in existing electronic bus stop signs, improving the environmental adaptability and ease of use of the equipment, and enhancing the inclusiveness and convenience of public transportation information services.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- JIANGXI YUNBEN DIGITAL TECH CO LTD
- Filing Date
- 2026-04-29
- Publication Date
- 2026-07-14
AI Technical Summary
Existing electronic bus stop signs cannot understand natural language or support dialect interaction. They rely on physical touch operation, have limited functionality, and are difficult to upgrade to intelligent systems at low cost, resulting in inconvenience and insufficient service.
A modular transformation scheme is adopted, integrating a sound pickup module, a voice playback module, and a local AI large voice model unit into the traditional electronic bus stop sign. This enables voice interaction in Mandarin and dialects, combines edge computing for semantic understanding and route planning, supports touchless operation, and automatically switches between silent and voice interaction modes.
It achieves low-cost, non-destructive improvement of the intelligence level of electronic bus stop signs, supports natural language and dialect interaction, reduces maintenance costs and power consumption, improves the environmental adaptability and ease of use of the equipment, and enhances the inclusiveness and convenience of public transportation information services.
Smart Images
Figure CN122392342A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of electronic bus stop technology, specifically to an electronic bus stop for directions based on a large language model. Background Technology
[0002] Currently, urban public transportation, as a typical public service, generally faces the predicament of long-term losses, with its operators heavily reliant on local government subsidies for daily maintenance and equipment investment. Under this rigid budget constraint, public transportation companies find it difficult to afford large-scale, high-standard replacement of equipment with entirely new intelligent systems. Therefore, utilizing and upgrading existing equipment to achieve a leap in service capacity at minimal cost has become a realistic choice and an urgent need for the industry's development.
[0003] Currently, cities across China have deployed a large number of traditional electronic bus stop signs, forming the core terminal network for public transportation information dissemination. However, due to early technological architecture and cost considerations, these devices have significant technical shortcomings and applicability deficiencies in terms of functional design, human-computer interaction, maintenance economy, and level of intelligence. Specifically: Existing bus stop sign interactions are unnatural and unable to understand complex travel needs. The human-computer interaction of existing devices heavily relies on pre-defined menu structures and hierarchical selection operations. Their underlying logic is command-based, lacking the ability to understand natural language and thus unable to handle open-ended, ambiguous, or multi-constrained questions and transfer requests such as "How do I transfer to get to the train station?", "I want to go to the nearest subway station?", or "How long until the next express train to the city center?". This essentially stems from their lack of semantic understanding and intent reasoning capabilities, resulting in an extremely low level of intelligence.
[0004] Secondly, current bus stop signs rely heavily on physical contact and have poor adaptability to different scenarios. Operation methods generally use touchscreens or physical buttons, requiring users to approach and perform precise touch or press actions. This presents significant physical interaction barriers in scenarios such as rainy days (when the screen is wet), carrying large luggage, wearing gloves in winter, and for the elderly, disabled, or other people with mobility impairments. The operation is cumbersome and has a high learning and usage threshold.
[0005] Existing electronic bus stop signs offer limited language support, creating a significant "digital divide." The vast majority of traditional devices only support standard Mandarin commands or text input, lacking dialect recognition and comprehension capabilities. For many elderly passengers, local residents, and some migrant workers who primarily communicate in their local dialects, these devices effectively constitute a service barrier, failing to achieve the goal of universal and equitable public services and contradicting the public welfare nature of public transportation.
[0006] Furthermore, current electronic bus stop signs suffer from limited functionality and a lack of service expansion capabilities. Existing equipment is limited to displaying arrival information in a preset format, failing to provide one-stop travel guidance services such as dynamic route planning, multi-modal transfer guidance, and nearby points of interest search. Neither its software nor hardware architecture supports hot-scaling or remote iteration, resulting in extremely limited optimization and improvement of the public transportation information service experience, making it difficult to meet the requirements of modern cities for convenient, refined, and intelligent infrastructure.
[0007] In conclusion, the public transportation industry urgently needs a solution that can closely adapt to its current financial situation, avoid high investment in new construction, and enable rapid, low-cost, and non-destructive upgrades to a large amount of existing traditional equipment.
[0008] In order to fundamentally solve the above-mentioned technical challenges and achieve a universal leap in public transportation information service capabilities, the solution must have the ability to interact with natural language and dialects, low physical contact dependence, high environmental reliability, low maintenance and power consumption, and strong replicability and scalability. Summary of the Invention
[0009] To address the aforementioned technical issues, this invention provides an electronic bus stop sign based on a large language model. This sign is designed for low-cost modular upgrades and transformations of traditional electronic bus stop signs, and features multifunctional intelligent capabilities, including providing directions, queries, and transfer guidance in both dialects and standard Mandarin using an AI large voice model.
[0010] To achieve the above objectives, the present invention provides the following technical solution: an electronic bus stop sign for directions based on a large language model, comprising: The display unit is used to display regular bus information in the absence of voice interaction, and to simultaneously display the voice directions guidance interface and voice query results in the interactive state. The microphone module is used to collect the user's voice input; The voice playback module is used to broadcast the directions and guidance results in voice format; The local AI large speech model unit has a built-in trained dialect-Mandarin mixed speech recognition model and natural language understanding model, which is used to perform speech recognition, semantic understanding, intent parsing and route planning on the speech input collected by the sound pickup module, and generate speech broadcast content. The edge computing unit is used to run the local AI large speech model unit and coordinate data communication and task scheduling between modules; The real-time bus data interface is used to obtain real-time bus arrival data, route data, and operation announcement data. The positioning module is used to obtain the current geographical location information of the bus stop sign; The power management unit is used to supply power to the above-mentioned units; The aforementioned electronic bus stop sign for directions primarily utilizes the modular transformation and upgrading of existing traditional electronic bus stop signs. By adding a sound pickup module, a voice playback module, and an edge computing unit integrating the local AI large voice model unit inside the original bus stop sign, and then connecting to the original bus stop sign's display unit, power management unit, and data transmission unit via standardized interfaces, a non-destructive and functionally unaffected intelligent upgrade is achieved. The electronic wayfinding sign is configured to automatically and seamlessly switch between the following two modes: Silent mode without voice: When the sound pickup module does not detect valid voice input, the display unit maintains all the conventional information display functions of a traditional electronic bus stop sign, and the fixed area of the display unit continuously displays the voice directions guidance interface to prompt the user to inquire by voice; AI voice interaction mode: When the sound pickup module detects valid voice input, the device automatically wakes up the local AI large voice model unit to recognize and understand the open-ended directions input by the user in Mandarin, local dialect, or a mixture of both. Combining the real-time data obtained from the bus real-time data interface with the location information obtained from the positioning module, the device generates the optimal travel guidance plan and outputs it synchronously through the voice playback module and the display unit. After the AI voice interaction is completed, the device automatically returns to the no-voice silent mode after a delay.
[0011] As a preferred technical solution for a wayfinding electronic bus stop sign based on a large language model according to the present invention, the modular transformation and upgrading mode of the existing traditional electronic bus stop sign is as follows: the original bus stop sign's box, display screen, power system, communication module, installation foundation and external lines are retained unchanged, and the original main structure is not replaced or damaged; only a sound pickup module, the voice playback module and the edge computing unit integrating the local AI large voice model unit are added inside the original bus stop sign; and data communication, display linkage and command response are realized with the original bus stop sign's control system through at least one of a general serial port, network interface or IO interface.
[0012] As a preferred technical solution for electronic bus stop signage based on a large language model according to the present invention, in the silent mode without voice, the conventional information displayed by the display unit includes at least: bus route number and direction, real-time distance and estimated time of vehicle arrival, current station name and adjacent station names, first and last bus times and fare information, system time and date, weather information, and operation announcements and emergency information.
[0013] As a preferred technical solution for a wayfinding electronic bus stop sign based on a large language model according to the present invention, the voice wayfinding guidance interface is fixedly displayed in the preset area of the display unit in the form of text or icons, and is used to prompt users to initiate voice wayfinding, transfer inquiry and destination guidance operations in Mandarin or local dialect; the guidance interface includes at least one of the following guidance prompts: "Please say your destination directly", "Voice wayfinding does not require touching", "Supports dialect inquiry".
[0014] As a preferred technical solution for electronic bus stop signs for directions based on a large language model according to the present invention, the local AI large speech model unit has a dialect-Mandarin mixed speech recognition model built in it, which supports end-to-end recognition and understanding of local dialects, Mandarin, and mixed spoken language, and can automatically identify at least one of the following user intent types: destination inquiry, vehicle arrival query, transfer plan planning, station information query, and first and last bus time query.
[0015] As a preferred technical solution for electronic bus stop signs for directions based on a large language model, the AI voice interaction mode completely eliminates the dependence on touch screens and physical buttons. Users can complete the query and guidance operations entirely through voice input. The device does not have a touch interface or physical button interface for query operations.
[0016] As a preferred technical solution for the electronic bus stop sign based on a large language model, the electronic bus stop sign eliminates the reliance on touch screen query operations. The display unit of the electronic bus stop sign can adopt a non-touch outdoor high-reliability display screen, thereby reducing the overall power consumption, reducing the maintenance workload and replacement cost caused by touch screen failure, and making the modified equipment have low maintenance and low power consumption.
[0017] As a preferred technical solution for electronic bus stop signs for directions based on a large language model according to the present invention, the local AI large voice model unit is deployed inside the edge computing unit, and completes local inference operations for speech recognition, semantic understanding and route planning in an offline state, without relying on cloud servers for real-time interaction.
[0018] As a preferred technical solution for electronic bus stop signs for directions based on a large language model, the sound pickup module adopts a multi-microphone array structure, has a far-field sound pickup capability of 3-5 meters, and has built-in environmental noise suppression, echo cancellation and wind noise filtering algorithms to adapt to effective voice collection in noisy and windy outdoor environments.
[0019] As a preferred technical solution for electronic bus stop signs based on a large language model according to the present invention, the standardized interface includes a data communication interface and a power interface. The data communication interface adopts at least one of a general serial port, a network interface or an IO interface, and is used to interface with the original system of existing traditional electronic bus stop signs of different manufacturers, years and models, and has the ability to be replicated and promoted across models and batches.
[0020] Compared with the prior art, the beneficial effects of the present invention are: 1. This technical solution addresses the problems of "closed architecture and lack of intelligent upgrade interfaces" and "huge investment and serious waste in overall replacement" pointed out in the background technology. This invention adopts a modular installation scheme: while retaining the original enclosure, display screen, power supply, data transmission unit, and wiring layout, only an AI large language model module, a multi-microphone pickup module, and a voice playback module are added through standardized interfaces. This modification does not damage the original structure, does not affect the original functions, and does not interrupt daily operations. It is fast to install, highly compatible, and can be replicated in batches. Therefore, it completely avoids the huge financial expenditure and resource waste caused by traditional "demolition and reconstruction" upgrades, enabling public transportation companies to achieve intelligent utilization of massive amounts of existing equipment with minimal investment, significantly extending the equipment life cycle, and aligning with the public service nature and low-budget operational realities of the public transportation industry.
[0021] 2. This solution addresses the interactive shortcomings of existing devices, such as "inability to understand natural language" and "lack of dialect recognition." This invention integrates an AI-powered large-scale speech model unit, supporting direct input of Mandarin, local dialects, and a mixture of both. Users do not need to learn any menu structures or operation steps; they can complete complex intent expressions such as open-ended directions (e.g., "How do I get to the train station?"), fuzzy destination queries (e.g., "The nearest subway station"), and multi-condition transfer inquiries (e.g., "How long until the next express train to the city center?") using only natural spoken language. Combining the geographical location of bus stops with real-time bus data, the system automatically completes speech recognition, semantic understanding, intent parsing, route planning, and speech synthesis, and outputs the information synchronously with the on-screen text. This technological advantage allows elderly passengers unfamiliar with Mandarin, local residents, and rural residents migrating to the city to use the service without barriers, truly realizing the universalization, age-friendliness, and localization of public transportation services.
[0022] 3. This technical solution completely eliminates the reliance on touch screens and physical buttons, greatly improving environmental adaptability and equipment reliability. In response to the problems pointed out in the background technology, such as "touch screens are easily damaged outdoors, have a high failure rate and high maintenance costs" and "difficult to use in rainy weather, when wearing gloves, or when mobility is limited", this invention adopts a fully touchless voice interaction mode. Users can complete all queries and guidance operations simply by issuing voice commands, without having to approach or touch any physical interface. This not only fundamentally eliminates common fault points such as touchscreen blackouts, malfunctions, and damage, significantly reducing the workload and replacement costs of later maintenance, but also greatly improves the user experience in special scenarios such as rainy days, carrying luggage, wearing gloves, and physical disabilities, thereby enhancing the accessibility of public transportation services.
[0023] 4. Addressing the limitations of existing equipment, such as "limited functionality and poor expandability," this invention designs a silent mode without voice input and an AI voice interaction mode, supporting automatic, delayed, and seamless switching. In the absence of voice input, the device retains all the basic functions of a traditional electronic bus stop sign (route information, real-time arrival times, first and last bus times, weather announcements, etc.), without altering the original information dissemination model and operational logic, ensuring zero disruption to daily operations during the upgrade process. Simultaneously, a fixed area in the display interface adds voice directions and guidance prompts, proactively reducing the learning curve for users. Upon detecting valid voice input, the device automatically enters AI interaction mode and automatically returns to silent mode after the interaction is complete. This dual-mode design ensures both complete functional continuity and the addition of intelligent service capabilities, achieving an upgrade effect of "no loss of original functions and easy use of new functions."
[0024] 5. This invention not only supports static route queries but also dynamically generates optimal transfer plans, walking guidance suggestions, and surrounding landmark information by combining real-time bus data, station location, and user intent. Through voice and screen collaborative output, users can obtain complete, clear, and timely travel guidance, solving the problem of traditional devices' limited functionality—"only showing timetables, not providing route information." This greatly improves the convenience, refinement, and intelligence of modern urban public transportation services, enhancing public willingness and satisfaction to choose public transportation. Attached Figure Description
[0025] To make the content of this invention easier to understand, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings.
[0026] Figure 1 This is a block diagram of the overall system structure of the present invention; Detailed Implementation The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0027] Example 1 See attached document Figure 1The modular transformation and upgrade of existing traditional electronic bus stop signs are as follows.
[0028] This embodiment uses a traditional electronic bus stop sign already in operation in a certain city as an example to describe in detail the specific implementation of the present invention. The original cabinet, 7-inch LCD display screen, power system, 4G communication module, basic installation bracket and external power supply line of the traditional electronic bus stop sign are all intact, but it only has fixed menu-style query and arrival information display functions, and does not have voice interaction capabilities.
[0029] First, preparatory work is carried out before the renovation. The target traditional electronic bus stop sign is inspected on-site to confirm whether its internal physical space meets the installation requirements of the new module, whether its power system can provide additional power redundancy, and whether its original control system has available data communication interfaces (such as Universal Asynchronous Receiver / Transmitter Interface (UART), RS232 / RS485 serial port, Ethernet interface, or general input / output IO interface).
[0030] Upon inspection, the traditional electronic bus stop sign in this embodiment has approximately 150mm×100mm×50mm of free installation space inside. Its power system can provide an additional 12V / 2A power supply capacity, and its main control board has a reserved RS232 serial communication interface.
[0031] The second step is to prepare the following new modules: AI Large Language Model Module: It adopts an integrated edge computing module, with a built-in pre-trained dialect-Mandarin mixed speech recognition model and natural language understanding model, with a computing power of 1.0 TOPS, and supports offline local inference.
[0032] Multi-microphone pickup module: It adopts a 4-microphone array structure, which has a far-field pickup capability of 3-5 meters. It has built-in environmental noise suppression, echo cancellation and wind noise filtering algorithms, making it suitable for noisy and windy outdoor environments.
[0033] Voice playback module: It adopts a 3W moisture-proof speaker unit and a matching audio power amplifier circuit.
[0034] While maintaining the integrity of the original bus stop sign box structure, proceed with the installation and connection according to the following steps: ① Open the back cover of the bus stop box, and without removing any existing components, fix the AI large language model module, multi-microphone pickup module and voice playback module to the empty area inside the box.
[0035] ② Point the microphone array opening of the multi-microphone pickup module toward the passenger standing area in front of the bus stop sign to ensure the correct sound pickup direction; point the speaker outlet of the voice playback module toward the front or lower side of the bus stop sign to ensure effective sound propagation.
[0036] ③ Connect the AI large language model module to the main control board of the existing bus stop sign via an RS232 serial cable to achieve display linkage, data exchange, and command response. At the same time, draw power from the bus stop sign's power system to provide 12V DC power to the new module.
[0037] ④ Establish a network connection between the AI large language model module and the real-time bus data interface (reusing the network channel of the original 4G communication module) to obtain vehicle GPS information, arrival data, route information, station information, and operation scheduling information. At the same time, the module integrates a positioning module to obtain the geographical location information of the current bus stop sign.
[0038] ⑤ After completing the hardware installation, configure the software: Set the fixed area of the display unit as the voice navigation guidance interface, and display guidance prompts such as "Please just say your destination", "Voice navigation without touching", and "Supports dialect search" in the interface.
[0039] After the modification is completed, the device will power on and start. After system initialization, it will enter silent mode by default.
[0040] In this mode, the device operates like a normal electronic bus stop sign, with the display unit showing the following standard information in real time: All bus route numbers and routes that pass through this station; Real-time arrival distance and estimated time for the next train on each route; Current site name and names of adjacent sites; First and last bus times and ticket prices; System time, date, and weather information; Operational announcements, public service announcements, emergency information, etc.
[0041] Meanwhile, the fixed area in the lower right corner of the display unit continuously displays the aforementioned voice-guided directions interface to inform passing passengers that the bus stop now has voice interaction capabilities. In this mode, the microphone module remains in low-power standby listening mode, while the voice playback module remains in standby and silent mode.
[0042] When a passenger arrives at the bus stop and sees the directional signs, they can ask for directions directly in natural language without touching or pressing any buttons. For example (in Mandarin): "How do I get to the train station?" or (in the local dialect): "Where do I get to the train station?" or (in a mix of spoken and natural language): "I want to go to the train station, how long until the next train?" The sound pickup module uses a microphone array to collect ambient sound in real time. When valid voice input is detected (voice energy exceeds a preset threshold and duration is greater than 200ms), the system is automatically woken up from the silent mode and enters the AI voice interaction mode.
[0043] The voice pickup module performs front-end processing on the acquired voice signals: beamforming algorithm enhances the voice signal from a range of 3-5 meters in front of the bus stop sign, while suppressing ambient noise from the sides and rear; echo cancellation algorithm eliminates echo interference that may be generated by the voice playback module itself; and wind noise filtering algorithm reduces the impact of outdoor wind noise. After the above processing, the clean voice stream is transmitted to the local AI large voice model unit in the edge computing unit.
[0044] In this technical solution, after the local AI large speech model unit receives the speech stream, it performs the following processing: ① Speech Recognition: The built-in dialect-Mandarin hybrid speech recognition model is used for end-to-end speech recognition. This model has been fine-tuned and trained with local dialect corpora, and can accurately recognize spoken expressions input by users in Mandarin, local dialect, or a mixture of both. For example, for the dialect input "how to get to the train station", the model converts it into the standard semantic expression "the route to the train station".
[0045] ② Semantic understanding and intent parsing: Natural language understanding is performed on the identified text to automatically identify the user's intent type. In this embodiment, the system can identify at least one of the following intent types: Ask for directions at your destination (e.g., "How do I get to the train station?"); Vehicle arrival time query (e.g., "How long until the next train"); Transfer planning (e.g., "How to transfer from here to the city center hospital"); Station information query (e.g., "Does this station have a bus to the stadium?"); Check first and last bus times (e.g., "What time is the last bus?").
[0046] ③ Entity extraction: Extract key entities from user statements, such as destination name ("train station"), time constraints ("next train"), preference conditions ("fastest", "fewest transfers"), etc.
[0047] Based on the parsed user intent, the edge computing unit sends a request to the bus dispatch center server through the real-time bus data interface to obtain the following data: Real-time GPS location and estimated arrival time of all vehicles on all routes at the current station; Information on the route to the target station (such as a train station); Operational scheduling information (such as temporary route changes, service suspension notices, etc.).
[0048] At the same time, the edge computing unit reads the precise latitude and longitude coordinates of the current bus stop obtained by the positioning module.
[0049] Based on the above data, the edge computing unit calls the built-in path planning algorithm to generate the optimal travel guidance plan. For example, for the query "How to get to the train station", the system may generate the following plan: "You can take bus K1 from this station, which will take about 12 minutes after 5 stops; or you can take bus 3 to the Municipal Cultural Palace Station and transfer to Metro Line 2, which will take about 18 minutes in total." The system will output the generated results synchronously in two ways: Voice Broadcast: The voice playback module broadcasts the results using natural and fluent synthesized speech. The broadcast content adopts a human-like tone, such as, "To get to the train station, you can take bus K1 from this station, which takes about 12 minutes. Do you need me to tell you which stop to get off at?" Screen display: The display unit synchronously displays text information, route numbers, station lists, and transfer step diagrams. For example, the K1 route is highlighted on the screen, and the station sequence and estimated time for "current station → train station" are marked.
[0050] If the user is not satisfied with the result or needs further information, they can continue to make a voice call, such as: "How long until the next K1 bus?" The system will repeat the above steps for multiple rounds of dialogue.
[0051] If the system does not detect any new valid voice input within a preset delay time (e.g., 15 seconds), it will automatically exit the AI voice interaction mode and return to the silent mode. The display unit will resume normal information display, the voice playback module will go into standby mode, and the sound pickup module will resume low-power monitoring mode.
[0052] Throughout the entire interaction process, users do not need to touch the screen, press any physical buttons, or approach the bus stop to perform any manual operations, achieving a truly touchless voice interaction.
[0053] Since the reliance on touchscreen queries is eliminated, the display unit in this embodiment can use a non-touchscreen outdoor high-reliability display (such as a high-brightness industrial-grade LCD). This type of screen does not contain a touch-sensing layer, has a much lower failure rate than touchscreens, reduces overall power consumption by approximately 30%, and eliminates the need for regular touch calibration maintenance. The modified equipment possesses high stability and low maintenance, significantly reducing the subsequent operating costs and maintenance burden for public transportation companies.
[0054] Example 2 See attached document Figure 1 The integrated deployment mode of the new intelligent electronic bus stop sign is as follows.
[0055] This embodiment is applicable to newly built bus stops, without the need to modify existing equipment, and instead adopts an integrated whole-machine solution for direct deployment.
[0056] In this mode, the bus stop sign integrates all the functional modules described in Example 1 when it leaves the factory, including a display unit, a multi-microphone pickup module, a voice playback module, a local AI large voice model unit, an edge computing unit, a real-time bus data interface, a positioning module, and a power management unit. The connection and communication between the modules have been pre-tested in the factory.
[0057] After the device is powered on, its workflow is completely consistent with the dual-mode automatic switching logic in Example 1: the default is a silent mode without voice, which displays regular information and a guidance interface; after voice is detected, it enters the AI voice interaction mode to complete recognition, understanding, planning, broadcasting and display; after the interaction is completed, it automatically returns to the silent mode.
[0058] The difference from Example 1 is that the integrated deployment mode does not need to consider compatibility with the original bus stop control system, the interface design is more flexible, and the cabinet structure can be acoustically optimized for voice acquisition and broadcasting, such as directional layout of microphone array and speaker positions, to further improve far-field sound pickup and broadcasting clarity.
[0059] The following is a supplementary explanation of the preferred implementation method: Regarding the optimization of the microphone module: In outdoor environments with strong wind noise, the system can automatically activate the wind noise detection algorithm. When the wind speed exceeds the set threshold, the beamforming parameters of the microphone array are dynamically adjusted to enhance the frontal voice signal and attenuate the lateral wind noise, ensuring that the voice recognition accuracy does not decrease significantly.
[0060] Regarding local AI model updates: The local AI large speech model unit supports remote model updates via the real-time bus data interface. When bus companies need to support new dialect regions or optimize recognition accuracy, the central server can push incremental training models, and the device will automatically complete the model replacement during low-power standby periods without on-site manual operation.
[0061] Regarding power management: The power management unit can dynamically adjust power consumption based on the current time and usage scenario. For example, during low-traffic periods from late night to early morning, the wake-up sensitivity of the microphone module can be lowered, or the output power of the voice playback module can be reduced to further save energy.
[0062] Regarding the adaptation to multiple output languages: In tourist cities or areas with a large number of foreigners, the local AI large speech model unit can be additionally loaded with speech recognition and synthesis models in English or other languages to achieve multilingual interaction support and expand the service range of the device.
[0063] Using a specific embodiment of the present invention, 100 traditional electronic bus stop signs in a certain city were upgraded in batches. Statistical data after 6 months of operation following the upgrade shows: The mean time between failures (MTBF) was improved by approximately 2.5 times compared to the original touchscreen solution. The number of maintenance work orders due to touchscreen malfunctions has dropped to 0 (because there are no touchscreens left). The average number of voice queries per day reaches 120 per device, of which about 35% are in dialects; User satisfaction surveys show that the willingness of elderly passengers and those unfamiliar with Mandarin to use the service has increased by more than 80%. The average power consumption of the entire machine was reduced from 15W before the modification to 9W, a reduction of 40%.
[0064] The above data fully demonstrates the significant progress made by the technical solution proposed in this invention in reducing maintenance costs, improving service accessibility, and achieving low-power operation.
[0065] The specific implementation method provided by this invention has a clear structure, high modularity, and standardized interfaces, which can be quickly adapted to traditional electronic bus stop signs of different manufacturers, years, and models. It does not require changes to the existing operational logic and information dissemination processes of public transportation companies, has a short construction cycle (the average time for a single unit modification is no more than 2 hours), and can be replicated and promoted in batches, possessing extremely high industrial application value.
[0066] The above description is merely a preferred embodiment of the present invention and is not intended to further limit the present invention. All equivalent changes made based on the description and drawings of the present invention are within the protection scope of the present invention.
Claims
1. An electronic bus stop sign for directions based on a large language model, comprising: The display unit is used to display regular bus information in the absence of voice interaction, and to simultaneously display the voice directions guidance interface and voice query results in the interactive state. The microphone module is used to collect the user's voice input; The voice playback module is used to broadcast the directions and guidance results in voice format; The local AI large speech model unit has a built-in trained dialect-Mandarin mixed speech recognition model and natural language understanding model, which is used to perform speech recognition, semantic understanding, intent parsing and route planning on the speech input collected by the sound pickup module, and generate speech broadcast content. The edge computing unit is used to run the local AI large speech model unit and coordinate data communication and task scheduling between modules; The real-time bus data interface is used to obtain real-time bus arrival data, route data, and operation announcement data. The positioning module is used to obtain the current geographical location information of the bus stop sign; The power management unit is used to supply power to the above-mentioned units; Its features are, The aforementioned electronic bus stop sign for directions primarily utilizes the modular transformation and upgrading of existing traditional electronic bus stop signs. By adding a sound pickup module, a voice playback module, and an edge computing unit integrating the local AI large voice model unit inside the original bus stop sign, and then connecting to the original bus stop sign's display unit, power management unit, and data transmission unit via standardized interfaces, a non-destructive and functionally unaffected intelligent upgrade is achieved. The electronic wayfinding sign is configured to automatically and seamlessly switch between the following two modes: Silent mode without voice: When the sound pickup module does not detect valid voice input, the display unit maintains all the conventional information display functions of a traditional electronic bus stop sign, and the fixed area of the display unit continuously displays the voice directions guidance interface to prompt the user to inquire by voice; AI voice interaction mode: When the sound pickup module detects valid voice input, the device automatically wakes up the local AI large voice model unit to recognize and understand the open-ended directions input by the user in Mandarin, local dialect, or a mixture of both. Combining the real-time data obtained from the bus real-time data interface with the location information obtained from the positioning module, the device generates the optimal travel guidance plan and outputs it synchronously through the voice playback module and the display unit. After the AI voice interaction is completed, the device automatically returns to the no-voice silent mode after a delay.
2. The electronic bus stop sign for directions based on a large language model according to claim 1, characterized in that: The modular transformation and upgrade mode of the existing traditional electronic bus stop signs is as follows: the original bus stop sign's box, display screen, power system, communication module, installation foundation and external lines are retained unchanged, and the original main structure is not replaced or damaged; only the sound pickup module, the voice playback module and the edge computing unit integrating the local AI large voice model unit are added inside the original bus stop sign; and data communication, display linkage and command response are realized with the original bus stop sign's control system through at least one of the general serial port, network interface or IO interface.
3. The electronic bus stop sign for directions based on a large language model according to claim 1, characterized in that: In the silent mode without voice, the regular information displayed by the display unit includes at least: bus route number and direction, real-time distance to the station and estimated time, current station name and adjacent station names, first and last bus times and fare information, system time and date, weather information, and operation announcements and emergency information.
4. The electronic bus stop sign for directions based on a large language model according to claim 1, characterized in that: The voice directions guidance interface is displayed in a fixed area of the display unit in the form of text or icons, and is used to prompt users to initiate voice directions, transfer inquiries and destination guidance operations in Mandarin or local dialect; the guidance interface includes at least one of the following guidance prompts: "Please just say your destination", "Voice directions do not require touching", "Dialect search supported".
5. The electronic bus stop sign for directions based on a large language model according to claim 1, characterized in that: The local AI large speech model unit has a built-in dialect-Mandarin mixed speech recognition model that supports end-to-end recognition and understanding of local dialects, Mandarin, and mixed spoken language. It can automatically recognize at least one of the following user intent types: asking for directions to the destination, checking vehicle arrival time, planning transfer routes, querying station information, and querying first and last train times.
6. The electronic bus stop sign for directions based on a large language model according to claim 1, characterized in that: The AI voice interaction mode completely eliminates the reliance on touch screens and physical buttons. Users can complete the query and guidance operations entirely through voice input. The device does not have a touch interface or physical button interface for query operations.
7. The electronic bus stop sign for directions based on a large language model according to claim 6, characterized in that: Since the reliance on touchscreen queries has been eliminated, the display unit of the wayfinding electronic bus stop sign can use a non-touchscreen outdoor high-reliability display screen, thereby reducing overall power consumption, reducing maintenance workload and replacement costs due to touchscreen failures, and making the upgraded equipment low-maintenance and low-power.
8. The electronic bus stop sign for directions based on a large language model according to claim 1, characterized in that: The local AI large voice model unit is deployed inside the edge computing unit and completes local inference operations for speech recognition, semantic understanding and route planning in an offline state, without relying on cloud servers for real-time interaction.
9. The electronic bus stop sign for directions based on a large language model according to claim 1, characterized in that: The pickup module adopts a multi-microphone array structure, which has a far-field pickup capability of 3-5 meters, and has built-in environmental noise suppression, echo cancellation and wind noise filtering algorithms to adapt to effective voice acquisition in noisy and windy outdoor environments.
10. The electronic bus stop sign for directions based on a large language model according to claim 1, characterized in that: The standardized interface includes a data communication interface and a power interface. The data communication interface adopts at least one of a general serial port, a network interface, or an I / O interface, and is used to interface with the original systems of existing traditional electronic bus stop signs from different manufacturers, years, and models, and has the ability to be replicated and promoted across models and batches.