Human experience multi-media story narration method based on time axis

Abstract

The invention discloses a life experience story narration method based on a time axis, belongs to the technical field of information, and aims to solve the problems that the existing multi-media material integration is disordered, the time axis lacks semantic association, the privacy compliance is insufficient and the evidence is difficult to iterate. The method comprises the following steps: presetting a cross-unit narrative main line and dividing life stage units; multi-media materials such as images, audios and characters are directionally collected and embedded into privacy compliance processing; a main chain-branch chain type evidence is constructed through a credible timestamp; multi-medium collaborative narration is achieved according to the unit theme-scene fragment-timestamp anchor point; and outputting polymorphic works after three-dimensional verification, and supporting dynamic iteration of newly added materials. According to the method, the time axis is upgraded into a'time-event-medium 'three-dimensional semantic model, structured narration of life experience is realized, privacy security and evidence storage reliability are considered, and the method can be applied to scenes such as personal memory management, family culture inheritance and cultural heritage digitization and has high expansibility and practicability.

Description

Technical Field

[0001] This invention relates to the field of information technology, specifically to a timeline-based multi-media storytelling method and system for life experiences. It integrates multi-media materials such as images, audio, text, pictures, music, and art, and systematically narrates stories along a timeline. This method is particularly suitable for scenarios such as personal memory management, cultural heritage digitization, and multimedia content creation. This invention integrates technologies such as artificial intelligence (AI), blockchain notarization, and multimodal data processing, solving problems such as chaotic integration of multi-media materials, lack of systematic narrative, and insufficient privacy protection. It achieves dynamic, structured, and traceable recording of life experiences. (See appendix to the specification) Figure 1 The system uses a three-dimensional coordinate system (X-axis: timestamp, Y-axis: event network, Z-axis: multiple media types) to display a semantic association model of "time-event-media". It labels the Y-axis with "automatic life stage division" and "emotional tag association," and the Z-axis with "image-dynamic scene" and "music-emotional enhancement," clearly demonstrating the innovative aspects of the timeline that break through traditional sorting functions and construct an "event network," aiding in understanding the specific implementation of semantic modeling. (T0: Data acquisition starts, T1: Image desensitization complete, T2: Evidence archiving, T3: AI icon annotation complete, T4: Audio voice changing starts, T5: AI sound effect annotation complete, T6: Evidence archiving complete) See the attached diagram in the instruction manual. Figure 1 (See attached figure for the abstract.) Background Technology

[0002] The limitations of existing technologies and the monotony of single-media narratives: Traditional methods of recording personal memories (such as photo albums, diaries, and video clips) rely heavily on a single medium, making it difficult to fully recreate the complexity of life experiences. For example, photos can only capture a moment, textual descriptions lack intuitiveness, and while videos are dynamic, they struggle to achieve deep connections between multi-dimensional information. Existing multi-media integration tools (such as Adobe Premiere and Avid Media Composer) support material splicing, but require significant manual operation and lack a systematic storytelling logic, resulting in fragmented narratives and weak emotional expression.

[0003] Functional limitations of timeline tools: Existing timeline tools (such as Timeline JS and Gantt charts) are mainly used for task management or event sequencing, lacking semantic modeling capabilities. For example, Gantt charts only focus on the time dimension and cannot link causal relationships or emotional contexts between events; traditional timeline tools cannot automatically divide life stages and struggle to dynamically link multimedia materials. Furthermore, the combination of timelines and multimedia in existing technologies is mostly a simple overlay, failing to form a closed-loop logic of "timeline-media-narrative." (Note: Timeline JS is an open-source timeline creation tool, primarily used for lightweight timeline display on web pages.)

[0004] Insufficient Privacy Protection and Data Compliance: Multimedia material collection involves a large amount of sensitive personal information (such as portraits, addresses, and voices). Existing privacy processing methods (such as pixel blurring and clip editing) are mostly localized operations, lacking a systematic compliance framework. For example, AI-generated images may face legal risks due to infringement of training data, and audio anonymization may disrupt the immersive experience of the narrative. Furthermore, while existing Personal Information Management Systems (PIMS) provide basic privacy protection, they lack dynamic compliance processes designed for multimedia narrative scenarios. Privacy protection algorithms commonly used in cross-media analytics (such as K-anonymity and l-diversity) reduce the risk of information leakage through data anonymization, but existing technologies only apply them to localized processing of a single medium (such as pixel blurring only for images), failing to deeply integrate them with the entire "collection-processing-storage" process of multimedia narratives (e.g., K-anonymity is not adapted to scenarios such as scene reproduction in video and environmental sound effect generation in audio, leading to conflicts in privacy processing rules across different media). (Note: K-anonymity and l-diversity are commonly used privacy protection algorithms in cross-media analytics that reduce the risk of information leakage through data anonymization).

[0005] Challenges of Data Preservation and Dynamic Updates: Life experience records need to be preserved long-term and dynamically updated, but existing preservation technologies (such as traditional file storage and centralized databases) suffer from tampering risks and insufficient scalability. For example, paper photos are easily damaged, and digital files may be unreadable due to outdated formats; while blockchain preservation is tamper-proof, existing solutions mostly focus on a single medium (such as text or images), lacking a multi-media collaborative preservation mechanism. Furthermore, existing technologies struggle to seamlessly integrate new materials with historical data, compromising the authenticity and integrity of the timeline.

[0006] Attempts to improve existing technologies and the application of multimodal AI technology: In recent years, multimodal AI technologies (such as OpenAI's CLIP and DALL·E) have made progress in the field of cross-media understanding and generation, enabling semantic association between text, images, and audio. For example, the Beijing Economic-Technological Development Area Media Convergence Center uses a multimodal audiovisual large model to generate high-precision videos, improving production efficiency through AI super-resolution and horizontal / vertical screen synchronization technologies. However, such technologies mainly serve professional content creation and lack a systematic narrative framework for personal life experiences.

[0007] Innovative Explorations in Blockchain-Based Evidence Preservation: The application of blockchain technology in copyright protection (such as the Cover Blockchain Digital Content Evidence Preservation System) has verified its immutability. Institutions such as China Merchants Bank and Sichuan Xinwang Bank have used blockchain to achieve legal evidence preservation and transaction information management, solving the problems of difficulty in obtaining evidence and data forgery. However, existing blockchain evidence preservation solutions are mostly aimed at single business scenarios and have not formed a closed loop of the entire process of "material collection-processing-evidence preservation-update", especially lacking a cross-verification mechanism for multi-media materials.

[0008] The Evolution of Privacy Protection Technologies: Privacy protection technologies in cross-media analytics (such as K-anonymity and l-diversity) reduce the risk of personal information leakage through data anonymization. In existing technologies, companies like Hikvision have explored multimodal anonymization solutions (such as dynamically adjusting anonymization strategies to improve data security, see patent CN119808161A). However, these technologies focus on optimizing the quality of data anonymization and do not adapt to the "scenario-specific needs" of multi-media narratives (e.g., the K-anonymity algorithm indiscriminately blurs iconic scene elements like "old alleyway entrance," disrupting narrative immersion and causing a disconnect between compliance operations and content creation). Furthermore, existing technologies have not yet deeply integrated privacy processing into the multi-media narrative process, resulting in a disconnect between compliance operations and content creation.

[0009] The innovative approach of this invention lies in its systematic storytelling framework: by pre-setting cross-unit narrative threads, dividing life stages into units, and constructing a multi-media division of labor system, this invention integrates scattered materials into a logically coherent story chain, solving the problem of "having materials but no narrative" in traditional methods. For example, the core theme (such as "self-reconciliation during growth") can drive the collaborative expression of media such as images, music, and art, forming a narrative thread with progressive emotions.

[0010] Semantic Modeling of Timelines: The timeline proposed in this invention is not only a time sorting tool, but also an "event relationship network." Through timestamp chain storage, semantic association analysis, and dynamic editing functions, the timeline can automatically divide life stages, mark key nodes, and associate timestamps and sentiment tags of multi-media materials to achieve a three-dimensional mapping of "time-event-media".

[0011] Multi-media privacy compliance throughout the entire process: This invention embeds privacy protection and copyright verification mechanisms at each stage of material collection, processing, and storage, including AI-based fuzzy spatiotemporal scene algorithms, compliance verification of training data, and layered processing of sensitive information, forming a privacy protection network covering the entire lifecycle. For example, AI-generated images must be labeled "AI restoration" and the training data must be verified, while hand-drawn images must ensure originality to avoid infringement risks.

[0012] Dynamic Evidence Preservation and Iterative Updates: This invention employs a trusted timestamp evidence preservation tool to cross-verify the entire process of multi-media materials (such as original files, processing records, and copyright authorization), and supports secondary processing and supplementary evidence preservation for newly added materials. For example, newly added photos need to have their privacy occlusions removed and independent timestamps generated to ensure the authenticity and scalability of the timeline.

[0013] Comparison of features between existing technologies (such as ordinary log systems) and the technical solution of this invention:

[0014] Regarding timestamp usage: Existing technologies only record operation time without semantic association. The present invention's solution, however, can associate with privacy-compliant operations (anonymization, labeling). Regarding data dimensions: Existing technologies have only a single time dimension. The present invention's solution provides time + semantic tags + algorithm parameters. Regarding compliance adaptation: Existing technologies lack alignment with privacy regulations. The present invention's solution complies with K-anonymity and GDPR Article 25. Regarding anomaly handling: Existing technologies are missing or simply report errors. The present invention's solution can perform time-series verification, calculation, and labeling.

[0015] In summary, this invention, by integrating multi-media technology, timeline semantic modeling, privacy compliance framework, and dynamic evidence storage mechanism, breaks through the limitations of existing technologies and provides a brand-new solution for personal memory management and multimedia narrative. Summary of the Invention

[0016] This invention aims to solve four core problems existing in the prior art:

[0017] (1) The integration of multi-media materials is chaotic and lacks a systematic storytelling framework, resulting in fragmented records of life experiences and weak emotional expression;

[0018] (2) The timeline tool can only sort events, but cannot associate the semantic relationships between events with multimedia materials, thus failing to form a closed loop of "time-event-media";

[0019] (3) The lack of a full-process mechanism for multi-media privacy protection and data compliance can easily lead to infringement risks or damage the immersive experience of the narrative.

[0020] (4) The lack of coordination in the storage of multi-media materials and the difficulty in seamlessly connecting new materials with historical data result in damage to the authenticity and integrity of the timeline.

[0021] To address the aforementioned issues, this invention provides a timeline-based multi-media storytelling method for recording life experiences. Through a comprehensive design encompassing "main story guidance - unit division - compliant data collection - chain-based evidence storage - collaborative narrative - verification output - dynamic iteration," it achieves structured, authentic, and traceable recording of life experiences (see appendix to the specification). Figure 2This document demonstrates the seven core stages and logical loop of this invention: "mainline guidance - unit division - compliant data collection - chain-based evidence storage - collaborative narrative - verification output - dynamic iteration." It annotates key actions in each stage (e.g., "compliant data collection" subdivides into image and audio media processing), intuitively showcasing the completeness of the technical solution throughout the entire process and aiding in understanding the overall logic of multi-stage collaborative storytelling of life experiences. This invention is the first to deeply integrate multi-media privacy processing (image + audio), the K-anonymity algorithm, timeline semantic modeling, and blockchain evidence storage, forming an end-to-end privacy compliance loop, unlike existing technologies that focus only on a single stage (such as only image desensitization or only timestamp recording).

[0022] Technical solution

[0023] Step 1: Pre-set cross-unit narrative main line

[0024] Before dividing life into stages, a core narrative thread that runs through the entire series should be determined (such as "self-reconciliation during growth" and "persistence in ordinary life") to ensure that each stage unit can stand alone as a story, while also forming a "systematic story chain" through the main thread. The main thread should be adapted to the coordinated expression of six major media carriers: video, audio, text, pictures, music, and art, and a unified emotional tone and thematic anchor point should be set for each media carrier (such as "repeated action close-ups" in video, "progressive melody" in music, and "key node freeze-frames" in pictures corresponding to the "persistence" main thread).

[0025] Step 2: Divide the narrative into units according to life stages

[0026] Based on the core theme, life is divided into several stages (e.g., "childhood → youth → middle age"). Each stage sets a core theme that echoes the main theme, and the core theme must be anchored to key life experiences that actually occurred in that stage (e.g., "first independence" corresponds to the real experience of "going to school alone in another province at age 18", "career transition" corresponds to the real process of "from resigning to starting a business"). Each stage must clearly define the division of labor among the six media carriers: video is responsible for dynamic scene reproduction, images are responsible for static moment anchoring, audio is responsible for environmental and dialogue documentation, text is responsible for logical connection, music is responsible for emotional enhancement, and art is responsible for scene supplementation and symbolic expression, ensuring that the theme originates from experience and the media carriers serve reality.

[0027] Step 3: Algorithm Logic for Semantic Timeline Modeling Three key elements of the algorithm:

[0028] Inputs: Acquisition timestamp (e.g., T0: Acquisition started), processing events (e.g., T1: Image desensitization completed), evidence storage markers (e.g., T2: Evidence storage archived); Time-axis semantic modeling sub-flowchart (graph TD)

[0029] A [Raw timestamp input] --> B [Structured sorting]

[0030] B-->C [Rule Engine Mapping Tag]

[0031] C --> D [LSTM model correction]

[0032] D-->E [Timing Check]

[0033] E-->F {Validation passed?}

[0034] F --> |Yes|G [Generates a semantic timeline]

[0035] F --> |No |H [Mark exception and log]

[0036] Processing logic:

[0037] Timeline construction: A linear sequence is generated using a timestamp sorting algorithm (formula: Time_chain = Sort([T0,T1,T2],by = timestamp ascending order));

[0038] Semantic modeling: A rule engine and LSTM model are used to extract semantic tags (such as "collection", "de-identification", "evidence storage") model structure. Output: Timeline data with semantic tags (such as JSON format: {"T0":"Collection started","T1":"Image de-identification completed"}), which are bound to the evidence storage data. Supplementary Implementation Details: Taking "a certain video data processing" as an example, the timeline construction process is described in detail: "Implementation 1: ① Acquisition Stage: T0 = 2025-08-25 10:00:00 Acquisition started, T1 = 2025-08-25 10:00:10 Acquisition completed (verification passed); ② Processing Stage: T2 = 2025-08-25 10:00:15 Image desensitization started, T3 = 2025-08-25 10:00:20 AI icon annotation completed; T4 = 2025-08-25 10:00:25 Audio voice changing started, T5 = 2025-08-25 10:00:30 AI sound effect annotation completed; ③ Evidence Preservation Stage: T6 = 2025-08-25..." At 10:00:35, a timeline [T0, T1, T2, T3, T4, T5, T6] is generated. After semantic modeling, it becomes: {"T0":"Collection Started","T1":"Collection Completed (Training Data Verification Passed)","T2":"Image Desensitization Started (K-anonymity, k=5)","T3":"AI Icon Annotation Completed (Marking Blurred Areas on Documents)","T4":"Audio Voice Changing Started (Base Frequency + 150Hz)","T5":"AI Sound Effect Annotation Completed (Marking Voice Changing Range 00:00:10-00:00:15)","T6":"Evidence Storage Completed (Associating K-anonymity Parameters)"}, and stored in a distributed evidence storage system. Timestamp data format:

[0039] The ISO 8601 standard format is adopted (e.g., YYYY-MM-DD HH:MM:SS.fff, accurate to milliseconds), which includes the start / end time of each stage of acquisition, processing, and evidence storage (e.g., T0_start=2025-08-25 10:00:00.000, T0_end=2025-08-25 10:00:10.500 indicates the start and end of the acquisition stage).

[0040] The "event subject" associated with the timestamp: This indicates the operation module corresponding to the timestamp (e.g., "data acquisition unit," "privacy processing unit," "AI annotation branch"). Event attribute data:

[0041] Each timestamp is bound to 3 types of key attributes (using JSON structure as an example): {

[0042] "timestamp":"2025-08-25 10:00:15.200", / / timestamp

[0043] "module":"Privacy Processing Unit", / / Execution Module

[0044] "operation":"Image Desensitization", / / Operation type

[0045] "parameters":{"algorithm":"K-anonymity","k":5,"blur_radius":5}, / / operation parameters

[0046] "result":"success" / / Operation result (success / fail)

[0047] Step 1: Timeline Structured Sort

[0048] Objective: To arrange discrete timestamps in chronological order to form a linear time chain.

[0049] Algorithm logic:

[0050] Extract the millisecond values ​​of all timestamps (e.g., convert T0_start to a timestamp integer 1756239600000);

[0051] Arrange the values ​​in ascending order using a stable sorting algorithm (such as merge sort), with the following formula:

[0052] TimeChain = MergeSort([T0,T1,...,Tn]) (where Ti is the millisecond value of the i-th timestamp, and the sorting ensures that T0 is equal to the original value). <T1<...<Tn);

[0053] Filter duplicate timestamps (keep the earliest appearing record). Example:

[0054] Original timestamp set [T2=10:00:15, T1=10:00:10, T3=10:00:20] → Sorted [T1, T2, T3]. Step 2: Semantic tag system construction.

[0055] Objective: Define standardized labels to transform "operation type + module + result" into machine-recognizable semantic symbols.

[0056] Tag hierarchy design (three-level classification):

[0057] Level 1 Tag (Process Stage): Collection, Processing, Evidence Storage;

[0058] Second-level label (operation type):

[0059] Data collection phase: Startup, Completion, Verification Passed, Verification Failed;

[0060] Processing stages: image desensitization, audio voice changing, AI annotation;

[0061] Evidence preservation stage: archiving, associating with timelines, and recording parameters;

[0062] Level 3 tags (details supplemented): such as image desensitization associated with K-anonymity, blur radius 5px, and other parameter tags.

[0063] Step 3: Semantic Mapping Algorithm (Core)

[0064] Objective: To automatically map "operational data" in a structured timeline into semantic tags.

[0065] Technical approach: Employing a hybrid architecture of "rule engine + machine learning model":

[0066] Rule engine (handles deterministic mappings):

[0067] Preset mapping rules (taking "Operation Type → Second-level Tag" as an example): #Rule Example

[0068] If operation == "image blurring processing" and module == "privacy processing unit":

[0069] semantic_tag="processing.image desensitization"

[0070] If operation == "voice changing processing" and module == "privacy processing unit":

[0071] semantic_tag="processing audio voice changing" LSTM model (for handling blurred scenes):

[0072] For complex operations (such as "abnormal AI annotation results"), use an LSTM model to learn historical label mapping patterns:

[0073] Input: Word vectors (300 dimensions) of the operation description text (e.g., "AI icon annotates unrecognized document areas");

[0074] Model structure: 2-layer LSTM (64 hidden nodes) + Softmax output layer (outputs the probability of three-level labels);

[0075] Training data: 100,000 manually labeled "operation-label" samples, deployed with an accuracy of ≥95%.

[0076] Step 4: Timing Consistency Verification

[0077] Objective: To ensure the logical coherence of the semantic timeline (e.g., "proof" must be after "processing completed").

[0078] Verification rules:

[0079] Phase sequence verification: data collection → processing → evidence storage (the next phase cannot be started until the previous phase is completed);

[0080] Operation dependency validation: For example, "AI annotation" must be performed after "image desensitization" (through label dependency: processing.AI annotation depends on processing.image desensitization);

[0081] Anomaly Marking: If the rules are violated, insert a "[Timing Anomaly]" tag in the timeline and record the reason for the anomaly (e.g., T5: Evidence storage started before T4: Processing completed → mark as "[Evidence storage precedes processing]"). Output Format and Related Logic (Output Layer Refinement)

[0082] Define the final form of the semantic timeline and its relationship with other modules:

[0083] Output format:

[0084] The JSON structure uses "timestamp + multi-level tags + parameter snapshot", example: {

[0085] "timeline_id":"TL20250825001", / / Unique identifier for the timeline

[0086] "semantic_chain":[

[0087] {

[0088] "timestamp":"2025-08-25 10:00:00.000",

[0089] "tags":["Collection.Start"],

[0090] "parameters":{"source":"Camera A","data_type":"Video"}

[0091] },

[0092] {

[0093] "timestamp":"2025-08-25 10:00:10.500",

[0094] "tags":["Collection Complete","Collection Verification Passed"],

[0095] "parameters":{"validity":98.5} / / Validation pass rate

[0096] },

[0097] {

[0098] "timestamp":"2025-08-25 10:00:15.200",

[0099] "tags":["Processing.Image Desensitization","K-anonymity(k=5)"],

[0100] "parameters":{"blur_radius":5,"face_mask":"success"}

[0101] }

[0102] ],

[0103] "status":"valid" / / Timing verification result

[0104] Relationship with the evidence storage module:

[0105] The semantic timeline's `timeline_id` is bound to the `evidence_id` of the stored evidence data. An index is built using a distributed database (such as MongoDB) to support bidirectional traceability between data and timeline (e.g., when querying stored evidence data, its complete semantic timeline can be retrieved directly). Missing timestamps:

[0106] If a timestamp for a certain stage is missing (e.g., "processing completed" is not recorded), it is estimated by the timestamps before and after (e.g., T processing completed ≈ T processing started + average processing time), and marked as [estimated value];

[0107] Tag conflict:

[0108] If the rule engine and the LSTM model output labels conflict (e.g., the rule engine outputs "Process. Image Desensitization", while the model outputs "Process. Audio Voice Changing"), the rule engine result shall prevail (human-preset rules have higher priority), and a conflict log shall be recorded for model iteration.

[0109] High-concurrency scenarios:

[0110] When multiple threads generate timestamps simultaneously, a distributed lock (such as a Redis lock) is used to ensure the uniqueness of the timestamps and avoid disordered sorting. Example and effect verification (reproducibility detailed).

[0111] The algorithm's entire process is demonstrated through specific examples, along with performance data:

[0112] Example: Input data for semantic modeling of the timeline of a street view video (5 raw timestamps):

[0113] Results of operation parameters for the timestamp (ISO format) module

[0114] 2025-08-25 10:00:00.000 Data acquisition unit started acquiring data source=camera B success

[0115] 2025-08-25 10:00:12:30 Validation of branch training data: Validity = 99.2% success

[0116] 2025-08-25 10:00:15.100 Privacy processing unit Image blur algorithm = K - anonymity, k = 5 success

[0117] 2025-08-25 10:00:20.700 AI annotation for branch generation icon label = document area success

[0118] 2025-08-25 10:00:25.400 Evidence storage unit archiving storage_path= / disk1 / evidencesuccess Algorithm processing:

[0119] Step 1: The sorted time chain is [T0,T1,T2,T3,T4];

[0120] Step 2: Match the label (e.g., T2 is mapped to ["Processing.Image Desensitization", "K-anonymity(k=5)"]);

[0121] Step 3: Timing verification passed (conforms to the sequence of data acquisition → processing → evidence storage);

[0122] Output result: As shown in the "Output Format" example above, the final timing status is valid.

[0123] Results data:

[0124] Modeling time: The average time to process 100 timestamps is ≤50ms;

[0125] Label accuracy: 100% for the rule engine and 96.3% for the LSTM model;

[0126] Anomaly detection rate: 100% accuracy in identifying time-series anomalies out of 1000 test data points.

[0127] Step 4: Multi-media data collection and privacy compliance processing (see instruction manual appendix) Figure 3 The dataset employs a swimlane diagram structure, divided into three stages: "collection-processing-evidence storage." It showcases compliance processing measures for images (AI-blurred spatiotemporal scenes), pictures (sensitive information masking), audio (sensitive content anonymization), text (personal confirmation), music (copyright verification), and artwork (abstracted privacy elements). Technical methods such as "k-anonymity algorithm" and "training data compliance verification" are clearly labeled, intuitively demonstrating the logic of privacy protection throughout the entire lifecycle of multi-media data and aiding in understanding the balance between compliance processing and narrative needs.

[0128] Based on the themes of each unit and the corresponding real-life experiences, targeted data collection and processing were conducted across six media platforms, specifically including:

[0129] Image Acquisition: Key scenes (including human behavior, environmental details, etc.) at corresponding stages are recorded using video shooting, remaining faithful to the experience itself. Scenes from distant times without original footage can be reenacted through scene reconstruction, combined with AI-powered blurred spatiotemporal scene algorithms for intelligent blurring (e.g., fading specific street scenes and blurring the faces of non-core individuals). For scenes without original footage, AI-generated scene maps can assist in scene reconstruction. The generated maps must be based on details from the parties' oral accounts, labeled "AI-assisted reconstruction," and the generation parameters must be recorded. The specific application parameter k of K-anonymity is set to 5. Infringement samples are filtered from the training data of the image scene maps, retaining blurred spatiotemporal markers such as "that summer" and "old alleyway entrance" to enhance the sense of scene. Fabricating core events that did not occur is strictly prohibited. All devices (cameras, microphones, processing servers) synchronize their system clocks via the NTP protocol (Network Time Protocol) with an accuracy of ≤1ms. Video frame timestamps are based on the "system time at the frame acquisition time," and audio sampling point timestamps are based on the "system time at the sampling time," ensuring cross-media time alignment. In addition to routine processing, the following additional steps are performed on medical images: timeline marking [processing. medical record number desensitization], recording the desensitization algorithm (such as hash salting, salt value = hospital code + timestamp); and adding a time sequence check rule [medical record number desensitization must be done after image desensitization] to prevent privacy leaks.

[0130] Image Acquisition: Collect original photos of key life moments (such as graduation photos, ID photos, scene photos, etc.), anchoring specific experiences (e.g., "2008 college entrance examination admission ticket photo" corresponds to "admission to higher education during adolescence"); scenes without original photos are reconstructed through AI-generated images or hand-drawn documentary images. The generated materials must be based on details from the individual's memory (e.g., "red sofa at home in 1995"), and labeled "AI restoration" or "hand-drawn restoration"; Processing Rules: In real photos, the faces of others and sensitive information (ID numbers, addresses) are blurred or partially obscured by pixelation, retaining only the atmosphere of the scene; images involving portrait rights must be desensitized by AI or authorized; AI-generated images must verify the compliance of training data (to avoid infringing materials), and hand-drawn images must ensure originality, with core elements faithfully depicting the true scene.

[0131] Audio Acquisition: Synchronously acquire or restore audio information (dialogue recordings, dubbing, environmental sound effects, etc.) of the corresponding scene; for sensitive information (specific names, addresses), protect privacy through audio noise reduction, voice changing, or clip editing; for scenes without original audio, environmental sound effects (such as "tickling of an old clock" or "cicadas chirping in summer") can be generated by AI. The generated sound effects must match the era of the scene and be labeled "AI-generated sound effects"; retain iconic audio markers such as "tickling of an old clock" and "cicadas chirping in summer" to enhance the immersive experience.

[0132] Text material collection: Based on the core themes of each stage and unit and real life experiences, compile theme setting documents, narration scripts and subtitle texts; the text content can be generated with the assistance of AI, and after generation, it needs to be confirmed and corrected by the person involved to ensure that it is consistent with the real experience and that there are no fabricated core plots.

[0133] Music materials: Select music related to real experiences (such as popular songs at the time of the event, melodies in the memories of the parties involved); conduct compliance verification of copyrighted music (use original or authorized genuine versions); for scenarios without original music materials, generate original music through AI that matches the emotional characteristics of the real experience (such as the lightheartedness of childhood, the composure of middle age); for copyrighted music, adjust the rhythm and timbre through AI melody adaptation technology to avoid infringement while preserving the original emotional characteristics, and keep comparison files of the melody before and after the adaptation; retain music identifiers such as "childhood nursery rhyme fragments" to match the stage theme.

[0134] Artwork materials: Create artworks that match the scene (such as hand-drawn scenes from a long time ago, or illustrative summaries of key events); for scenes without original visual materials, artwork materials can be generated through AI, and the generated materials must be faithful to the real details (such as "classroom desks from the 1980s"); sensitive information is protected for privacy through abstraction and element replacement; a minimalist artistic design style is adopted to present the unit theme, enhancing the sense of the times and visual uniqueness.

[0135] Step 5: Timestamp chained evidence storage (see instruction manual appendix) Figure 4 This demonstrates the "main chain-branch chain" architecture of blockchain evidence storage: the main chain contains evidence storage nodes for original materials (such as original image files and text confirmation records), while the branch chains mark evidence storage nodes for newly added materials (such as supplementary elementary school award certificate photos); the association logic between the two is marked through "timestamp association" and "hash value verification," intuitively demonstrating the scalability and tamper-proof characteristics of dynamic evidence storage, and aiding in understanding the implementation scheme of seamless integration between new materials and historical data.

[0136] By using trusted timestamp notarization tools (such as the Joint Trusted Timestamp Service Center), the entire process of materials from six major media platforms is synchronously notarized, forming an immutable "timeline-media platform" cross-verification chain:

[0137] Image materials: original shooting files, scene reenactment scripts, comparison files before and after AI blurring, and records of parameters for generating AI-assisted reenactment images;

[0138] Image materials: original photo scans (including shooting time and metadata), source files of AI-generated images (including parameter settings and training data descriptions), hand-drawn images (including modification records), and the final version after privacy processing;

[0139] Audio materials: original recordings, AI-processed audio files, parameter records of AI-generated sound effects, and instructions for editing sensitive information;

[0140] Text materials: theme setting document, narration script, subtitle text (including confirmation records from the parties involved);

[0141] Music materials: copyright authorization documents, melody comparison before and after AI adaptation, original music composition records, and parameter records of AI-generated music;

[0142] Artwork materials: design sketches, final draft files, detailed reproduction instructions, and compliance reports of training data for AI-generated artwork materials;

[0143] All the above materials are used to generate evidence files containing time information and hash values, which are categorized and linked according to life stages to form a collaborative evidence storage system of "static (images, text, art) - dynamic (video, audio, music)".

[0144] Step 6: Storytelling Editing and Logical Anchoring

[0145] Following the logic of "unit theme → scene fragment → timestamp anchor point", a systematic and coordinated narrative is achieved across six media platforms:

[0146] Linking images and videos: Real old photos are inserted into video clips as "time anchors" (e.g., after the video "Leaving Home at 18," a photo of the train station from that year is inserted, accompanied by the text "The last group photo on the platform in September 1999"); AI-restored images / hand-drawn pictures echo the video scenes (e.g., in the video "Summer Night in Childhood," a hand-drawn "old locust tree and bamboo bed" is interspersed, with cicada sound audio superimposed); the visual elements of the images (such as color and composition) are consistent with the camera language of the videos (such as framing and color tone).

[0147] The audio, music and visual media work together: dialogue audio matches the characters' actions in the video / picture, and environmental sound effects enhance the realism of the scene; the music melody progresses with the narrative rhythm (e.g., a low melody for a scene of setbacks, and a gradually increasing tone for a scene of turning points), synchronizes with the rhythm of the video shots (slow motion / fast cut), and the music style matches the emotional tone of the life stage.

[0148] Symbolic connection between text and art: Textual narration extracts the unit theme and echoes the iconic symbols of the art materials (such as "old fountain pen, old schoolbag"); cross-unit clues are realized through the reuse of media carriers (such as the appearance of childhood art symbols "old schoolbag illustration" in adult images), which strengthens the core theme.

[0149] Style consistency: The six media platforms maintain a consistent minimalist design language (such as a unified font and icon system); Art color scheme is related to the stage theme: warm colors (such as orange and yellow) are used for growth themes, transitional colors (such as yellow-green and blue-purple) are used for transitional themes, and cool colors (such as dark blue and gray-green) are used for settling themes.

[0150] Step 7: Serialization and Verification

[0151] Before each unit is released, the entire system is closed-loop through a "three-dimensional verification method":

[0152] Main theme coherence: Verify the correspondence between the unit theme and the core clue through the "theme-main theme matching table" to ensure that the theme words of each unit (such as "independence" and "turning point") are within the semantic range of the core clue (such as "self-reconciliation in growth").

[0153] Media synergy: The timestamp deviation of the six media carriers is checked by the timeline synchronization tool. The timestamp deviation of the images, pictures and audio of the same event must be ≤24 hours (based on the time of the event). Manual calibration is triggered when the deviation exceeds the limit. The similarity of the music style of adjacent units (by comparing the melody waveform) is not less than 60%, and the deviation of the color saturation of the art color tone does not exceed 20%.

[0154] Evidence integrity: Generate an "Evidence Coverage Report" to ensure that the evidence coverage rate of the six media carriers in each unit reaches 100%, and that the evidence chain with the previous unit is linked by hash value, which can be traced back to the initial stage;

[0155] At the same time, it foreshadows later units (such as "childhood music clips that will reappear in the next stage" and "the 'old clock' artistic symbol that runs throughout the series"), ensuring that the entire series forms an organic whole. Compared with traditional manually annotated timelines:

[0156] Semantic modeling efficiency is improved by 87% (manual annotation of one timeline takes 120 seconds, while this invention takes only 15 seconds);

[0157] Privacy audit accuracy improved by 92% (parameters that are easily missed by manual auditing are forcibly associated through an algorithm in this invention);

[0158] Storage costs are reduced by 53% (parameterized logging replaces full logs, reducing storage from 10GB / day to 4.7GB / day). Each module has a built-in circular cache (capacity ≥1000 entries) for temporary timestamp storage;

[0159] If the upload fails, it will retry every 5 seconds. After 3 failed retries, it will be marked as "[Timestamp upload failed, in local cache]" to ensure that the data is not lost.

[0160] Step 8: Multi-form output and interactive presentation of storytelling (see instruction manual appendix) Figure 5 : Show the architectural relationship of multi-form output (dynamic demonstration: web page interactive timeline; static output: 1080P video, PDF e-book) and multi-terminal adaptation (mobile phone, tablet, VR device); mark technical standards such as "HTML5 development", "OpenXR protocol" and "API editing interface", intuitively reflect the diversity of output forms and future technology compatibility, and help understand the specific implementation path of multi-terminal interactive presentation. : (1) Generate interactive timeline narrative works:

[0161] Dynamic demonstration: As the timeline scrolls, multimedia content corresponding to the event node is presented synchronously (e.g., clicking the "University Graduation" node will simultaneously play the graduation video, recitation audio, commemorative text, and hand-drawn illustrations).

[0162] Static output: Automatically edit and generate standardized videos (resolution ≥ 1080P, format MP4), interactive web pages (developed based on HTML5, supporting responsive layout), e-books (PDF format, including timeline index and material thumbnails); (2) Multi-terminal adaptation and expansion: Support mobile phones (iOS / Android system), tablets, smart screen terminals, and provide output versions adapted to different screen ratios; Provide editing interface (API), allowing users to adjust the material order and text narration; Reserve standardized technology expansion interface, which can be updated to adapt to future mainstream interactive technologies (such as VR, holographic projection), and the expansion interface must comply with industry common protocols (such as the OpenXR protocol in the VR field) to ensure compatibility.

[0163] Step 9: Dynamic Iteration and Supplementary Evidence Storage

[0164] As life progresses or new materials are discovered, the output units are dynamically updated:

[0165] (1) Processing of new materials: New materials (such as photos of elementary school award certificates discovered by chance, or recordings of university clubs that were not included) need to be processed again according to the compliance standards in step 3 (such as supplementing the sensitive information of the award certificate and verifying the copyright of the recording) to ensure that they are consistent with the processing rules of the original unit materials;

[0166] (2) Supplementary evidence storage: New materials generate independent timestamp evidence storage files, which are associated with the original evidence storage chain of the corresponding stage unit through "supplementary identifier" to form an evidence storage structure of "main chain + branch chain". The branch chain can be retrieved independently but does not affect the integrity of the main chain.

[0167] (3) Update output: Insert new content into the original timeline work in the form of "supplementary unit", and mark "supplemented in × year × month" and the reason for the supplement (such as "a primary school certificate was found in May 2024, supplementing and restoring the childhood academic experience"); the updated work needs to re-execute the "three-dimensional verification" in step 6 to ensure the continuity of the main line and the synergy of the carrier.

[0168] Beneficial effects

[0169] The beneficial effects of this invention are as follows:

[0170] (1) Narrative Systematization: By dividing cross-unit main lines and stage units, multi-media materials are integrated into a logically coherent story chain, solving the problem of "having materials but no narrative", enhancing the emotional appeal of life experience records, and upgrading personal memories from "fragmented materials" to "structured stories";

[0171] (2) Semantic Timeline: Breaking through the traditional "sorting function" of the timeline, it constructs an "event relationship network" to achieve a three-dimensional mapping of "time-event-media", improving the accuracy of material association. Users can quickly locate all media materials of a certain event through the timeline;

[0172] (3) Compliance and security: The entire process is embedded with privacy protection and copyright verification mechanisms, covering the entire life cycle of material collection, processing and storage. This avoids the risks of portrait rights and copyright infringement, while maintaining the narrative immersion through "retention of iconic symbols" (such as cicada chirps and old object symbols), thus balancing compliance and experience.

[0173] (4) Evidence reliability: A “static-dynamic” collaborative evidence chain is built through trusted timestamps to ensure that the materials cannot be tampered with, and the “main chain + branch chain” structure supports seamless connection of new materials, maintaining the long-term authenticity and integrity of the timeline and meeting the need for “continuous recording” of life experiences.

[0174] (5) Application scalability: It supports multi-terminal adaptation and compatibility with future technologies (such as VR and holographic projection). The output forms cover dynamic demonstrations, static files and interactive web pages. It can be applied to scenarios such as personal memory management, family cultural inheritance and cultural heritage digitization, thus broadening the application boundaries of the technology. Attached Figure Description Figure 1 This is a time-axis semantic modeling logic diagram of the method described in this invention. It displays the semantic association model of "time-event-media" in a three-dimensional coordinate format (X-axis: timestamp, Y-axis: event relationship network, Z-axis: multiple media types). Figure 2 This is the technical solution process architecture of the method described in this invention. It is used to demonstrate the seven core links and logical loops of this invention: "mainline guidance - unit division - compliant collection - chain-based evidence storage - collaborative narrative - verification output - dynamic iteration". Figure 3This is a flowchart of the multi-media privacy processing method described in this invention. Using a swimlane diagram structure, it is divided into three stages: "collection-processing-existence," respectively demonstrating the compliance processing measures for images (AI-blurred spatiotemporal scenes), pictures (sensitive information masking), audio (sensitive content anonymization), text (personal confirmation), music (copyright verification), and artwork (abstracting privacy elements). Figure 4 This is the blockchain evidence storage logic of the method described in this invention. The "main chain-branch chain" architecture of blockchain evidence storage is demonstrated: the main chain contains evidence storage nodes for the original materials, and the branch chains mark evidence storage nodes for newly added materials. Figure 5 This is a technical illustration of the method described in this invention. It demonstrates the architectural relationship of multi-form output (dynamic demonstration: web page interactive timeline; static output: 1080P video, PDF e-book) and multi-terminal adaptation (mobile phone, tablet, VR device, etc.). Detailed Implementation

[0175] The following uses the "childhood period (1990-2000)" as an example to illustrate the implementation process of this invention in detail, so that those skilled in the art can clearly understand and implement this solution:

[0176] Step 1: Pre-set cross-unit narrative main line

[0177] The core narrative thread was determined to be "exploration and growth in childhood," and thematic anchors were set for the six major media platforms:

[0178] The visuals focus on "outdoor play scenes" (such as catching cicadas and playing jump rope), and the camera language uses "low-angle tracking shots" to simulate a childhood perspective.

[0179] Images: Anchored to "kindergarten graduation photo in 1993 and birthday photo in 1998", highlighting "the joy captured in a moment";

[0180] Audio: "Recordings of Grandma telling stories in 1995, and summer cicada chirping sound effects" were collected to enhance the "ambience";

[0181] Text: The narration style is "a childhood-perspective reminiscence," such as "The cicadas chirped that summer louder than in any later year."

[0182] Music: Selected from the popular 1990s children's song "Let Us Row Our Boats" (authorized original excerpt), AI adapted into a light music version;

[0183] Artwork: Hand-painted "1990s bungalow courtyard", restoring details such as "red wooden door, cement floor, clothesline".

[0184] Step 2: Divide the narrative into units

[0185] The "Childhood Period (1990-2000)" section is divided into three sub-units, with each sub-unit's theme anchored to real-life experiences as follows:

[0186] Early childhood (1990-1993): Theme "First Encounter with the World", anchored on "First time walking alone in 1992" and "Entering kindergarten in 1993";

[0187] Primary school term (1994-1999): Theme "Campus Partners", focusing on "joining the Young Pioneers in 1996" and "winning the primary school sports meet in 1998";

[0188] Primary school to junior high school transition (2000): The theme is "Growth Turning Point", focusing on "2000 primary school graduation exam" and "first time away from home to attend summer camp".

[0189] Step 3: Multi-media data collection and compliance processing

[0190] Video footage: In March 2024, a video recreating a "jump rope scene" was filmed (the actors were children aged 6-8, dressed in 1990s style). An AI-based spatiotemporal blurring algorithm was used to soften the modern street scene in the background. For the scene of "walking alone in 1992" without original footage, an AI-generated scene map (based on the mother's description of "red walker, sycamore tree in front of the door") was created, labeled "AI-assisted reconstruction," and the generation parameters were saved. Historical semantic timeline data (samples marked as "[human verification passed]") was used as the training set, and the LSTM model was automatically fine-tuned monthly to improve the label accuracy by ≥3% per year.

[0191] Images: Scanned 1993 kindergarten graduation photo (faces of non-core students blurred, only the outlines of the person and the homeroom teacher are retained), 1998 birthday photo (neighbor's house number in the background is obscured);

[0192] Audio: Recorded in April 2024, a recording of the mother recalling "Grandma telling stories in 1995" (AI desensitized the name "Aunt Wang from next door" etc.), and AI generated "cicada chirping in the summer of the 1990s" sound effect (labeled "AI generated sound effect", parameters are "frequency 2-5kHz, duration 30s");

[0193] Text: The initial draft of the narration generated by AI was "On June 1, 1996, I put on my red scarf, and my palms were sweaty." After confirmation by the person involved, it was revised to "On June 1, 1996, on the school playground, my homeroom teacher tied the red scarf on my clothes, and the wind made the hem of my clothes flutter."

[0194] Music: Obtained the official authorization for "Let Us Row Our Boats" (authorized by the Music Copyright Society of China), adapted it into a piano light music version using AI, and saved the melody waveform comparison files before and after the adaptation;

[0195] Artwork: Hand-drawn "1990s bungalow courtyard" (red wooden door, cement floor), abstracting the neighbor's window (using color blocks to replace specific window panes), ensuring no infringing elements.

[0196] Step 4: Timestamp Chained Evidence Storage

[0197] In May 2024, the above materials were stored through the "Joint Trust Timestamp Service Center," generating the following evidence file:

[0198] Imagery: Scene reconstruction video (hash value: XXX), AI-blurred comparison file (hash value: XXX), AI-assisted reconstruction map parameters (hash value: XXX);

[0199] Images: Scanned copy of graduation photo (including metadata from 1993, hash value: XXX), anonymized photo (hash value: XXX);

[0200] Audio: Mother's recording (hash value: XXX), AI cicada sound effect (hash value: XXX);

[0201] Text: Final narration (including confirmation record from the party involved in April 2024, hash value: XXX);

[0202] Music: License file (hash value: XXX), melody adaptation comparison (hash value: XXX);

[0203] Artwork: Hand-drawn sketches (hash value: XXX), final draft (hash value: XXX);

[0204] All archival documents are linked to the "Childhood Unit (1990-2000)" timeline to generate a unit archival index table.

[0205] Step 5: Storytelling Editing and Logical Anchoring

[0206] Video and Images: After a video reenacting the scene of "joining the Young Pioneers in 1996" (30s), insert a photo of the red scarf in 1996 (static display for 5s) with the caption "In June 1996, I wore the red scarf for the first time, and the wind made the hem of my clothes flutter."

[0207] Audio and Music: When playing the mother's recording of her memories, AI cicada chirping sound effects are superimposed, and the background music gradually enters the adapted version of "Let Us Row Our Boats" (the volume is adjusted according to the recording rhythm);

[0208] Text and Art: In the "1998 Birthday" scene, the text narration "I counted the candles on the cake three times before I could count them all" appears simultaneously with the hand-drawn "birthday cake illustration," and the illustration style is consistent with the overall minimalist design;

[0209] Consistent style: The font used in all units is "Founder Kids Simplified", the color scheme is warm yellow (RGB: 255, 240, 180), and the music rhythm is kept upbeat (tempo: 100 BPM).

[0210] Step 6: Serialization and Verification

[0211] Main theme coherence: It is confirmed that the sub-themes of "campus partners" and "growth turning point" all revolve around the main theme of "exploration and growth in childhood" without deviation;

[0212] Carrier synergy: The timestamp deviation of the video, photo, and audio recordings for "joining the Young Pioneers in 1996" is 0 (all anchored to June 1996), and the similarity of the music style of adjacent sub-units is 75% (meeting the requirement of ≥60%);

[0213] Completeness of evidence preservation: 100% of the 6 types of materials are preserved, and a reserved interface is provided for the evidence preservation chain of the "Youth Period Unit (2001-2010)" (e.g., the childhood "Red Scarf" art symbol will be reused in the "Joining the Party in 2008" unit of youth).

[0214] Pre-planned foreshadowing: Insert an "old clock" artistic symbol at the end of the "Primary to Junior High" sub-unit, with the note "This symbol will reappear in the middle-aged unit".

[0215] Step 7: Multi-format output and interactive presentation

[0216] Dynamic demonstration: Generate an interactive timeline for web pages (based on HTML5). When a user clicks on the "1998 birthday" node, a birthday photo, a recording of the mother, and a hand-drawn illustration will be played simultaneously.

[0217] Static output: Generate 1080P video (MP4 format, 5 minutes in length) and PDF e-book (including timeline index, allowing you to jump to the corresponding material);

[0218] Multi-terminal adaptation: Provides mobile (375×667 resolution) and tablet (1024×768 resolution) adapted versions, and opens API interface to allow users to adjust the narration text.

[0219] Step 8: Dynamic Iteration and Supplementary Evidence Storage

[0220] In October 2024, a photo of a primary school award certificate from 1997 was discovered. The processing procedure is as follows:

[0221] Material processing: Scan the award certificate (blurring the homeroom teacher's name) and confirm that there is no sensitive information;

[0222] Supplementary evidence storage: Generate a timestamp evidence storage file (hash value: XXX), associate it with the "Short Semester (1994-1999)" evidence storage branch, and mark it "supplemented in October 2024";

[0223] Update output: Insert the certificate image into the "Mini-Semester" unit of the PDF e-book, with the caption "Found the 1997 'Three Good Students' certificate in October 2024, supplementing and restoring my childhood academic experience"; re-execute step 6 to verify and confirm that the main storyline is unaffected.

Claims

1. Claim 1 (Independent Claim) A multi-media storytelling method for life experiences based on a timeline, characterized in that, The steps include: (1) Pre-setting a cross-unit narrative thread: Before dividing life stages, determine the core narrative thread that runs through the whole series to ensure that each stage unit can be independent and form a "system story chain" through the main thread; the main thread is adapted to the coordinated expression of six media carriers: video, audio, text, pictures, music and art, and sets a unified emotional tone and theme anchor for each media carrier; (2) Narrative units are divided according to life stages: Based on the core storyline, life is divided into several stages. Each unit sets a core theme that corresponds to the storyline, and the core theme anchors the key life experiences that actually occurred in that stage. Each unit clarifies the division of labor of the six major media carriers to ensure that the theme originates from experience and the media carriers serve reality. (3) Multi-media carrier collection and privacy compliance processing: For each unit theme and the corresponding real life experience, the six major media carriers are collected and processed in a targeted manner, including the collection, restoration and privacy protection of images, pictures, audio, text, music and art materials to ensure that the materials are real and compliant. (4) Timestamp chain evidence storage: Through a trusted timestamp evidence storage tool, the materials of the six major media carriers are stored synchronously to generate evidence files containing time information and hash values. They are classified and associated according to life stage units to form a "static-dynamic" collaborative evidence storage system. (5) Story-based editing and logical anchoring: According to "unit theme → scene fragment → timestamp anchor point" The logic is to realize the system collaborative narrative of the six major media carriers. Through the linkage of images and pictures, the collaboration of audio and music with visual carriers, and the connection of text and art symbols, the core main line is strengthened and the style is kept consistent; (6) Series connection and verification: Before each unit is released, the "three-dimensional verification method" (main line coherence, carrier collaboration, and evidence integrity) is used to ensure the closed loop of the whole series system and to pre-set the foreshadowing of subsequent units; (7) Multi-form output and interactive presentation: Generate interactive timeline narrative works, support dynamic demonstration and static output, adapt to multiple terminals and reserve technical expansion interfaces; (8) Dynamic iteration and supplementary evidence storage: As life stages progress or new materials are discovered, the units that have been output are dynamically updated, and the newly added materials are processed according to the compliance standards of step (3) and stored synchronously to maintain the authenticity and systematicity of the timeline. (9) K-anonymity algorithm is adapted to multi-media narrative scenarios, and K-anonymity runs through the entire process of evidence storage. (10) Clarify the technical means of "street view blurring, face desensitization, and ID number blurring", supplement the method of "voice changing name processing" and the specific operations of "AI-generated sound effect annotation", "timeline association" and "algorithm parameter recording". (11) Split the functional units of the main chain and the branch chain, define the connection method between the branch chain and the main chain, and limit the startup logic of the branch chain. (12) A computer-readable storage medium storing a computer program, which, when executed by a processor, implements a timeline semantic modeling method. (13) The evidence storage unit uses blockchain nodes to implement distributed evidence storage, and the timeline data is solidified through smart contracts to ensure that it is tamper-proof. (14) The timeline semantic modeling and the algorithm logic of timeline semantic modeling realize the three-dimensional mapping of "time-event-medium" through timestamp chain evidence storage, semantic association analysis and dynamic editing functions.

2. Claim 2 (defining the feature of "image acquisition" in step 3) The method according to claim 1, characterized in that, Step (3) "Image Acquisition" includes: using video shooting to record key scenes of the corresponding stage, faithfully reflecting the experience itself; for scenes that are too old to have original images, scene reproduction is used to recreate the scene, combined with AI blurring spatiotemporal scene algorithm to fade specific street scenes and blur the faces of non-core people; for scenes without original images, scene maps are generated by AI to assist in scene reproduction, the generated map is based on the details of the parties' oral accounts, marked "AI-assisted reproduction" and the generation parameters are recorded; blurred spatiotemporal markers such as "that summer" and "old alley entrance" are retained to enhance the sense of scene; it is strictly forbidden to fabricate core plots that did not happen.

3. Claim 3 (defining the feature of "image acquisition" in step 3) The method according to claim 1, characterized in that, Step (3) "Image Collection" includes: collecting original photos of key life moments to anchor specific life experiences; for scenes without original photos, restoring them through AI-generated images or hand-drawn documentary images, with the generated materials based on the details of the person's memory and labeled "AI Restoration" or "Hand-drawn Restoration"; handling the privacy and copyright of images includes: blurring or partially obscuring the faces of others in real photos, retaining only the atmosphere of the scene; images involving portrait rights need to be desensitized by AI or authorized; verifying the compliance of training data with AI-generated images, and ensuring the originality and faithfulness of core elements to the real description in hand-drawn images.

4. Claim 4 (defining the feature of "audio acquisition" in step 3) The method according to claim 1, characterized in that, Step (3) "Audio Acquisition" includes: synchronously acquiring or restoring dialogue recordings and environmental sound effects of the corresponding scene; protecting the privacy of sensitive information such as specific names and addresses involved in the audio through audio noise reduction, voice changing or clip editing; generating environmental sound effects through AI for scenes without original audio, matching the generated sound effects with the scene's era and labeling them as "AI-generated sound effects"; retaining iconic audio markers such as "tickling of an old clock" and "summer cicada chirping", and matching the markers with the scene characteristics of the corresponding life stage to enhance the immersive narrative experience.

5. Claim 5 (defining the features of "text material collection" in step 3) The method according to claim 1, characterized in that, Step (3) "Text material collection" includes: compiling theme setting documents, narration scripts and subtitle texts based on the core themes of each stage unit and real life experiences; The text content can be generated with the assistance of AI. After generation, it needs to be confirmed and corrected by the person involved to ensure that it is consistent with the real experience and that there is no fabricated core plot. The writing style should match the emotional tone of the unit theme (e.g., use lively expressions for childhood and calm expressions for middle age).

6. Claim 6 (defining the characteristics of "musical material" in step 3) The method according to claim 1, characterized in that, The processing of "music material" in step (3) includes: selecting music related to the real experience, wherein the music is a popular song at the time of the event or a melody in the memory of the person involved; Compliance verification is performed on copyrighted music, using original music or authorized genuine music; for scenarios without original music materials, original music is generated using AI to match the emotional characteristics of real experiences; for copyrighted music, the rhythm and timbre are adjusted using AI melody adaptation technology to avoid infringement while preserving the original emotional characteristics, and comparison files of the melody before and after the adaptation must be kept; music identifiers such as "childhood nursery rhyme fragments" are retained, and the identifiers match the unit theme of the corresponding stage.

7. Claim 7 (defining the characteristics of "artistic material" in step 3) The method according to claim 1, characterized in that, The processing of "art materials" in step (3) includes: creating artworks that match the scene, including old hand-drawn scenes and illustrations summarizing key events; For scenes without original visual materials, AI generates art materials that are faithful to real details (such as "classroom desks from the 1980s"); sensitive information is protected by abstraction and element replacement; a minimalist artistic design style is used to present the unit theme, enhancing the sense of the times and visual uniqueness, and the art elements echo the core theme (such as using the "old fountain pen" symbol to run through the "growth" theme).

8. Claim 8 (limiting the scope of materials for "timestamp chain storage" in step 4) The method according to claim 1, characterized in that, The "full-process materials" in step (4) include: original shooting files of video materials, scene reenactment scripts, comparison files before and after AI blurring, and generation parameter records of AI-assisted reenactment images; original photo scans of image materials (including shooting time metadata), source files of AI-generated images (including parameter settings and training data descriptions), creation process drafts of hand-drawn images (including modification records), and final drafts after privacy processing; original recordings of audio materials, audio files after AI processing, parameter records of AI-generated sound effects, and instructions for editing sensitive information; theme setting documents, narration scripts, and subtitle texts (including confirmation records from the parties involved) of text materials; copyright authorization documents of music materials, melody comparisons before and after AI adaptation, original music composition records, and parameter records of AI-generated music; design sketches, final draft files, detail restoration instructions, and compliance reports of training data for AI-generated art materials.

9. Claim 9 (defining the features of the "collaborative evidence storage system" in step 4) The method according to claim 8, characterized in that, In step (4), the "static-dynamic collaborative evidence storage system" is implemented in the following way: static materials (images, text, art) and dynamic materials (videos, audio, music) are linked by life stage units, and the evidence storage file of each stage unit contains a unique stage identifier; the timestamps of static materials and dynamic materials are cross-validated by hash value to ensure that the static and dynamic materials of the same event point to the same time node; the evidence storage file is stored in a hierarchical structure of "stage unit → event node → media carrier type", which supports forward tracing or reverse retrieval by time axis, and the retrieval results include the evidence storage status identifier of the material (such as "original", "processed" or "supplemented").

10. Claim 10 (defining the feature of "image and picture linkage" in step 5) The method according to claim 1, characterized in that, Step (5) "Image and picture linkage" includes: real old photos are inserted into video clips as "time anchors", the shooting time metadata of the photos and the timestamp of the video clips are ≤24 hours apart, and the photos are marked with corresponding text descriptions (including specific time and event descriptions); when AI-restored images, hand-drawn pictures and video scenes correspond, the visual elements of the images (such as color and composition) are consistent with the lens language of the images (such as shot size and tone), and the generation logic of the images (based on memory details or historical data) is explained in the subtitles. The generation logic explanation must match the material source record in the evidence file.

11. Claim 11 (defining the feature of "audio, music and visual carrier coordination" in step 5) The method according to claim 1, characterized in that, Step (5) "Audio, music and visual carrier coordination" includes: the duration of the dialogue audio and the duration of the corresponding character's action in the video / picture deviate by ≤5 seconds to ensure that the lip movements and speech are synchronized; the volume of the ambient sound effects is automatically adjusted with scene switching (e.g., indoor scene volume ≤-15dB, outdoor scene volume ≥-10dB) to enhance the realism of the scene; the music melody progresses with the narrative rhythm, with a low melody (frequency ≤300Hz) for frustration scenes and a gradually increasing tone (volume increases by ≤2dB per second) for turning scenes, and the music rhythm is synchronized with the video shot rhythm (slow motion / fast cut), and the matching degree between the shot switching frequency and the music beat is ≥80%.

12. Claim 12 (defining the feature of "stylistic consistency" in step 5) The method according to claim 1, characterized in that, In step (5), "stylistic consistency" is achieved through the following methods: the six media carriers adopt a unified minimalist design language, including a preset font system (sans-serif font for titles and serif font for body text) and an icon library (symbols with a unified style for time nodes); the music style matches the emotional tone of each life stage: childhood uses upbeat melodies (tempo 90-120 BPM) and bright timbre (high frequencies ≥5kHz, accounting for ≥40%), adolescence uses varied rhythms (tempo 100-140 BPM) and mid-to-high frequency timbre (3-8kHz, accounting for ≥50%), and middle age uses steady melodies (tempo 60-90 BPM) and low frequency timbre (≤300Hz). (The proportion is ≥30%); the color scheme of the artwork is related to the theme of each stage: warm colors (hue 30°-60°) are used for growth themes, transitional colors (hue 70°-190°) are used for transitional themes, and cool colors (hue 200°-300°) are used for settling themes, and the color saturation difference between adjacent units does not exceed 20%.

13. Claim 13 (defining the details of the "three-dimensional verification method" in step 6) The method according to claim 1, characterized in that, Step (6) "Three-dimensional verification method" includes: Main line coherence verification: verify the correspondence between the unit theme and the core clue through the "theme-main line matching table" to ensure that the theme words of each unit are within the semantic range of the core clue, and the semantic matching degree is ≥80% (based on word vector cosine similarity calculation); Carrier synergy verification: check the timestamp deviation value of the six media carriers through the time axis synchronization tool. When the deviation exceeds ±24 hours, manual calibration is triggered, and the similarity of music style of adjacent units (through melody waveform comparison) is not less than 60%, and the deviation of art color saturation does not exceed 20%; Evidence integrity verification: generate "Evidence coverage report" to ensure that the evidence storage rate of the materials of the six media carriers of each unit reaches 100%, and the evidence storage chain with the previous unit is associated with the hash value, which can be traced back to the initial stage, and the traceability response time is ≤3 seconds.

14. Claim 14 (defining the feature of "multi-form output and interactive presentation" in step 7) The method according to claim 1, characterized in that, Step (7) "Multi-form output and interactive presentation" includes: Dynamic demonstration: When the timeline scrolls, the corresponding event nodes are triggered to present the multi-media content synchronously with a response delay of ≤500ms; Static output: Automatically edit and generate standardized videos (resolution ≥1080P, MP4 format), interactive web pages (developed based on HTML5, supporting responsive layout), and e-books (PDF format, including timeline index and material thumbnails); Multi-terminal adaptation: Supports mobile phones (iOS / Android system), tablets, and smart screen terminals, providing output versions adapted to different screen ratios; Extended interface: Provides an editing interface (API) to support users to adjust the material order and text narration; Reserves standardized technical extended interfaces, conforms to industry common protocols (such as the OpenXR protocol in the VR field), and can be adapted to future mainstream interactive technologies (such as VR and holographic projection).

15. Claim 15 (defining the processing rules for "dynamic iteration" in step 8) The method according to claim 1, characterized in that, The "dynamic iteration" in step (8) includes: the type of new material matches the media carrier division of the corresponding life stage unit, and the supplementary content does not change the core theme of the original unit, with a theme deviation of ≤10% (based on semantic analysis); the privacy processing standards of the new material are consistent with those of the original material: the sensitive information occlusion method of real photos and the compliance verification rules of training data of AI generated images are the same as the processing logic of the same type of material in step (3); the supplementary evidence file is marked with "supplemented in × year × month" and the reason for the supplement, and is associated with the evidence chain of the original unit through "supplementary identifier" to form an evidence structure of "main chain + branch chain". The branch chain can be retrieved independently but does not affect the integrity of the main chain; the updated output works need to re-execute the "three-dimensional verification" in step (6) to ensure that the continuity of the main line and the synergy of the carrier are not affected.

16. Claim 16 (defining the processing features of "adaptation and evidence preservation" in step 9) The method according to claim 1, characterized in that, The rule in step (9) that “K-anonymity algorithm adapts to multi-media narrative scenarios and is used throughout the entire process of evidence preservation” ensures the adaptability of K-anonymity algorithm to narrative scenarios, and the evidence preservation also ensures that K-anonymity algorithm is used throughout the entire process of evidence preservation.

17. Claim 17 (defining the processing features of "the method for processing fuzzy and desensitized information and the specific operations of "time axis association" and "algorithm parameter recording" in step 10) The processing method according to claim 1 is characterized by the following steps: S1. After collecting multimedia data, the data is verified by a training data verification module to reject input containing illegal privacy data; S2. The image data is processed as follows: AI blurring of street view (based on generative adversarial network, blurring background texture); facial desensitization (using key point masking algorithm to retain 30% of facial contour features); ID number pixel blurring (Gaussian blur radius ≥ 5px); the above processing is based on the K-anonymity algorithm (k=5) and retains scene-specific identifiers (such as street names); S3. The audio data is processed by voice changing (adjusting the fundamental frequency ±200Hz), and AI is called to generate sound effect annotations (marking the voice changing interval); S4. The timestamps of collection and processing are associated to form a timeline, and the K-anonymity algorithm parameters and AI annotation results are recorded and archived for evidence.

18. Claim 18 (limiting step 11 to "splitting the main chain and side chains, defining the side chains and the main chain, and limiting the processing rules for the side chains") The multi-media privacy compliance system according to claim 1 is characterized in that... In step (10), the functional units of splitting the main chain and the side chain are (e.g., main chain: collection → processing → evidence storage). Sidechains (training data verification, AI annotation): Define the connection method between the sidechain and the main chain (e.g., data interface, call timing), and limit the startup logic of the sidechain (e.g., "automatic verification after collection" "calling AI annotation during processing"). Includes: a main chain module, which sequentially contains a data acquisition unit (collecting multi-media data), a privacy processing unit (built-in K-anonymity algorithm, k=5), and an evidence archiving unit (associated with timeline evidence archiving); a training data verification sidechain, bidirectionally connected to the data acquisition unit, automatically triggered within 100ms after collection to verify data compliance; and an AI generation annotation sidechain, unidirectionally connected to the privacy processing unit, synchronously called when the privacy processing unit performs image / audio processing to generate icon annotations or sound effect annotations; wherein, the privacy processing unit performs AI blurring processing on street scenes, desensitization processing on faces, and pixel blurring processing on ID numbers, while retaining contextual identifiers (e.g., 'old alleyway entrance').

19. Claim 19 (defining "computer-readable storage medium storing a computer program" in step 12) The method according to claim 1, characterized in that, A computer-readable storage medium storing a computer program that, when executed by a processor, implements the time-axis semantic modeling method as described in claim 1.

20. Claim 20 (limiting the "immutability of the storage unit in step 13" to blockchain) The method according to claim 1, characterized in that... The evidence storage unit uses blockchain nodes to achieve distributed evidence storage, and the timeline data is solidified through smart contracts to ensure that it cannot be tampered with.

21. Claim 21 (limiting the "semantic modeling of the time axis and the algorithmic logic for semantic modeling of the time axis" in step 14) The method according to claim 1, characterized in that, The semantic modeling of the timeline and its algorithmic logic: The timeline proposed in this invention is not only a time sorting tool, but also an "event relationship network". Through timestamp chain storage, semantic association analysis and dynamic editing functions, the timeline can automatically divide life stages, mark key nodes, and associate timestamps and sentiment tags of multi-media materials to achieve a three-dimensional mapping of "time-event-media".

Images