Methods for controlling the return of media files in real-time translation scenarios, control terminals, and head-mounted acquisition devices.

By employing a dual-wireless link architecture and dynamic backhaul licensing conditions, the resource contention between translation services and media file transmission in real-time translation scenarios is resolved. This achieves a balance between the stability of low-latency translation services and the efficiency of large-volume media file transmission, thereby improving the communication stability and data backhaul reliability of translation sessions.

CN123124130APending Publication Date: 2026-11-17IFLYTEK CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611503324.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-24
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing technologies struggle to balance the stability of low-latency translation services with the efficiency of transmitting large media files in real-time translation scenarios, resulting in translation data latency jitter and stuttering.

Method used

A dual-wireless link architecture is adopted, with the first link used for real-time translation data and the second link used for media file transmission. The preset return permission conditions are determined by real-time activity status and communication status, and the media file transmission is dynamically scheduled to ensure physical isolation between translation services and media transmission.

Benefits of technology

It achieves a balance between the stability of low-latency translation services and the efficiency of large-volume media file transmission, avoids interference of media file transmission with translation services, and improves the communication stability of translation sessions and the reliability of data return.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123124130A_ABST
    Figure CN123124130A_ABST
Patent Text Reader

Abstract

This application discloses a method for controlling the return of media files in a real-time translation scenario, a head-mounted acquisition device, and a control terminal, relating to the fields of smart wearable devices and mobile communication technology. The method includes: establishing a first wireless link between the control terminal and the head-mounted acquisition device to interact with real-time translation data; when there is a media object to be returned, determining whether preset return permission conditions are met based on the real-time activity status of the translation session and the communication status of the second wireless link; and, if the conditions are met, establishing a second wireless link between the control terminal and the head-mounted acquisition device to receive the original media file. This application can achieve dual-link separate transmission of translation data and media files, avoiding media return from occupying real-time translation link resources, ensuring low-latency stability in the translation process, and ensuring that media files are accurately merged into the translation record.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application relates to the fields of smart wearable devices and mobile communication technology, and more specifically, to a method for controlling the return transmission of media files in a real-time translation scenario, a control terminal, and a head-mounted acquisition device. Background Technology

[0002] With the rapid development of artificial intelligence and wearable technology, head-mounted devices such as smart glasses with real-time voice translation capabilities have been widely used in scenarios such as international conferences, cross-border tourism, and foreign language learning. These devices typically establish a wireless connection with a control terminal to achieve real-time audio data interaction and translation playback. In a typical translation session, the system not only needs to process low-latency sensitive audio streams, control commands, and subtitle status data, but also frequently encounters media acquisition behaviors triggered by the user, such as taking photos, recording audio, or recording short videos. In existing collaborative communication solutions, a single main wireless communication link is usually maintained between the head-mounted device and the control terminal to carry various business data, including translation data and media files. The mainstream technical approach is to transmit data through a single link, or to use time-division multiplexing mechanisms at the physical layer to coordinate the coexistence of different wireless standards (such as Bluetooth and WiFi) to ensure that data can be transmitted to the other end device.

[0003] However, in complex real-time translation applications, existing technologies face the challenge of balancing the stability of low-latency translation services with the efficiency of transmitting large media files. Because translated audio frames and control commands are extremely sensitive to latency, while media files such as photos and videos are massive and take a long time to transmit, competition for limited wireless channel resources within the same session can easily lead to transmission congestion. This resource contention often results in significant fluctuations in end-to-end latency of real-time translation data, and may even cause translation interruptions or stuttering during media transmission, impacting the user's final experience. Summary of the Invention

[0004] In view of the above problems, this application proposes to provide a method, control terminal, and head-mounted acquisition device for controlling the return transmission of media files in real-time translation scenarios, so as to balance the stability of low-latency translation services with the efficiency of large-volume media file transmission. The specific solution is as follows:

[0005] Firstly, a method for controlling the return transmission of media files in a real-time translation scenario is provided, applied to a control terminal working in conjunction with a head-mounted acquisition device. The method includes:

[0006] A first wireless link is established with the head-mounted acquisition device, and relevant data is translated in real time by interacting with the head-mounted acquisition device through the first wireless link during the translation session;

[0007] When there is a media object to be returned, at least based on the real-time activity status of the translation session and the communication status of the second wireless link, it is determined whether the preset return permission conditions are met, wherein the media object is the original media file that was acquired and locally cached by the head-mounted acquisition device during the translation session.

[0008] If the preset return permission conditions are met, a second wireless link is established with the head-mounted acquisition device, and the original file of the media object returned by the head-mounted acquisition device is received through the second wireless link.

[0009] In one possible design, in another implementation of the first aspect of the embodiments of this application, the first wireless link is a Bluetooth link or a Bluetooth Low Energy link; the second wireless link is a WiFi link.

[0010] In one possible design, in another implementation of the first aspect of the embodiments of this application, the real-time activity state includes one or more of the following: whether there is active speech, whether a translation is being broadcast, and whether a translated segment is being processed.

[0011] In one possible design, in another implementation of the first aspect of the embodiments of this application, when determining whether the preset return permission conditions are met, in addition to relying on the real-time activity status of the translation session and the communication status of the second wireless link, one or more of the following are also considered:

[0012] Return window status, first wireless link load, and control terminal device status;

[0013] The return window state includes at least one of the following: a post-meeting return window triggered after the translation session ends, an active import window triggered by the user, and a silent return window that satisfies the requirement that the real-time activity state is not in a state susceptible to return interference.

[0014] In one possible design, in another implementation of the first aspect of the embodiments of this application, the process of determining whether the preset return permission conditions are met includes:

[0015] The backhaul permission score is calculated based on at least one of the following: the real-time activity status of the translation session, the communication status of the second wireless link, the backhaul window status, the load of the first wireless link, and the device status of the head-mounted acquisition device.

[0016] When the backhaul permission score is higher than a preset threshold, it is determined that the preset backhaul permission conditions are met.

[0017] In one possible design, in another implementation of the first aspect of the embodiments of this application, the device status of the head-mounted acquisition device includes power level and / or battery temperature, wherein the power level is positively correlated with the backhaul permission score and the battery temperature is negatively correlated with the backhaul permission score;

[0018] The process of determining whether the preset return permission conditions are met also includes:

[0019] When the battery level is lower than a preset battery level threshold, it is determined that the preset data transmission permission conditions are not met. This process continues until the head-mounted data acquisition device enters the charging state, at which point the data transmission permission score is calculated, and the data transmission permission score is used to determine whether the data transmission permission conditions are met.

[0020] In one possible design, in another implementation of the first aspect of the embodiments of this application, before determining whether the preset return permission conditions are met, the method further includes:

[0021] The device receives lightweight metadata of the media object uploaded by the head-mounted acquisition device through the first wireless link. The lightweight metadata is generated and uploaded by the head-mounted acquisition device after acquiring the media object.

[0022] A media anchor object matching the media object is created based on the lightweight metadata. The media anchor object is used to locate the insertion position of the media object in the translation record.

[0023] After receiving the original file of the media object from the head-mounted acquisition device, the original file of the media object is merged into the translation record based on the media anchor object matched by the media object.

[0024] In one possible design, in another implementation of the first aspect of the embodiments of this application, the lightweight metadata includes the acquisition timestamp and media identifier of the media object;

[0025] The process of creating a media anchor object that matches the media object based on the lightweight metadata includes:

[0026] Based on the acquisition timestamp and the current segment timeline maintained locally by the control terminal, the segment number where the acquisition timestamp is located is determined and used as the segment reference of the media object;

[0027] Create a media anchor object that includes an anchor identifier, the acquisition timestamp, the media identifier, the segment reference, and the return status, wherein the return status is marked as pending return.

[0028] In one possible design, in another implementation of the first aspect of this application, the process of merging the original file of the media object into the translation record based on the media anchor object matched by the media object includes:

[0029] Retrieve the segment reference from the media anchor object that matches the media object;

[0030] If the segment reference represents the j-th translated segment, then the original file of the media object is inserted after the j-th translated segment in the translation record, and the return status in the media anchor object is updated to complete;

[0031] If the segment reference represents the space between the j-th and j+1-th translated segments, then the original file of the media object is inserted into the translation record between the j-th and j+1-th translated segments, and the postback status in the media anchor object is updated to complete.

[0032] In one possible design, in another implementation of the first aspect of the embodiments of this application, the media anchor object further includes at least one of the following:

[0033] Session identifier, record identifier, start and end timestamps of the translated segment corresponding to the segment referenced, and offset position within the sentence;

[0034] The session identifier is used to identify the current translation session, the record identifier is used to identify the current translation record, and the sentence offset position indicates the relative position of the acquisition timestamp within the start and end time interval of the translation segment corresponding to the sentence segment reference.

[0035] In one possible design, in another implementation of the first aspect of the embodiments of this application, when the media object is an audio or video segment covering two or more translated sentence segments, the acquisition timestamp is the start and end timestamp of the acquisition of the audio or video segment;

[0036] The process of determining the segment number of the acquired timestamp as the segment reference of the media object based on the acquired timestamp and the current segment timeline maintained locally by the control terminal includes:

[0037] Based on the start and end timestamps of the data collection and the current segment timeline maintained locally by the control terminal, the starting segment number of the data collection start timestamp and the ending segment number of the data collection end timestamp are determined, and the starting segment number and the ending segment number are used as segment references for the media object.

[0038] The process of merging the original file of the media object into the translation record based on the media anchor object matched by the media object includes:

[0039] Based on the media anchor object matched by the media object, the original files of the audio or video clips are merged into the translation record, and the segment intervals corresponding to the starting segment number and the ending segment number are defined.

[0040] In one possible design, in another implementation of the first aspect of the embodiments of this application, the media anchor object includes a record identifier, the record identifier being used to identify the current translation record;

[0041] The translation record corresponding to the translation session contains the source text, translated text, and audio data of the translation session. After the media object is merged, it forms a composite record under the same record identifier with the source text, translated text, and audio data.

[0042] In one possible design, in another implementation of the first aspect of the embodiments of this application, the media anchor object further includes an anchor identifier, and the photo-type media objects form a record-bound media collection under the same record identifier. When the record-bound media collection triggers a postback at different import entry points, it performs unified merging and deduplication through the same anchor identifier and the record identifier.

[0043] In one possible design, in another implementation of the first aspect of the embodiments of this application, the process of receiving the original file of the media object transmitted back by the head-mounted acquisition device via the second wireless link includes:

[0044] The second wireless link prioritizes receiving preview data of media objects with a size lower than a first preset threshold, and delays receiving the original file of media objects with a size higher than the second preset threshold.

[0045] The preview data and the original file are associated with the same anchor point identifier, and the preview data is replaced with the original file after the original file is received.

[0046] In one possible design, another implementation of the first aspect of the embodiments of this application further includes:

[0047] When the second wireless link is unavailable, the backhaul is interrupted, or the file verification fails, the backhaul status in the media anchor object is marked as pending backhaul or retrying, and the backhaul will continue to be received when the preset backhaul permission conditions are met in the future.

[0048] In one possible design, in another implementation of the first aspect of the embodiments of this application, before receiving the original file of the media object via the second wireless link, the following is further included:

[0049] Query the return status in the media anchor object that matches the media object. If the return status is marked as completed, skip receiving; if the return status is marked as pending return or retrying, perform the operation of receiving the original file of the media object.

[0050] Secondly, a method for controlling the return transmission of media files in a real-time translation scenario is provided, applicable to a head-mounted acquisition device that works in conjunction with a control terminal. The method includes:

[0051] A first wireless link is established with the control terminal, and relevant data is translated in real time by interacting with the control terminal through the first wireless link during the translation session;

[0052] During the translation session, media objects are captured, and the original files of the media objects are cached locally.

[0053] In response to the feedback instruction issued by the control terminal, a second wireless link is established with the control terminal, wherein the feedback instruction is generated by the control terminal after determining that the preset feedback permission conditions are met based at least on the real-time activity status of the translation session and the communication status of the second wireless link;

[0054] The original file of the media object is sent to the control terminal via the second wireless link.

[0055] In one possible design, in another implementation of the second aspect of the embodiments of this application, after acquiring the media object, it further includes:

[0056] Lightweight metadata of the media object is generated and sent to the control terminal via the first wireless link. The lightweight metadata is used to trigger the control terminal to establish a media anchor object that matches the media object. The media anchor object is used to locate the insertion position of the media object in the translation record.

[0057] In one possible design, in another implementation of the second aspect of the embodiments of this application, the process of sending the original file of the media object to the control terminal via the second wireless link includes:

[0058] Preview data of the media object with a size lower than a first preset threshold is sent first via the second wireless link, while the original file of the media object with a size higher than a second preset threshold is sent later.

[0059] Thirdly, a control terminal is provided, including: a memory and a processor;

[0060] The memory is used to store programs;

[0061] The processor is configured to execute the program to implement the various steps of the media file return control method in the real-time translation scenario as described in any of the first aspects above.

[0062] Fourthly, a head-mounted acquisition device is provided, including a multimedia acquisition module and playback model, a communication module, a memory, and a processor;

[0063] The memory is used to store programs;

[0064] The processor is configured to execute the program to implement the various steps of the media file return control method in the real-time translation scenario as described in any of the second aspects above.

[0065] By employing the aforementioned technical solution, this application establishes a first wireless link on the control terminal side, which works in conjunction with the head-mounted acquisition device, to interactively translate relevant data in real time. When there is media to be transmitted back, the system dynamically determines whether preset transmission permission conditions are met based on the real-time activity status of the translation session and the communication status of the second wireless link. If the conditions are met, the second wireless link is established to receive the original media file. Through this dual-link architecture and dynamic admission mechanism, low-latency translation-related data always exclusively occupies the first wireless link, while the original media file is only scheduled for transmission when the translation risk is controllable and the second link is available. This achieves effective isolation between translation transactions and media transmission transactions in terms of physical link and time, balancing the stability of low-latency translation services with the transmission efficiency of large-volume media files.

[0066] Furthermore, based on the perception of real-time activity status and the determination of return permission conditions, this application ensures that media return operations automatically give way during critical periods such as active audio, translated text broadcasting, or sentence / segment processing, avoiding bandwidth contention and congestion of the real-time translation link caused by the transmission of large amounts of media objects. Therefore, it avoids translation stuttering and disconnections caused by the immediate transmission of media files, significantly improving communication stability and data return reliability during the translation session. Attached Figure Description

[0067] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0068] Figure 1 A schematic diagram of an implementation system architecture for a media file return control method in a real-time translation scenario provided in this application embodiment;

[0069] Figure 2This is a schematic diagram of a media file return control method in a real-time translation scenario provided by an embodiment of this application from the perspective of the control terminal;

[0070] Figure 3 This is a schematic diagram of another method for controlling the return of media files in a real-time translation scenario, provided from the perspective of the control terminal, according to an embodiment of this application.

[0071] Figure 4 A schematic diagram of a data transmission link between a head-mounted acquisition device, a control terminal, and a translation server provided in an embodiment of this application;

[0072] Figure 5 This is a schematic flowchart illustrating a method for controlling the return of media files in a real-time translation scenario from the perspective of a head-mounted acquisition device, as provided in this application embodiment.

[0073] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0074] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0075] It is understood that before using the technical solutions disclosed in the various embodiments of this application, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in this application in an appropriate manner in accordance with relevant laws and regulations, and user authorization should be obtained.

[0076] This application provides a media file return control scheme for real-time translation scenarios, referring to... Figure 1 As shown, the solution in this application is applied to a system consisting of a head-mounted data acquisition device and a collaborative control terminal. Optionally, the system may also include a translation server.

[0077] The control terminal can perform speech transcription, translation, and speech synthesis locally. Alternatively, the control terminal can send the audio stream to a translation server, which will then perform speech transcription, translation, and speech synthesis, and return the results to the control terminal.

[0078] In this context, "head-mounted data acquisition device" refers to any electronic device that can be worn on the head and has the function of acquiring data (such as audio, video, and images), such as AI glasses, smart glasses, audio glasses, smart helmets, headsets / earphones, smart masks, and goggles. Subsequent embodiments of this application will use smart glasses as an example for illustrative purposes.

[0079] The control terminal is an electronic device that communicates with the head-mounted acquisition device, such as a mobile phone, tablet, desktop computer, smartwatch / wristband, laptop, translator, voice recorder, or portable gateway. Application software can be installed on the control terminal to collaboratively execute media file transmission control methods in real-time translation scenarios with the head-mounted acquisition device.

[0080] In real-time translation scenarios, a head-mounted acquisition device captures the audio stream to be translated and sends it to the control terminal via a first wireless link (real-time translation link). The control terminal processes the audio stream locally or through a translation server, performing transcription, translation, and speech synthesis. The translated and target language synthesized audio streams are then returned to the head-mounted acquisition device via the first wireless link, whereby the head-mounted acquisition device outputs the translated subtitles and the target language synthesized audio stream.

[0081] In addition, the first wireless link is also used to carry control commands, translate relevant status data, and translate relevant low-latency data in real time.

[0082] During a translation session, the user can trigger actions such as taking photos, recording audio, and capturing videos using a head-mounted acquisition device to obtain multimedia objects. Generally, multimedia objects are large in size. To avoid multimedia objects competing for transmission resources with translation-related data, this application determines that, when preset backhaul permission conditions are met, a second wireless link (media backhaul link) is established between the head-mounted acquisition device and the control terminal to transmit the multimedia objects back to the control terminal via this second wireless link. The control terminal then incorporates the multimedia objects into the translation record. The translation record may include the source text of the translation session, the translated text, and synthesized target audio data.

[0083] Next, this application will describe the media file return control method from the perspectives of the control terminal and the head-mounted acquisition device.

[0084] Reference Figure 2 First, from the perspective of the control terminal, the implementation process of the media file return control method is introduced, which may include the following steps:

[0085] Step S100: Establish a first wireless link with the head-mounted acquisition device, and interact with the head-mounted acquisition device in real time to translate relevant data during the translation session.

[0086] To achieve normal translation sessions between the head-mounted acquisition device and the control terminal, a dedicated communication channel, namely the first wireless link, needs to be built between them to carry low-latency services.

[0087] The primary wireless link carries real-time translation-related data between the head-mounted acquisition device and the control terminal. Its core transmission characteristics are low latency and low power consumption. Bluetooth or Bluetooth Low Energy links offer advantages such as rapid connection establishment, low power consumption, and a transmission protocol stack adapted to high-frequency interactions with small data packets. This meets the stringent real-time requirements of translation sessions for data such as translated audio segments, translation control commands, translation status data, real-time subtitle data, and translated text playback control. Using a Bluetooth or Bluetooth Low Energy link as the primary wireless link ensures that the transmission of real-time translation data remains uninterrupted even when large media file transfers consume system resources, thus maintaining session continuity and smoothness.

[0088] This first wireless link operates independently of the media file return channel within the system architecture, ensuring that core translation data is not interfered with by large file transfers. During a translation session, the control terminal continuously receives the audio stream captured by the head-mounted acquisition device via this first wireless link and transmits translated data or control commands returned from the local machine or translation server to the head-mounted acquisition device in real time, thereby ensuring the smoothness of simultaneous interpretation or dialogue translation.

[0089] Step S110: Determine whether the preset return permission conditions are met.

[0090] Specifically, when there are media objects to be transmitted back, it is determined whether the preset transmission permission conditions are met, based at least on the real-time activity status of the translation session and the communication status of the second wireless link.

[0091] The media objects to be transmitted back can refer to the original media files captured and cached locally by the head-mounted acquisition device during the translation session. Their data volume is usually much larger than that of real-time translation data frames, such as photographs, audio files, or short video clips.

[0092] When a media object awaiting return is identified, the return permission determination logic is triggered. This logic comprehensively assesses the resource contention between the current translation task and the communication environment, and the determination is based on at least the real-time activity status of the translation session and the communication status of the second wireless link. The real-time activity status of the translation session reflects the busyness and anti-interference capability of the current translation task, including whether there is active speech, whether the translation is being broadcast, or whether there are translation segments being processed.

[0093] The second wireless link is a dedicated communication channel for transmitting raw media files. It can employ high-bandwidth wireless communication technologies, such as WiFi links (including WiFi Direct, access point access, and local area networks). Alternatively, cellular data or other high-throughput channels can be used.

[0094] The second wireless link is logically and physically isolated from the first wireless link and undertakes media backhaul tasks with large data throughput.

[0095] The communication status of the second wireless link reflects the availability and quality of the alternative backhaul channel, including the signal strength, connection status, or bandwidth margin of the WiFi link.

[0096] The preset backhaul permission conditions are admission rules set based on the above indicators to ensure that media backhaul actions do not interfere with the real-time translation experience. For example, when the real-time activity status shows that the active voice input or translation playback stage is in progress, even if the second wireless link communication status is good, it can be determined that the preset backhaul permission conditions are not met, thereby delaying the backhaul operation to ensure the continuity and low latency of the translation.

[0097] If the preset return permission conditions are met, proceed to step S120.

[0098] Step S120: Establish a second wireless link with the head-mounted acquisition device, and receive the original file of the media object returned by the head-mounted acquisition device through the second wireless link.

[0099] When the preset return transmission permission conditions are met, it indicates that the current environment allows for large file transfers without affecting the translation experience. At this point, the control terminal establishes a second wireless link with the head-mounted acquisition device. This second wireless link is a dedicated communication channel for transmitting the original media files. After the second wireless link is successfully established, the control terminal receives the original files of the media objects returned by the head-mounted acquisition device through this link. These original files are complete data stored in the local cache of the head-mounted acquisition device at the time of media acquisition, such as uncompressed high-definition photos or complete audio files. During transmission, the connection of this link is maintained until the original file data of the media object is completely received. Through this separate transmission strategy, the first wireless link can continue to maintain the interaction of real-time translation-related data without being affected by the large-volume media file return transmission on the second wireless link, thereby achieving concurrent execution and resource isolation of translation and media return transmission services.

[0100] If the preset data return permission conditions are not met, the head-mounted acquisition device is instructed to continue caching the original file of the media object locally and not return the original file.

[0101] The method provided in this embodiment solves the problem of resource contention between the real-time translation link and the media file return by physically isolating real-time translation data interaction and media file return to two independent wireless links, and combining this with a dynamic permission determination mechanism based on real-time activity and communication status. The dedicated nature of the first wireless link ensures the transmission stability of low-latency data such as translated audio and control commands, while the on-demand establishment of the second wireless link ensures that media objects are transmitted only when the system is idle or resources are sufficient. This dual-link architecture and intelligent scheduling strategy effectively avoids latency jitter caused by large media files crowding out real-time translation bandwidth, and provides a data transmission foundation for the accurate merging of subsequent media objects into the translation record.

[0102] In some possible implementations, real-time activity status may include: whether there is active speech, whether a translation is being played, and whether one or more of the translated segments are being processed.

[0103] Real-time activity status, as a set of key indicators for quantifying the continuity risk of translation sessions, refers to the instantaneous state of several links in a translation session that are directly related to ensuring real-time performance. Among them, the presence of active speech is used to characterize the activity level of the translation input source. This is determined by monitoring the acoustic signal strength of the microphone of a head-mounted acquisition device (such as AI glasses) and the results of speech endpoint detection. When a speech signal higher than a preset energy threshold is detected, or when the acoustic feature model identifies a valid speech frame, active speech is determined to exist.

[0104] Whether the translation is being played indicates the occupancy status of the output feedback channel. This is determined by monitoring the audio playback stream status or screen subtitle refresh status of the control terminal (such as a mobile app). When the audio decoding buffer is outputting an audio stream, or the subtitle display module is performing rendering updates, it is determined that the translation is being played.

[0105] The presence of a processing segment indicates the occupancy status of background computing resources. This is determined by monitoring the inference queue of the Natural Language Processing (NLP) model or the request response status of the translation service. If the translation engine is performing feature extraction, decoding, or text generation on the current input speech stream, it is determined that a processing segment is in progress.

[0106] Based on the fine-grained definition of real-time activity states described above, the system can accurately identify critical periods in the translation process that cannot be interrupted. When any sub-state is triggered as "yes," it indicates that the system is currently in a phase with high real-time requirements. Introducing a large file transfer task at this time would highly likely cause audio stuttering, translation delays, or sentence / segment loss. Therefore, this real-time activity state is directly used as a criterion for suppressing media transfer, ensuring that media transmission only occurs during periods when all the aforementioned states are negative (i.e., silent and idle), thereby guaranteeing the smoothness and integrity of the translation interaction.

[0107] In some embodiments of this application, when determining whether the preset return permission conditions are met in the aforementioned step S110, in addition to relying on the real-time activity status of the translation session and the communication status of the second wireless link, one or more of the following may also be considered:

[0108] The data includes the return window status, the load of the first wireless link, and the device status of the control terminal.

[0109] The return window status includes at least one of the following: a post-meeting return window triggered after the translation session ends, an active import window triggered by the user, and a silent return window that meets the requirement that the real-time activity status is not in a state susceptible to return interference.

[0110] The post-transmission window state defines the time policy boundaries for allowing media data retransmission, and its construction is based on the phase division of the entire lifecycle of the translation task. This process divides the translation session's timeline into intervals such as session in progress, session ended, and idle interaction state, and configures corresponding transmission permission flags for each interval. Specifically, the post-transmission window is activated when a translation session end signal is detected; at this time, since there is no longer a need for real-time translation data interaction, system resources can be completely released for media transmission. The active import window is determined by monitoring whether the user enters a specific application interface, such as the translation record page or device management page, to respond to the user's need to view historical records immediately. The silent retransmission window is defined by analyzing the voice activity detection and translation playback status in the real-time activity state. It is only considered a valid window when a period of duration exceeding a preset silent threshold (e.g., 2 seconds) is detected without any background processing tasks, ensuring that the retransmission action does not interrupt the currently ongoing voice interaction. These three window states are logically mutually exclusive and complementary, together forming the retransmission timing scheduling framework.

[0111] The first wireless link load, as a physical indicator of resource utilization, reflects the current congestion status of the real-time translation link. This data is typically obtained by the link status monitoring module by statistically analyzing data throughput, buffer queue length, and retransmission rate per unit time. When the first wireless link load exceeds a preset load threshold, it indicates that the current link bandwidth is nearing saturation; scheduling media backhaul tasks at this time can easily lead to transmission delays in translated audio frames or control signaling. Therefore, when determining backhaul permission conditions, this load indicator is considered as a negative weighting factor to ensure that media backhaul requests are suppressed or queued during periods of high real-time translation activity, thereby guaranteeing the low-latency characteristics of core translation services.

[0112] The device status of the control terminal is used to characterize the computing resources and power margin for performing backhaul tasks. Its monitoring dimensions include, but are not limited to, remaining battery power, CPU utilization, memory usage, and device temperature. In low-power mode, to extend battery life, the system tends to shut down high-power wireless transmission modules; under high temperature or high load conditions, to prevent system thermal throttling or response lag, concurrent transmission of large files is also limited. Therefore, in some optional implementations, a device is only considered to have met the backhaul permission conditions when its status indicates sufficient battery power, normal temperature, and ample computing resources.

[0113] By combining the backhaul window status, the load of the first wireless link, the device status of the control terminal, and the real-time activity status and communication status of the second wireless link, a multi-dimensional backhaul decision mechanism is constructed. This mechanism not only focuses on network-level connectivity but also deeply perceives the activity patterns at the service level and the resource health status at the device level. This allows it to accurately select the optimal backhaul time in complex and ever-changing real-world usage scenarios that neither interferes with the translation experience nor hinders execution, thus solving the problem of transmission conflicts or resource exhaustion caused by single-condition judgments.

[0114] In some embodiments of this application, a specific implementation method is provided for determining whether preset backhaul permission conditions are met. This process can calculate a backhaul permission score according to a set backhaul decision formula. Specifically:

[0115] The backhaul permission score is calculated based on at least one of the following: the real-time activity status of the translation session, the communication status of the second wireless link, the backhaul window status, the load of the first wireless link, and the device status of the head-mounted acquisition device.

[0116] The backhaul permission score, a comprehensive quantitative indicator representing the suitability of enabling the media backhaul link at the current moment, is obtained through normalization and weighted mapping of multi-dimensional status indicators. This score aims to reflect the dynamic balance between translation continuity risk, link resource sufficiency, and user intent.

[0117] During the calculation process, the system first extracts features and maps values ​​to each discrete or continuous state information. For the communication state of the second wireless link, the system obtains the current Received Signal Strength Indicator (RSSI) value of the WiFi through a signal strength detection mechanism and maps it to a communication quality coefficient between 0 and 1; signal strength is positively correlated with this coefficient. For the real-time activity status of the translation session, the system assesses the translation continuity risk based on factors such as the presence of active speech, whether the translation is being broadcast, and whether there are currently processed translation segments; the higher the risk level, the greater the corresponding penalty weight. For the return window status, the system identifies the current time window type, including post-session return windows after the translation session ends, user-triggered active import windows, or silent return windows; if it falls within a window period with extremely low interference to real-time services, a higher reward score is assigned to encourage data transmission during this period.

[0118] In one embodiment, the evaluation process of the above-mentioned multidimensional states can be formalized as the following multi-objective decision function:

[0119]

[0120] In the formula, RouteScore_t represents the backhaul permission score at time t, W_t represents the availability and quality of the second radio link at time t, G_t represents the backhaul window coefficient, U_t represents the weight actively triggered by the user, L_t represents the load of the first radio link, A_t represents the current translation continuity risk, which is related to the real-time activity status, and b1 to b5 are weight coefficients.

[0121] The physical meanings and settings of each variable are as follows: W_t is used to quantify the transmission capacity of the second wireless link (e.g., WiFi), and its value is usually determined based on RSSI, signal-to-noise ratio, or link layer rate; G_t is used to distinguish the priority of different backhaul windows. For example, the value of the post-meeting backhaul window is higher than that of the silent backhaul window to utilize idle bandwidth during non-critical periods; U_t is used to characterize the intensity of the user's immediate needs. When the user actively clicks the import operation, this variable is set to a higher value to increase the backhaul priority; L_t is used to characterize the busyness of the first wireless link (e.g., Bluetooth). The higher the load, the less suitable it is to start the second link operation that may cause resource contention; A_t is used to characterize the possibility of interference with real-time translation. For example, the value is set to a higher value when active voice is detected, thereby suppressing the backhaul to avoid occupying processing resources; the weight coefficients b1 to b5 are used to adjust the proportion of each factor in the total score. Among them, the constraint priorities represented by b4 and b5 are usually set to be higher than those in ordinary file synchronization scenarios to ensure the absolute priority of the core translation business.

[0122] When the backhaul permission score is higher than the preset threshold, the preset backhaul permission conditions are determined to be met.

[0123] A preset threshold serves as a quantitative threshold for determining whether to allow the backhaul link to be enabled. Its setting requires a comprehensive balance between user experience expectations and system resource constraints. During system initialization, historical operational data can be collected and statistically analyzed to determine key performance indicators such as translation stuttering rate and transmission success rate caused by enabling backhaul within different score ranges. This initial threshold can then be determined based on the business's tolerance for misjudgments. For example, if the strategy prioritizes ensuring an extremely smooth translation experience, the threshold can be appropriately increased, allowing backhaul only when resources are extremely abundant; if the strategy prioritizes rapid synchronization of media files, the threshold can be appropriately decreased. This threshold can be a fixed value during operation or dynamically adjusted based on system performance degradation (such as decreased processing power due to equipment aging).

[0124] The calculated backhaul permission score is compared with a preset threshold to execute the judgment logic. If the backhaul permission score is higher than the preset threshold, it indicates that the comprehensive evaluation of the current system status supports enabling the backhaul link, and the benefits of backhaul (such as meeting users' immediate needs and utilizing idle bandwidth) outweigh the potential risks (such as increased translation latency). Therefore, the preset backhaul permission conditions are met. At this time, the control terminal triggers subsequent link establishment instructions, allowing the head-mounted acquisition device to send the original media file through the second wireless link. If the backhaul permission score is not higher than the preset threshold, it indicates that there are bottleneck factors such as poor link quality, high real-time translation risk, or heavy device load, and the backhaul permission conditions are not met. The system will maintain the state of communicating only through the first wireless link, continue to cache the media file without backhauling, thereby avoiding interference with the core real-time translation business.

[0125] The above solution maps multi-dimensional information, including the real-time translation activity status, the second wireless link communication status, the return window status, the first wireless link load, and the device status of the head-mounted acquisition device, into a unified return permission score. This score is then quantified using a preset threshold, enabling refined decision-making regarding the timing of media return. This mechanism overcomes the limitations of single-rule judgment, dynamically adjusting the return strategy based on the system's current overall resource status. While ensuring the continuity of real-time translation, it maximizes the use of idle network resources for media file return, thereby improving business processing efficiency and user interaction experience throughout the entire translation session.

[0126] In some embodiments of this application, the device status of the head-mounted acquisition device may include battery power and / or battery temperature, with battery power being positively correlated with the backhaul permission score and battery temperature being negatively correlated with the backhaul permission score.

[0127] In this decision-making logic, battery power and battery temperature are introduced as key influencing factors to dynamically adjust the backhaul permission score. The battery power parameter reflects the current remaining energy reserve of the head-mounted acquisition device, and its value is positively correlated with the backhaul permission score; that is, when the battery power is sufficient, the system tends to assign a higher backhaul priority. The battery temperature parameter characterizes the heat dissipation pressure inside the device, and its value is negatively correlated with the backhaul permission score; that is, when the temperature rises, the system lowers the backhaul permission score to suppress high-energy-consuming transmission behavior. This differentiated weighting mechanism can effectively balance media backhaul needs with device physical security limitations, preventing translation sessions from being interrupted due to battery overheating or depletion caused by long-term, high-intensity data backhaul.

[0128] Furthermore, the process of determining whether the preset return permission conditions are met may also include:

[0129] When the battery level is lower than the preset battery threshold, it is determined that the preset data transmission permission conditions are not met. The process continues until the head-mounted data acquisition device enters the charging state, at which point the data transmission permission score is calculated, and the data transmission permission score is used to determine whether the data transmission permission conditions are met.

[0130] In this embodiment, an absolute power threshold is set to ensure the minimum operational safety of the device.

[0131] The system reads the battery management system interface of the head-mounted acquisition device in real time to obtain the current battery level and compares it with a preset battery threshold. Once the battery level is detected to have fallen below the preset threshold, regardless of the excellent communication status of the second wireless link or the suitable return window status, it is directly and forcibly determined that the preset return permission conditions are not met. This determination logic is independent of the conventional decision-making process based on score calculation and has the highest execution priority. Essentially, it forcibly prohibits return operations when the energy reserves are insufficient to support a complete WiFi transmission and maintain the remaining session time. This hard constraint ensures that the battery life priority of the core translation function always takes precedence over non-real-time media transmission tasks.

[0132] Once the head-mounted acquisition device is confirmed to be charging, it means that continuous external power supply can offset the high energy consumption of WiFi backhaul. At this point, the system unlocks the backhaul function. The backhaul permission score calculation process is then reactivated. This calculation comprehensively considers parameters such as the real-time activity status of the translation session, the communication status of the second wireless link, the current battery level (which is increasing due to charging and contributes positively to the score), and battery temperature. Due to the intervention of external power, the negative impact of battery temperature and battery level on the backhaul decision is significantly weakened or even eliminated, thus the calculated backhaul permission score is usually significantly improved. Finally, based on the relationship between this improved backhaul permission score and a preset threshold, it is determined whether the backhaul permission conditions are met. If so, a second wireless link is established for media file backhaul. This closed-loop control logic—hard prohibition when low battery and automatic recovery after charging—ensures the device's survivability in low-battery mode and ensures that backlogged media backhaul tasks can be completed promptly and automatically after power is connected.

[0133] Reference Figure 3 This provides an alternative implementation of the media file return control method in real-time translation scenarios, specifically:

[0134] Before determining whether the preset return permission conditions are met in the aforementioned step S110, the following steps are also included:

[0135] Step S200: Receive lightweight metadata of the media object uploaded by the head-mounted acquisition device through the first wireless link.

[0136] Lightweight metadata is generated and uploaded after media objects are captured by a head-mounted capture device.

[0137] Transmitting lightweight metadata via a first wireless link (e.g., Bluetooth) instead of directly transmitting the original media file effectively avoids excessive bandwidth consumption. As a digital fingerprint of the media object, the lightweight metadata is generated immediately after the media acquisition action is completed at the head-mounted acquisition device. Taking a scenario where a user wears AI glasses to translate and take photos as an example, the lightweight metadata includes at least a media identifier and an acquisition timestamp. Further information can be included, such as media type (e.g., image, audio, video), session ID, file digest checksum, and local storage path pointer. In some optional implementations, after completing the media file writing operation, the head-mounted acquisition device extracts the acquisition timestamp and file size from the file system attributes, calculates a file digest using a hash algorithm to establish a unique identifier, and encapsulates it into a structured data packet based on the context information of the current translation session. This packet is then sent to the control terminal in real time via the first wireless link. Through this mechanism, the control terminal can detect the generation of the media object and its key attributes in advance without receiving large files.

[0138] Step S210: Create a media anchor object that matches the media object based on the lightweight metadata.

[0139] The media anchor object is used to locate the insertion position of the media object in the translation record.

[0140] During real-time translation, the control terminal dynamically maintains the timeline of the current sentence segment, which records the start and end times of each sentence segment in the translation session. As the translation activity progresses, whenever a new speech activity is identified or a sentence is completed, the control terminal updates the latest sentence segment number and its corresponding time range.

[0141] After receiving lightweight metadata, the control terminal parses the acquisition timestamp and compares it with the time intervals of each segment in the current segment timeline maintained locally by the control terminal to determine the segment number where the acquisition timestamp is located, which serves as the segment reference for the media object. Specifically, if the acquisition timestamp falls within the time interval [start_ts_j, end_ts_j] of the j-th segment, the media object is determined to belong to the j-th segment; if the acquisition timestamp falls within the silent interval between segments, its segment number is determined according to a preset strategy (such as the nearest principle or the sequential principle), thereby anchoring discrete event points on the continuous time stream to discrete segment sequences.

[0142] The process by which the control terminal compares and locates the collection timestamps in the lightweight metadata with the currently maintained translation segment timeline can be formally represented as the following decision function:

[0143]

[0144] In the formula, This represents the acquisition timestamp corresponding to media anchor point k; Indicates the end time of the j-th translated segment; Indicates the target position number where the media object should be inserted. When the time frame falls within the j-th sentence, the result is j, and the media object will be placed after the j-th sentence; when When located in a silent inter-sentence region, this function can find the position of the next sentence segment in time, thus avoiding the incorrect insertion of media content into the ongoing sentence.

[0145] After the control terminal receives the lightweight metadata and completes the segment location, it triggers the construction process of the media anchor object. This media anchor object is a structured data entity containing several key fields, such as anchor identifier, acquisition timestamp, media identifier, segment reference, and return status, where the return status is marked as pending return.

[0146] Anchor point identifiers are used to uniquely index the anchor point object in the control terminal's storage system. Acquisition timestamps and media identifiers are directly inherited from lightweight metadata, preserving the original time positioning information and file identity information. Segment references point to the segment sequence numbers determined through the preceding steps. The return status is a status flag used to track the progress of media file transmission, initially set to pending return. By encapsulating these fields into a unified logical object, the control terminal can establish a logical association between media objects and translation records without immediately acquiring large original files.

[0147] Further optionally, the media anchor object may also include a session identifier (session_id), a record identifier (record_id), the start and end timestamps of the translated segment corresponding to the segment reference, and the offset position within the segment (offset_in_segment), etc.

[0148] The media anchor object, as a key data structure connecting the acquisition moment and the final translation record, is internally designed to incorporate contextual information to support accurate merging across multiple scenarios. This media anchor object is generated after the head-mounted acquisition device produces a media object and the control terminal receives lightweight metadata. In addition to the basic media identifier and timestamp, it also includes auxiliary fields for precise positioning. Specifically, the media anchor object is stored in a structured data format, and its content includes the anchor identifier, media type, acquisition timestamp, segment references, and the aforementioned extended fields. The session identifier is used to globally and uniquely identify the current translation session to distinguish data generated by different translation tasks; the record identifier is used to identify the current translation record to ensure that media files can be accurately mounted to the corresponding user record entries; the translation segment start timestamp and end timestamp record the start and end times of the translation segment pointed to by the segment reference, respectively, providing time boundaries for the interval anchoring of media objects; the intra-sentence offset position represents the relative position of the acquisition timestamp within the start and end time interval of the translation segment corresponding to the segment reference, which is usually obtained by calculating the difference between the acquisition timestamp and the translation segment start timestamp. This value reflects the time point of the media acquisition action in the specific segment process.

[0149] Session identifiers maintain data isolation and uniqueness during multi-task concurrent or cross-time-segment data processing. Record identifiers ensure that even when multiple record segments or fragmented storage exist within the same session, media objects can be accurately assigned to the correct logical record unit, preventing incorrect cross-record associations. Segment references provide clear time window references for the start and end timestamps of the corresponding translated segments, enabling the system to accurately define the temporal relationship between media objects and translated text. This is particularly important for processing media content spanning multiple segments, such as audio recordings and short videos. Intra-segment offset positions precisely map the acquisition time of media objects to the time stream within the segment, supporting fine-grained synchronization at the sub-segment level. This allows for more accurate reproduction of the actual context at the time of media acquisition during playback or viewing.

[0150] By integrating the aforementioned extended fields into the media anchor object, the media anchor object can carry richer contextual information. It not only supports simply inserting media objects after a sentence, but also provides the necessary data foundation for handling complex temporal relationships (such as media cross-sentence and precise intra-sentence alignment), thereby improving the completeness and accuracy of the translation record.

[0151] Once created, the media anchor object is bound to the current translation session instance, logically pre-positioning the yet-to-arrive original media file in the correct position within the translation record. Compared to existing technologies that rely solely on file system timestamps for retrospective estimation, the media anchor object completes spatiotemporal positioning the instant the media is generated, providing a precise index for subsequent merging operations.

[0152] After receiving the original file of the media object returned by the head-mounted acquisition device in step S120, the method further includes:

[0153] Step S220: Based on the media anchor object matched by the media object, merge the original file of the media object into the translation record. This process may include:

[0154] S1. Retrieves a segment reference from the media anchor object that matches the media object.

[0155] This segment reference, calculated by the preceding steps based on the acquisition timestamp and the current segment timeline maintained locally on the control terminal, represents the logical affiliation of the media object's generation time within the translation sequence. By parsing this segment reference, it is possible to identify whether the media object belongs to a specific j-th translated segment or is located in a silent interval between two adjacent segments, thus providing a precise position index for subsequent merging operations.

[0156] S2. If the segment reference represents the j-th translated segment, then insert the original file of the media object after the j-th translated segment in the translation record, and update the postback status in the media anchor object to complete.

[0157] When a segment reference points to the j-th translated segment, it indicates that the user triggered media capture during the speech activity or processing of that j-th translated segment. The end position of the j-th translated segment is located in the translation record's data structure, and the original file of the media object is inserted immediately after it. This insertion method ensures that the media object is semantically closely related to its segment, avoiding incorrect attribution of captured media to the preceding or following sentence. After the insertion operation is complete, the return status flag in the corresponding media anchor object is immediately updated from pending return to completed, indicating that the media object has been successfully merged into the translation record, preventing subsequent retransmission or duplicate insertion. This process effectively solves the correspondence problem between media capture events occurring within a segment and the final record display, ensuring the logical coherence of the translation record.

[0158] S3. If the segment reference indicates the period between the j-th and j+1-th translated segments, then insert the original file of the media object into the position between the j-th and j+1-th translated segments in the translation record, and update the postback status in the media anchor object to complete.

[0159] When a sentence reference points to the interval between the j-th and j+1-th translated sentence segments, it indicates that the user triggered media capture during the interval between the end of the j-th translated sentence segment and the beginning of the j+1-th translated sentence segment. The gap between the j-th and j+1-th translated sentence segments is located in the translation record data structure, and the original file of the media object is precisely inserted into this interval. This insertion method accurately recreates the on-site timing of the user's snapshots or recordings between two sentences, avoiding the distortion of the timeline by awkwardly attaching inter-sentence captured media after a sentence segment. Similarly, after insertion, the return status in the media anchor object is updated to "completed" to synchronize the final state of the recorded object. This process, by differentiating between intra-sentence and inter-sentence insertion strategies, achieves fine-grained arrangement of media objects in the translation record, ensuring a high-fidelity reproduction of the conversation process in the composite translation record.

[0160] Through the above steps, based on the sentence and segment reference information carried in the media anchor object, a differentiated merging and insertion strategy and state update are automatically executed, so as to accurately integrate the delayed media object into the correct position of the translation record, so that the final generated record can truly reflect the spatiotemporal correspondence between media acquisition and speech translation during the translation session.

[0161] The method provided in this embodiment establishes a media anchor object carrying spatiotemporal location information in real time at the moment the media is generated, and strictly performs merging according to the anchor position after the subsequent file return is completed, ensuring that the media object can be accurately inserted into the correct position in the dialogue record. This method eliminates the uncertainty caused by inferring the attribution based solely on the file timestamp, ensuring that even if the original file arrives late, it can still be accurately inserted into the corresponding translated segment position, thereby guaranteeing the integrity and traceability of the composite translation record.

[0162] In some embodiments of this application, when the media object is an audio or video segment covering two or more translated sentence segments, the acquisition timestamp of the media object is specifically the start and end timestamps of the acquisition of the audio or video segment. Therefore, in the aforementioned embodiments, the process of determining the sentence segment number where the acquisition timestamp is located, based on the acquisition timestamp and the current sentence segment timeline maintained locally by the control terminal, as the sentence segment reference for the media object, includes:

[0163] Based on the start and end timestamps of the data collection and the current segment timeline maintained locally by the control terminal, the starting segment number of the data collection start timestamp and the ending segment number of the data collection end timestamp are determined, and the starting segment number and the ending segment number are used as segment references for the media object.

[0164] Based on the start and end timestamps of the data collection, the interval-based segment citations can accurately capture the dynamic span of long-duration media on the translation timeline. For a continuous recording spanning multiple segments, the segment citation clearly indicates that it begins at the m-th sentence and ends at the n-th sentence, avoiding content attribution bias caused by anchoring only at the start or end time, and ensuring accurate mapping of media metadata to the actual data collection situation.

[0165] Based on this, the process of merging the original files of the media objects into the translation record, according to the media anchor objects matched by the media objects, includes:

[0166] Based on the media anchor object matched with the media object (audio recording or video clip), the original file of the media object is merged into the translation record, and the sentence segment intervals corresponding to the start and end sentence segment numbers are defined.

[0167] Unlike the point insertion method for single images, this approach reads the start and end sentence / segment numbers from the media anchor object, locating the corresponding sentence / segment interval in the translation record. Subsequently, a logical association is established between the original audio or video clip and this sentence / segment interval. In the presentation of the translation record, this merging operation manifests as mounting or embedding the media object into the area defined by the start and end sentence / segments. For example, on the translation record's user interface, the audio or video clip is displayed as a complete entry spanning from sentence m to sentence n, or as an associated attachment marked on one side of the sentence / segment interval.

[0168] This interval merging method tightly couples audio or video clips with multiple consecutive sentences in the translation record, preserving the spatiotemporal consistency between the media content and the translation context. For users reviewing the translation record, it allows for a more intuitive understanding of the complete dialogue context corresponding to that long segment of media, thereby improving the information completeness and readability of the translation record.

[0169] In some embodiments of this application, a composite structure of translation records is further provided, specifically including the source text, translated text, and audio data of the translation session. After the media object is merged, it forms a composite record under the same record identifier with the source text, translated text, and audio data.

[0170] The record identifier serves as a unique index and a key primary key that links all output data of a specific translation session. It is assigned at the beginning of the translation session and persists throughout the entire lifecycle of the translation session, serving as a core association field in all subsequent data interactions.

[0171] The source text refers to the original language text content acquired through a head-mounted acquisition device and processed for recognition; it corresponds to the original sentences heard or spoken by the user in the translation scenario. The target text refers to the target language text content generated after machine translation or human-assisted processing of the source text, carrying the core information of the translation result. Audio data refers to the original speech stream acquired during the translation process, as well as the corresponding synthesized speech broadcast stream, used to record the real sound scene and machine pronunciation.

[0172] Compared to traditional data processing methods that store source text, translated text, and audio data as separate files or database entries, this application's embodiments solve the problem of fragile logical relationships caused by relying on vague timestamps or file names by introducing a unified record identifier. In traditional methods, audio files, text data, and media objects are stored in different directories or applications, making it easy to lose their correspondence during data migration, renaming, or cross-device synchronization. This application's embodiments, however, logically bind these heterogeneous data under the same entity, ensuring the stability of data relationships.

[0173] A composite record is a complete data structure that aggregates source text, translated text, audio data, and media objects, with the record identifier as the unique root node.

[0174] Once media objects are merged, they no longer exist as independent, discrete files. Instead, they are mounted into the composite record structure pointed to by the record identifier by updating the reference relationships in their associated media anchor objects. When a user views a translation record, the system can retrieve all source text, translation, audio playback links, and associated photo or video thumbnails for that session by querying the record identifier. This organization ensures that whether viewed locally, synchronized in the cloud, or exported and shared, the source text, translation text, audio data, and media objects are always processed as a whole unit, thus maintaining the integrity of the translation context and the consistency of the data.

[0175] In some embodiments of this application, for photo-type media objects, a record-bound media set can be formed under the same record identifier. When the record-bound media set triggers a return at different import entry points, it performs unified merging and deduplication through the same anchor point identifier and record identifier.

[0176] Media anchor objects contain anchor identifiers used for unique indexing. During the media object generation phase, the control terminal generates this anchor identifier, which serves as the basis for locating specific media objects during subsequent data flow and merging. All photo-type media objects belonging to the same record identifier constitute a record-bound media set. When the head-mounted acquisition device acquires photo-type media objects during a translation session, regardless of the stage of the session at which the acquisition occurs, the control terminal assigns the media objects to the corresponding record-bound media set based on the corresponding record identifier. For example, in a business meeting translation scenario, three consecutive photos taken by a user, although acquired at different times, logically constitute a single record-bound media set because they share the same record identifier. During the media anchor object creation phase, the control terminal establishes the hierarchical relationship between media objects and translation records at the data structure level by setting the anchor identifier and record identifier as a composite primary key for indexing.

[0177] For the return processing of record-bound media sets, the system supports triggering through multiple operation entry points, including the translation session end page, the historical translation record details page, and the media import interface in the device management page. Since the record-bound media sets have established a definite logical association through anchor identifiers and record identifiers, the merging logic executed in the control terminal's backend remains consistent regardless of the user's entry point for initiating the return command. In response to the return command, the control terminal obtains the record identifier based on the current context and queries all record-bound media sets under that identifier whose status is pending return. During the merging operation, the system uses the anchor identifier as the uniqueness criterion. If the anchor identifier of the media object to be returned already exists in the target translation record, it is determined to be a duplicate, only one media file is retained, its reference relationship is updated, and the duplicate data stream is discarded. This unified merging and deduplication mechanism based on anchor identifiers and record identifiers ensures the uniqueness and determinism of the media object's location in the database, effectively eliminating data dispersion or duplicate insertion problems caused by differences in trigger entry points.

[0178] In some embodiments of this application, an optional implementation of the aforementioned step S120, receiving the original file of the media object transmitted back by the head-mounted acquisition device via the second wireless link, is further described, specifically:

[0179] Preview data of media objects with a size lower than a first preset threshold are received preferentially via a second wireless link, while the original files of the media objects with a size higher than the second preset threshold are received later.

[0180] The preview data and the original file are associated with the same anchor point, and the preview data is replaced with the original file after the original file is received.

[0181] In this embodiment, the method implements a tiered transmission strategy based on the differences in data size of media objects during media return transmission. The size of the media object directly determines its transmission time and network resource consumption. Data versions that meet the needs of rapid display are defined as preview data, which includes, but are not limited to, low-resolution thumbnails of media objects, keyframe extracted images, or simplified audio summaries; the data size is typically compressed to tens to hundreds of KB. High-definition, complete versions without quality degradation are defined as original files, such as high-resolution photos, uncompressed audio, or high-definition video streams; the data size often reaches several MB or even larger.

[0182] The first and second preset thresholds are used to define the volume boundaries of the two types of data mentioned above. In actual operation, the first and second preset thresholds can be dynamically adjusted according to the specific network environment (such as WiFi (Wireless Fidelity) signal strength). For example, when the signal is good, the threshold can be appropriately increased to allow slightly larger preview data to be transmitted first, and when the signal is weak, the threshold can be decreased to strictly limit the priority transmission queue.

[0183] When the backhaul decision module determines that the second wireless link can be enabled, it scans the volume attributes of each media object in the backhaul queue. For media objects with a volume lower than a first preset threshold, they are marked as high-priority - immediate transmission, and a transmission request is immediately initiated through the second wireless link. This priority transmission mechanism allows the control terminal to quickly obtain a visual overview or auditory summary of the media object, thereby quickly generating a record placeholder on the user interface and providing the user with immediate feedback that the request has been received. At the same time, for original files with a volume higher than the second preset threshold, they are marked as low-priority - delayed transmission, and their transmission tasks are temporarily suspended, waiting for the high-priority queue to be cleared or the system to be in a more idle transmission period (such as the silent period after the user stops operating) before scheduling.

[0184] Compared to conventional first-in-first-out (FIFO) or filename-sorted transmission strategies, the aforementioned hierarchical transmission mechanism avoids the problem of large files blocking the network transmission queue and causing subsequent delays for smaller files. By breaking the strong binding between the physical transmission order and the file creation order, transmission resources are tilted towards small-volume data that can provide immediate feedback, thereby optimizing the user's perceived response speed within limited network bandwidth.

[0185] To ensure that preview data and subsequent arriving original files can be accurately identified as different versions of the same media object, a strong association is established between them using anchor identifiers. Anchor identifiers are unique identity fingerprints generated by the system when a media object is created.

[0186] During the transmission and storage phase, although the preview data and the original file are transmitted in a time-division multiplexing manner on the second wireless link and may temporarily exist independently in the control terminal's buffer, they share the same anchor point identifier. Specifically, when the preview data arrives, a temporary media entry is created at the corresponding position in the translation record (such as after a specific sentence reference) based on the anchor point identifier it carries, and the preview data is filled into the display area of ​​that entry. At this time, the anchor point identifier is recorded in the internal metadata of the media entry, and the current state is marked as preview awaiting replacement.

[0187] Subsequently, when the delayed original file arrives at the control terminal, a matching search is performed in the locally maintained media object index table by parsing the anchor identifier it carries. This search process involves traversing all media entries in the current translation record that are in the preview and awaiting replacement state, comparing the anchor identifier of each entry with the identifier of the original file. Once a match is found, the replacement execution logic is triggered. This replacement process uses atomic update operations: it keeps the original media entry position, segment reference relationship, and sort index unchanged, only updating the data source pointed to by the entry from the storage address of the preview data to the storage address of the original file, and updating the status of the media entry to "completed".

[0188] By employing an anchor-based association and replacement mechanism, a seamless transition from quick placeholders to full presentation is achieved. On one hand, this solves the problem of potential data misalignment during asynchronous transmission. Even if the original file experiences a long transmission delay or retransmission process on the network, it can still accurately locate and replace the previously left preview placeholders upon arrival, avoiding duplicate or misaligned media entries in the record. On the other hand, it eliminates the sense of interface blankness during the user's waiting process. Users can immediately see the complete record structure, including media thumbnails, after the translation session ends, while the original file is silently replaced in the background without interrupting the user's browsing flow.

[0189] In some embodiments of this application, the return control method may further include:

[0190] When the second wireless link is unavailable, the backhaul is interrupted, or the file verification fails, the backhaul status in the media anchor object is marked as pending backhaul or retrying, and the backhaul will continue to be received when the preset backhaul permission conditions are met.

[0191] The above processing flow establishes a failure compensation mechanism for the media return process. Through persistent management of the state machine, it ensures that media objects are not lost in the event of network fluctuations or abnormal interruptions, and that transmission can be automatically resumed when conditions permit, thereby guaranteeing the integrity of the translation record. The triggering conditions for this mechanism include detecting the unavailability of the second wireless link, unexpected interruption of the transmission connection, or the received original file failing verification (such as cyclic redundancy check, hash comparison, etc.).

[0192] Further optionally, before receiving the original file of the media object via the second wireless link in step S120, the following may also be included:

[0193] Query the return status in the media anchor object that matches the media object. If the return status is marked as completed, skip receiving; if the return status is marked as pending return or retrying, perform the operation of receiving the original file of the media object.

[0194] Specifically, a uniquely matching media anchor object can be retrieved from the record data corresponding to the current translation session using the media identifier. The return status field, which identifies the lifecycle stage, is then read from this media anchor object. This return status characterizes the processing progress of the media object in the current session; regardless of how many transmission requests are triggered, as long as the status indication is complete, the system can identify and intercept subsequent redundant operations. This mechanism effectively handles scenarios triggered by multiple entry points (such as users repeatedly clicking "import" on different pages) or automatic recovery after network anomalies and reconnections, preventing the system from getting stuck in an invalid data reception loop.

[0195] If the return status is marked as complete, skip receiving. When the return status field value is complete, it indicates that the original file corresponding to the media object has been successfully transmitted, verified, and merged via the second wireless link at a historical time. If the current receive request is determined to be a duplicate request, skip receiving.

[0196] If the return status is marked as pending return or retrying, the operation of receiving the original file of the media object will be performed. When the return status field value is pending return or retrying, it indicates that the media object has not been successfully merged or has a historical transmission failure record, and needs to be processed by this process.

[0197] Reference Figure 4 The example illustrates a data transmission link between a head-mounted acquisition device, a control terminal, and a translation server.

[0198] Establish a real-time translation session: The head-mounted acquisition device establishes a Bluetooth real-time translation link with the control terminal. The control terminal creates the current translation session object, including session identifier (session_id), record identifier (record_id), translation mode, and sentence / segment timeline.

[0199] Real-time translation data transmission: During the translation session, metadata of audio segments, control commands, and media objects is continuously exchanged via a Bluetooth real-time translation link. The control terminal can call the translation server for speech recognition (transcription), translation, and playback synthesis. Simultaneously, a sentence / segment timeline is generated based on the translation results, recording the start and end times of each segment, and this timeline is returned to the control terminal. The control terminal synchronously maintains the sentence / segment timeline and the current segment reference.

[0200] Media generation and local caching: When a user triggers a photo or generates additional media during the translation process, the head-mounted acquisition device first saves the original media file to the local cache area to be transmitted back, generates a media identifier, and reports lightweight metadata to the control terminal via Bluetooth. The lightweight metadata may include media type, acquisition timestamp, file summary, session identifier, and local file references.

[0201] Creating a media anchor object: The media anchor management module of the control terminal creates a media anchor object based on the reported lightweight metadata and the current segment timeline. This media anchor object is then bound to the current record_id, and its corresponding segment number, intra-segment offset position, and upload_state=pending (awaiting upload) status are recorded.

[0202] Backhaul Permission Determination: The backhaul decision module continuously acquires real-time Bluetooth link load, WiFi availability, battery level, temperature rise, whether the system is currently in active translation mode, and whether the user has actively triggered the import. Based on the backhaul permission score (RouteScore_t), it determines whether to allow the WiFi media backhaul link to be enabled. When the real-time translation risk exceeds the threshold, the system prohibits the original media object from being uploaded via Bluetooth.

[0203] Establish WiFi media backhaul link: When backhaul permission is granted, the control terminal sends a backhaul command to the head-mounted acquisition device. The head-mounted acquisition device activates the WiFi media backhaul link and transmits the media object to be backhauled to the control terminal in file or fragment format. The Bluetooth real-time translation link continues to maintain translation control and status synchronization.

[0204] Record merging and insertion: After receiving the returned media object, the record merging and deduplication module of the control terminal merges the media object into the original translation record based on the sentence reference and acquisition timestamp of the corresponding media anchor object. When the media is a photo during the translation process, it is preferred to insert it after the corresponding sentence segment, and keep the original text, translation, audio, and photo of the translation record as a composite object with the same record_id; when there are multiple photos to be returned under the same record_id, the photos are preferably merged and deduplicated uniformly according to the media set bound to the record.

[0205] Failure Compensation and Status Writeback: If WiFi is unavailable, the transmission is interrupted, file verification fails, or the user exits midway, the failure compensation module of the control terminal retains the media anchor object and the pending transmission status, updates the transmission status upload_state to pending, retrying, or failed, and continues transmission in the next transmission window or when the user re-enters the import entry; after successful transmission, the corresponding upload_state is updated to done.

[0206] The effectiveness of the media file return control method provided in this application for real-time translation scenarios will be explained next:

[0207] Traditional methods assign different QoS priorities or scheduling time slices to translated audio and media files on the same wireless link, allowing high-priority translation data to preempt. This approach fails in translation scenarios because media files are one to three orders of magnitude larger than a single translated frame. Even if downgraded to a lower priority, the segments already injected into the link still occupy buffering and retransmission opportunities, failing to guarantee zero yielding within active audio segments. This solution avoids scheduling within the same link. Instead, it physically isolates large media files to a second wireless link, with the first wireless link handling only low-latency translation transactions. A constraint is set that "latency-sensitive translation transactions must be physically isolated from large media files."

[0208] The traditional method is to upload the original file immediately after media is generated to ensure no loss. This approach fails in translation scenarios because media acquisition often coincides with active translation periods, and uploading would directly consume bandwidth. This solution reports only tens to hundreds of bytes of lightweight metadata via the first wireless link and caches the original file locally when the media is generated, delaying the return of the original file to the return window, thus setting a constraint of "decoupling the timing of media generation from the timing of return".

[0209] Traditional methods infer media ownership based on file system timestamps during post-meeting import. This approach fails in translation scenarios because pauses between sentences, rapid burst shooting, and device clock drift can cause mismatches between file timestamps and sentence timestamps, resulting in media being assigned to the wrong sentence or being removed from the recording. This solution creates a media anchor object carrying a media identifier and capture timestamp at the same moment the media is generated, moving the "ownership determination" forward to the moment of capture and directly determining it via the real-time sentence timeline. It also sets the constraint that "ownership information must be captured at the moment of generation, not reconstructed after the meeting."

[0210] Traditional methods save photos taken by head-mounted capture devices as regular photo album files, sorted chronologically. This approach fails in translation scenarios because the sentence-to-segment correspondences between photos, source text, translation, and audio are lost, making it impossible to create a retrievable composite translation record. This solution uses anchor point markers and record markers to incorporate media photos into the corresponding sentence-to-segment positions of the same composite translation record, establishing the constraint that "translation-related media are components of the record, not independent files."

[0211] Furthermore, considering that the availability of a second wireless link (such as WiFi) is independent of whether the translation is active, "available and transmit" may still occupy device wireless resources while users are still talking. This solution simultaneously considers the real-time activity status of the translation session and the communication status of the second wireless link to determine whether the preset return permission conditions are met. It encodes the constraint that "the return timing must give way to the active translation state" to ensure that the return of media files will not affect the real-time translation task.

[0212] Reference Figure 5 This application further describes the implementation process of a media file return control method from the perspective of a head-mounted acquisition device, which may include the following steps:

[0213] Step S300: Establish a first wireless link with the control terminal, and interact with the control terminal through the first wireless link to translate relevant data in real time during the translation session.

[0214] The head-mounted acquisition device first establishes a first wireless link with the control terminal. This link is used to carry real-time translation-related data interaction during the translation session. This data includes real-time acquired speech segments, translated text returned by the translation engine, real-time caption display instructions, and control signaling during the translation process. To ensure the real-time nature of the translation interaction, the first wireless link is configured as a communication channel with low latency characteristics. In specific implementations, this link can use Bluetooth or Bluetooth Low Energy protocols, or it can be configured as Ultra Wideband (UWB) or other proprietary low-latency wireless communication protocols depending on the actual scenario requirements. During the translation session, the first wireless link is kept active, enabling bidirectional data interaction, thereby supporting the control terminal's real-time monitoring and response to the translation process, and ensuring that the head-mounted acquisition device can receive translation results or control instructions from the control terminal in a timely manner.

[0215] Step S310: Acquire media objects during the translation session and cache the original files of the media objects locally.

[0216] While maintaining real-time translation data interaction via the first wireless link, media object acquisition is performed. Here, media objects refer to multimedia data generated by the head-mounted acquisition device in the translation scenario, such as photos taken by a camera, audio clips recorded by a microphone, or video clips. When a user-triggered media acquisition command is received, the sensor components of the head-mounted acquisition device acquire the raw media data and generate the corresponding raw file. Subsequently, this raw file is written to a preset cache area in the head-mounted device's local storage medium. This preset cache area is configured as non-volatile storage space, designed to save media content offline or in standby mode, preventing data loss due to network fluctuations or connection interruptions. By implementing a local caching strategy, the timing of large-volume media data transmission is decoupled from the timing of real-time translation data interaction, avoiding the consumption of bandwidth resources on the first wireless link by large file transmissions, thereby ensuring that the real-time performance of the translation session is not affected.

[0217] Step S320: In response to the feedback instruction issued by the control terminal, establish a second wireless link with the control terminal. The feedback instruction is generated by the control terminal after determining that the preset feedback permission conditions are met, based at least on the real-time activity status of the translation session and the communication status of the second wireless link.

[0218] The head-mounted acquisition device listens for and receives a feedback command from the control terminal via the first wireless link, thereby triggering the establishment of the second wireless link. This feedback command is a control signal generated by the control terminal based on a comprehensive judgment of preset feedback permission conditions. The control terminal comprehensively evaluates the real-time activity status of the translation session (such as whether there is active voice, whether the translation is being played, whether the translated segments are being processed, etc.) and the communication status of the second wireless link (such as signal strength, bandwidth availability, network connection stability, etc.), and only generates and sends the feedback command when it determines that the above status meets the preset feedback permission conditions. This passive response mechanism ensures that media feedback operations will not be initiated during critical translation activities or when the link status is poor, thereby guaranteeing the translation experience and transmission reliability.

[0219] The second wireless link is configured as a high-throughput communication channel for transmitting large media files. For example, this link can be a Wireless Fidelity (WiFi) link, including WiFi Direct mode or a LAN mode connected via an access point (AP), or, in specific scenarios, a cellular data link. The establishment process includes activating the corresponding high-bandwidth wireless communication module on the head-mounted acquisition device, performing link negotiation and authentication operations, until a stable data transmission connection is established with the control terminal. At this point, the first and second wireless links exist in parallel, respectively handling low-latency real-time communication and high-throughput media transmission.

[0220] Step S330: Send the original file of the media object to the control terminal via the second wireless link.

[0221] After the second wireless link is successfully established, the original file of the media object to be transmitted is read from the preset cache area of ​​the local storage medium. To ensure the integrity and traceability of data transmission, file retrieval can be performed based on the parameter information carried in the transmission command (such as the specified file identifier and transmission batch information). Subsequently, the read original file is sent to the control terminal via the second wireless link. For large original files, a fragmented transmission strategy is supported, that is, the large file is divided into multiple data fragments and sent sequentially, and then reassembled at the control terminal. During transmission, the connection status of the first wireless link is maintained to ensure that the interaction of real-time translation-related data is not interrupted due to the transmission of media files. Through this dual-link parallel working mode, the original file of the media object is quickly transmitted back via the high-throughput second wireless link, while the real-time translation task continues to be carried out by the low-latency first wireless link. The two do not interfere with each other, thereby achieving seamless transmission control of media files during the translation session.

[0222] In this embodiment, the head-mounted acquisition device and the control terminal handle real-time translation and media transmission tasks respectively through separate first and second wireless links, achieving rational scheduling of communication resources. The first wireless link ensures low latency and stability for translation interaction, while the second wireless link is established on demand under the control terminal's decision and is specifically used for batch transmission of large files. The head-mounted acquisition device, through a local caching strategy and a passive response mechanism to transmission commands, strictly cooperates with the control terminal's scheduling strategy to avoid resource conflicts that may be caused by the device's proactive transmission. This collaborative mechanism ensures that the transmission of media files is completely controlled by the real-time status and link quality of the translation session, thereby efficiently completing the archiving and transmission task of media objects while ensuring a smooth translation experience.

[0223] In some embodiments of this application, after acquiring the media object in the aforementioned step S310, the following steps may also be included:

[0224] Lightweight metadata of the media object is generated and sent to the control terminal via the first wireless link.

[0225] Lightweight metadata is used to trigger the control terminal to create a media anchor object that matches the media object. The media anchor object is used to locate the insertion position of the media object in the translation record.

[0226] After media object acquisition is complete, the head-mounted acquisition device extracts key attribute information of the media object to generate lightweight metadata. This lightweight metadata is not the original file of the media object itself, but rather a collection of information describing the basic characteristics of the media object. Its data size is typically in the tens to hundreds of bytes to ensure that it does not interfere with real-time translation data when transmitted over the low-bandwidth first wireless link. The lightweight metadata may include the media object's acquisition timestamp, media identifier, and identifier of the translation session to which it belongs. After generating the lightweight metadata, the head-mounted acquisition device sends it to the control terminal in real time using the established first wireless link. This process occurs after the original file of the media object is locally cached, aiming to notify the control terminal of the media object's creation event as soon as possible, allowing the control terminal to know the intention to insert the media object in advance without waiting for the large original file to be transmitted.

[0227] After receiving the lightweight metadata, the control terminal performs the media anchor object creation operation. The media anchor object is a logical index used to precisely locate the insertion position of a media object within the translation record structure. Its creation process relies on information such as the acquisition timestamp carried in the lightweight metadata. The control terminal generates the corresponding media anchor object based on the lightweight metadata. This object logically maps the media object's position on the timeline, thus completing the pre-positioning of the recording frame before the original file arrives. Through this lightweight metadata-based pre-positioning mechanism, the head-mounted acquisition device is responsible for acquisition and definition, while the control terminal is responsible for association and positioning. This clear division of labor ensures consistency in media merging logic between the two ends, resolving the media recording misalignment problem caused by large file transmission delays.

[0228] In some embodiments of this application, a differentiated transmission mechanism based on data volume grading is further provided. Specifically, the aforementioned step S330, which involves sending the original file of the media object to the control terminal via the second wireless link, may include:

[0229] Preview data of media objects with a size lower than the first preset threshold are sent first via the second wireless link, while the original files of media objects with a size higher than the second preset threshold are sent later.

[0230] When determining the transmission strategy for media objects, the data volume of the media object to be sent is first obtained and compared with a first preset threshold and a second preset threshold. The first preset threshold is used to define the range of small-volume data suitable for fast transmission, and the second preset threshold is used to define the range of large-volume data that requires traffic control.

[0231] For preview data media objects with a volume below a first preset threshold, a high transmission priority is assigned. The preview data is then instantly transmitted to the control terminal via a second wireless link, enabling the control terminal to quickly generate visual placeholder content in the translation record or interactive interface, thereby improving the perceived response speed for the user.

[0232] For raw file media objects whose size exceeds a second preset threshold, to avoid large file transfers instantly consuming significant network bandwidth or device processing resources, leading to communication congestion or system lag, the transmission operation for such files is delayed. Specifically, the transmission task for such raw files is suspended until the system is detected to be idle or other preset delayed transmission conditions are met, at which point the transmission process via the second wireless link is initiated.

[0233] The differentiated transmission mechanism based on data volume classification, as described above, enables staggered transmission of preview data and original files. This mechanism ensures the timely presentation of key visual information to optimize user experience while effectively mitigating the risk of large files blocking the transmission channel, thus balancing real-time performance and transmission efficiency.

[0234] This application also provides an electronic device in its embodiments. (See reference...) Figure 6 The diagram illustrates a structural schematic suitable for implementing the electronic device in the embodiments of this application. The electronic device in the embodiments of this application can be a control terminal of the aforementioned embodiments, such as a mobile phone or computer, or a head-mounted data acquisition device of the aforementioned embodiments, such as AI glasses. Figure 6 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0235] like Figure 6 As shown, the electronic device may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) 1, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 2 or a program loaded from a storage device 8 into a random access memory (RAM) 3, to implement the media file return control method in the real-time translation scenario of any of the preceding embodiments of this application. When the electronic device is powered on, the RAM 3 also stores various programs and data required for the operation of the electronic device. The processing unit 1, ROM 2, and RAM 3 are interconnected via a bus 4. An input / output (I / O) interface 5 is also connected to the bus 4.

[0236] Typically, the following devices can be connected to I / O interface 5: input devices 6 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 7 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 8 including, for example, memory cards, hard drives, etc.; and communication devices 9. Communication device 9 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 6 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown. More or fewer devices may be implemented or have alternatively.

[0237] This application also provides a computer program product, including computer-readable instructions, which, when executed on an electronic device, cause the electronic device to implement any of the media file return control methods provided in the real-time translation scenario of this application.

[0238] This application also provides a computer-readable storage medium that carries one or more computer programs. When the one or more computer programs are executed by an electronic device, the electronic device can implement any of the media file return control methods provided in the real-time translation scenario of this application.

[0239] It should also be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. In addition, in the device embodiment drawings provided in this application, the connection relationship between modules indicates that they have a communication connection, which can be implemented as one or more communication buses or signal lines.

[0240] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware, or it can be implemented by special-purpose hardware including application-specific integrated circuits, special-purpose CPUs, special-purpose memory, special-purpose components, etc. Generally, any function performed by a computer program can be easily implemented by corresponding hardware, and the specific hardware structure used to implement the same function can also be diverse, such as analog circuits, digital circuits, or special-purpose circuits. However, for this application, software program implementation is more often the preferred implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a readable storage medium, such as a computer floppy disk, USB flash drive, mobile hard disk, ROM, RAM, magnetic disk, or optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, training equipment, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0241] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product.

[0242] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, training device, or data center to another website, computer, training device, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can store or a data storage device such as a training device or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state drives (SSDs)).

[0243] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The various embodiments can be combined as needed, and the same or similar parts can be referred to each other.

Claims

1. A method for controlling the return of media files in a real-time translation scenario, characterized in that, A control terminal applied in conjunction with a head-mounted data acquisition device, the method comprising: A first wireless link is established with the head-mounted acquisition device, and relevant data is translated in real time by interacting with the head-mounted acquisition device through the first wireless link during the translation session; When there is a media object to be returned, at least based on the real-time activity status of the translation session and the communication status of the second wireless link, it is determined whether the preset return permission conditions are met, wherein the media object is the original media file that was acquired and locally cached by the head-mounted acquisition device during the translation session. If the preset return permission conditions are met, a second wireless link is established with the head-mounted acquisition device, and the original file of the media object returned by the head-mounted acquisition device is received through the second wireless link.

2. The method according to claim 1, characterized in that, The first wireless link is a Bluetooth link or a Bluetooth Low Energy link; the second wireless link is a WiFi link.

3. The method according to claim 1, characterized in that, The real-time activity status includes: whether there is active voice, whether a translation is being broadcast, and whether a translation segment is being processed.

4. The method according to claim 1, characterized in that, When determining whether the preset return permission conditions are met, in addition to considering the real-time activity status of the translation session and the communication status of the second wireless link, one or more of the following are also taken into account: Return window status, first wireless link load, and control terminal device status; The return window state includes at least one of the following: a post-meeting return window triggered after the translation session ends, an active import window triggered by the user, and a silent return window that satisfies the requirement that the real-time activity state is not in a state susceptible to return interference.

5. The method according to claim 4, characterized in that, The process of determining whether the preset return permission conditions are met includes: The backhaul permission score is calculated based on at least one of the following: the real-time activity status of the translation session, the communication status of the second wireless link, the backhaul window status, the load of the first wireless link, and the device status of the head-mounted acquisition device. When the backhaul permission score is higher than a preset threshold, it is determined that the preset backhaul permission conditions are met.

6. The method according to claim 5, characterized in that, The device status of the head-mounted data acquisition device includes battery level and / or battery temperature, wherein the battery level is positively correlated with the backhaul permission score and the battery temperature is negatively correlated with the backhaul permission score; The process of determining whether the preset return permission conditions are met also includes: When the battery level is lower than a preset battery level threshold, it is determined that the preset data transmission permission conditions are not met. This process continues until the head-mounted data acquisition device enters the charging state, at which point the data transmission permission score is calculated, and the data transmission permission score is used to determine whether the data transmission permission conditions are met.

7. The method according to claim 1, characterized in that, Before determining whether the preset return permission conditions are met, the following steps are also included: The device receives lightweight metadata of the media object uploaded by the head-mounted acquisition device through the first wireless link. The lightweight metadata is generated and uploaded by the head-mounted acquisition device after acquiring the media object. A media anchor object matching the media object is created based on the lightweight metadata. The media anchor object is used to locate the insertion position of the media object in the translation record. After receiving the original file of the media object from the head-mounted acquisition device, the original file of the media object is merged into the translation record based on the media anchor object matched by the media object.

8. The method according to claim 7, characterized in that, The lightweight metadata includes the acquisition timestamp and media identifier of the media object; The process of creating a media anchor object that matches the media object based on the lightweight metadata includes: Based on the acquisition timestamp and the current segment timeline maintained locally by the control terminal, the segment number where the acquisition timestamp is located is determined and used as the segment reference of the media object; Create a media anchor object that includes an anchor identifier, the acquisition timestamp, the media identifier, the segment reference, and the return status, wherein the return status is marked as pending return.

9. The method according to claim 8, characterized in that, The process of merging the original file of the media object into the translation record based on the media anchor object matched by the media object includes: Retrieve the segment reference from the media anchor object that matches the media object; If the segment reference represents the j-th translated segment, then the original file of the media object is inserted after the j-th translated segment in the translation record, and the return status in the media anchor object is updated to complete; If the segment reference represents the space between the j-th and j+1-th translated segments, then the original file of the media object is inserted into the translation record between the j-th and j+1-th translated segments, and the postback status in the media anchor object is updated to complete.

10. The method according to claim 8, characterized in that, The media anchor object also includes at least one of the following: Session identifier, record identifier, start and end timestamps of the translated segment corresponding to the segment referenced, and offset position within the sentence; The session identifier is used to identify the current translation session, the record identifier is used to identify the current translation record, and the sentence offset position indicates the relative position of the acquisition timestamp within the start and end time interval of the translation segment corresponding to the sentence segment reference.

11. The method according to claim 8, characterized in that, When the media object is an audio or video clip that covers two or more translated sentence segments, the acquisition timestamp is the start and end timestamp of the acquisition of the audio or video clip; The process of determining the segment number of the acquired timestamp as the segment reference of the media object based on the acquired timestamp and the current segment timeline maintained locally by the control terminal includes: Based on the start and end timestamps of the data collection and the current segment timeline maintained locally by the control terminal, the starting segment number of the data collection start timestamp and the ending segment number of the data collection end timestamp are determined, and the starting segment number and the ending segment number are used as segment references for the media object. The process of merging the original file of the media object into the translation record based on the media anchor object matched by the media object includes: Based on the media anchor object matched by the media object, the original files of the audio or video clips are merged into the translation record, and the segment intervals corresponding to the starting segment number and the ending segment number are defined.

12. The method according to claim 7, characterized in that, The media anchor object includes a record identifier, which is used to identify the current translation record; The translation record corresponding to the translation session contains the source text, translated text, and audio data of the translation session. After the media object is merged, it forms a composite record under the same record identifier with the source text, translated text, and audio data.

13. The method according to claim 12, characterized in that, The media anchor object also includes an anchor identifier. Photo media objects form a record-bound media collection under the same record identifier. When the record-bound media collection triggers a return at different import entry points, it performs unified merging and deduplication through the same anchor identifier and record identifier.

14. The method according to claim 8, characterized in that, The process of receiving the original file of the media object transmitted back by the head-mounted acquisition device via the second wireless link includes: The second wireless link prioritizes receiving preview data of media objects with a size lower than a first preset threshold, and delays receiving the original file of media objects with a size higher than the second preset threshold. The preview data and the original file are associated with the same anchor point identifier, and the preview data is replaced with the original file after the original file is received.

15. The method according to claim 8, characterized in that, Also includes: When the second wireless link is unavailable, the backhaul is interrupted, or the file verification fails, the backhaul status in the media anchor object is marked as pending backhaul or retrying, and the backhaul will continue to be received when the preset backhaul permission conditions are met in the future.

16. The method according to claim 15, characterized in that, Before receiving the original file of the media object via the second wireless link, the method further includes: Query the return status in the media anchor object that matches the media object. If the return status is marked as completed, skip receiving; if the return status is marked as pending return or retrying, perform the operation of receiving the original file of the media object.

17. A method for controlling the return of media files in a real-time translation scenario, characterized in that, A head-mounted data acquisition device applied in conjunction with a control terminal, the method comprising: A first wireless link is established with the control terminal, and relevant data is translated in real time by interacting with the control terminal through the first wireless link during the translation session; During the translation session, media objects are captured, and the original files of the media objects are cached locally. In response to the feedback instruction issued by the control terminal, a second wireless link is established with the control terminal, wherein the feedback instruction is generated by the control terminal after determining that the preset feedback permission conditions are met based at least on the real-time activity status of the translation session and the communication status of the second wireless link; The original file of the media object is sent to the control terminal via the second wireless link.

18. The method according to claim 17, characterized in that, After acquiring the media object, the following is also included: Lightweight metadata of the media object is generated and sent to the control terminal via the first wireless link. The lightweight metadata is used to trigger the control terminal to establish a media anchor object that matches the media object. The media anchor object is used to locate the insertion position of the media object in the translation record.

19. The method according to claim 17, characterized in that, The process of sending the original file of the media object to the control terminal via the second wireless link includes: Preview data of the media object with a size lower than a first preset threshold is sent first via the second wireless link, while the original file of the media object with a size higher than a second preset threshold is sent later.

20. A control terminal, characterized in that, include: Memory and processor; The memory is used to store programs; The processor is used to execute the program to implement each step of the media file return control method in the real-time translation scenario as described in any one of claims 1 to 16.

21. A head-mounted data acquisition device, characterized in that, Includes a multimedia acquisition module and playback model, a communication module, a memory, and a processor; The memory is used to store programs; The processor is used to execute the program to implement each step of the media file return control method in the real-time translation scenario as described in any one of claims 17 to 19.