Intelligent teleprompter system for training of a news anchor

Abstract

The present application relates to the technical field of broadcasting training, in particular to an intelligent word prompting self-adaptive scrolling system for broadcasting training, which acquires scripts and generates legal stop and link anchor points through structured processing, determines the current broadcast position in real time by aligning the voice, and detects the pause event; the system identifies the repair behavior characteristics after recognizing the pause, calculates the broadcast stability in combination with the recent pause history, and dynamically extends the effective promotion area accordingly; further, the system resets the judgment starting point of the promotion verification according to the differences of the repair behavior characteristics, and only outputs the control parameters for restoring the regular scrolling after confirming that the continuous scrolling has crossed the dynamic effective promotion area; the system realizes the decoupling of physical display and logical judgment, accurately matches the real error correction intention of the user, and effectively avoids the window misplacement and secondary jamming accidents caused by exploratory voice or serious withdrawal.

Description

Technical Field

[0001] This invention relates to the field of broadcasting and hosting training technology, specifically to an intelligent prompting and adaptive scrolling system for broadcasting and hosting training. Background Technology

[0002] Existing intelligent prompting systems typically combine speech recognition technology with text tracking algorithms, using the captured audio to automatically scroll and follow the text within the display window. This technology can smoothly advance the text according to the user's normal speaking speed, maintaining good synchronization between the audio stream and the display interface when the user is speaking fluently or making regular pauses for breath. This achieves the technical effect of freeing the announcer's hands and satisfying basic requirements for smooth and responsive audio delivery.

[0003] However, in real-world broadcasting training scenarios, users may exhibit non-standard corrective behaviors such as repeating mistakes or rewinding after errors. Existing technologies have significant drawbacks in addressing scrolling control issues in these disjointed states. Current systems often rely solely on a single voice activation state or a superficial understanding of the current recognition coordinates for mechanical scrolling start and stop control. They fail to deeply analyze the user's true intentions for error correction after pauses and lack a dynamic reset and verification mechanism for judging the user's gaze focus. This makes the teleprompter highly susceptible to severe visual misalignment between the physical display progress and the user's actual desired reading position when the user's gaze moves backward or the broadcasting status is unstable. This visual interference can lead to secondary glitches and stuttering incidents. Summary of the Invention

[0004] To address the problems in related technologies, this invention provides an intelligent prompting and adaptive scrolling system for broadcasting and hosting training, thereby overcoming the aforementioned technical problems existing in the prior art.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: an intelligent prompting and adaptive scrolling system for broadcasting and hosting training, comprising:

[0006] The anchor point generation module is used to acquire the text of the script to be broadcast and perform structured processing to generate legal stop anchor points that include anchor point position, anchor point range, initial effective advance zone and pre-reading requirement parameters.

[0007] The pause detection module is used to acquire the broadcast speech and align it with the text of the script to be broadcast to determine the current broadcast position; when the acoustic feature changes extracted from the broadcast speech meet the preset pause conditions, a pause event is generated and the broadcast position before the pause is recorded.

[0008] Anchor point matching module is used to select candidate anchor points from the legal pause anchor points based on the position already broadcast before the pause, and calculate the belonging degree of the candidate anchor points based on the distance between the position already broadcast before the pause and the position of the candidate anchor point, whether the pause event falls within the scope of the candidate anchor point, the matching degree between the pause duration and the allowable pause duration of the candidate anchor point, and whether the remaining effective text volume of the obtained prompting window meets the pre-reading requirements of the candidate anchor point, and the one with the highest belonging degree is the main anchor point;

[0009] The repair identification module is used to obtain the resumed broadcast position after the pause ends, calculate its positional offset from the broadcast position before the pause, and generate repair behavior feature data; when the positional offset indicates a re-reading, the re-reading point is recorded.

[0010] The advance adjustment module is used to obtain recent pause history data to calculate the broadcast stability; when the broadcast stability is lower than a preset threshold, the initial effective advance area of ​​the main anchor point is extended along the text advance direction according to the difference between the broadcast stability and the preset threshold to obtain a dynamic effective advance area;

[0011] The scrolling control module is used to determine whether the repair behavior feature data indicates a re-reading and re-rolling, and to determine whether the updated current broadcast position continuously crosses the dynamic effective advancement area from the determination starting point, and to output scrolling control parameters according to the determination result.

[0012] By employing the above technical solution, the present invention provides an intelligent prompting and adaptive scrolling system for broadcasting and hosting training, which has at least the following beneficial effects:

[0013] 1. This invention establishes a pause detection basis by aligning structured anchor points with real-time voice, characterizes the repair behavior by spatial position offset, and dynamically extends the effective advancement zone based on the time decay stability of recent pause history. It also resets the advancement verification starting point according to the repair type, transforming the scrolling recovery into a state machine verification process that continuously crosses dynamic boundaries. This decouples the physical display control from the underlying error correction logic, avoiding window misalignment and secondary jamming caused by tentative vocalizations or severe retractions.

[0014] 2. This invention transforms recent pause history data into broadcast stability with time decay weights, enabling the system to dynamically extend the effective propulsion zone through a segmented mapping function. Strict continuous crossing judgments are performed on this extended dynamic effective propulsion zone, which can effectively filter out the interference of students' tentative short vocalizations and ensure that the control parameters for restoring the regular scrolling screen are triggered only when a real and continuous broadcast propulsion intention appears.

[0015] 3. This invention maps disordered error correction actions into structured repair behavior features by calculating the offset direction and magnitude of absolute spatial coordinates before and after the pause in real time. In particular, when it is determined to be a reread, the verification starting point is forcibly shifted to the reread landing point, and the offset adaptation of the termination boundary is performed according to the reread distance. This ensures that the coordinate origin of the control verification is always absolutely anchored to the physical position of the user's actual reread, avoiding the spatial separation between logical judgment and visual focus.

[0016] 4. This invention sets the scrolling recovery as a closed-loop state machine verification that strictly and continuously crosses the termination boundary of the dynamic effective advancement area. If an acoustic change or new position jump is detected before reaching the boundary, the current locked state is forcibly maintained and the new event is packaged into the history record. The penal recalculation of the boundary is triggered cyclically. This can implement high-intensity logical interception when the user has not completely reconstructed the coherent speech flow, providing a reliable security guarantee for the smooth operation of the prompting system. Attached Figure Description

[0017] The above and other objects, features, and advantages of this application will become more apparent from the more detailed description of the embodiments of this application in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the embodiments of this application to explain this application and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0018] Figure 1 The data flow diagram of the intelligent prompting and adaptive scrolling system for broadcasting and hosting training provided by the present invention;

[0019] Figure 2 This is a schematic diagram of the module of the intelligent prompting and adaptive scrolling system for broadcasting and hosting training provided by the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Exemplary system:

[0022] In broadcasting and hosting training scenarios, teleprompter systems need to accurately identify the type of correction behavior exhibited by trainees after non-standard pauses (such as repetition or tentative re-reading) and adjust the scrolling recovery judgment benchmark accordingly to avoid misalignment between the text position in the teleprompter window and the trainee's actual broadcasting position. Existing teleprompter systems lack the ability to finely identify the type of correction behavior after a pause, and when a trainee repetitions, they still use the text position before the pause as the benchmark for scrolling recovery judgment, leading to visual misalignment and secondary stuttering incidents. To address this, this embodiment proposes an intelligent teleprompter adaptive scrolling system for broadcasting and hosting training.

[0023] In this embodiment, a news broadcast training script is used as an example to demonstrate the complete workflow of the entire system. Let the text T to be broadcast be a 200-word news article, with its text coordinate space as follows: , of which each Indicates the first The absolute physical coordinates of each character (in terms of character index) are displayed in the prompt window, which currently shows the content from the 80th to the 140th character. The student is currently reading the 105th character.

[0024] like Figures 1-2 As shown, the system includes:

[0025] The anchor point generation module is used to acquire the text of the script to be broadcast and perform structured processing to generate legal stop anchor points that include anchor point position, anchor point range, initial effective advance zone and pre-reading requirement parameters.

[0026] It should be noted that this module preprocesses the broadcast script text, extracting all legal pause / connection anchors through linguistic analysis and pre-configuring relevant attribute parameters for each anchor. Legal pause / connection anchors refer to text positions within the script text that, based on grammatical structure, punctuation, and semantic logic, allow the announcer to make standardized pauses or connections. This module provides the foundational data support for subsequent anchor matching and pause type determination.

[0027] The generation of legal anchor points includes:

[0028] First, punctuation marks are extracted based on their features. Then, non-punctuation marks, sentence group boundaries, logical transitions, and sentence initiation points are extracted by combining syntax tree sequence features and sentence semantic features.

[0029] Specifically, the system first scans the manuscript text for punctuation marks, marking the locations of punctuation marks such as periods, commas, pauses, semicolons, colons, question marks, and exclamation marks as punctuation pause positions. Then, the system uses a grammar analysis tool to construct a syntax tree from the manuscript text, extracting grammatical nodes such as subject-verb boundaries and verb-object boundaries as non-punctuation pause positions. Simultaneously, the system uses semantic analysis to identify sentence group boundaries at topic transitions, logical transition points at causal or adversative relationships, and sentence initiation points at the beginning of each paragraph where momentum is needed.

[0030] Secondly, the extracted punctuation stop positions and non-punctuation stop positions are marked according to the stylistic features of the language, and legal stop anchors of different anchor types are generated.

[0031] It should be noted that different genres (such as news broadcasting, literary recitation, and commentary) handle pauses differently. For example, news broadcasting requires short, brisk pauses, while literary recitation allows for longer, more emotional pauses. Therefore, the system categorizes and labels the extracted pause positions based on the genre characteristics of the text, generating valid pause anchors with different anchor type identifiers.

[0032] In this embodiment, anchor point types include, but are not limited to, sentence-end anchor points, sentence-middle comma anchor points, tone transition anchor points, and paragraph-start anchor points.

[0033] Finally, the permissible pause duration, initial effective advance zone, and pre-reading requirement parameters are pre-configured for the legal anchor points to match their corresponding anchor point types.

[0034] The allowable pause duration refers to the maximum pause duration permitted for this anchor point type. It limits the upper limit of the allowed standard pauses at different anchor points and can be set based on the statistical distribution of pause durations in standard broadcast samples under the same anchor point type, for example, by taking the upper quantile value. The sentence-end anchor point can be set to 1200ms, and the sentence-middle comma anchor point can be set to 600ms to adapt to the broadcast pattern of longer sentence-end pauses and shorter sentence-middle pauses. The initial effective propulsion zone refers to a basic length interval along the text propulsion direction from the anchor point position. It represents the minimum text distance that the trainee needs to smoothly and continuously cross when resuming broadcast after a pause. It is used to filter tentative short sounds and can be set according to the difference in character propulsion length between tentative reading and normal resumption broadcast. For example, the sentence-end anchor point can be set to 5 characters, and the sentence-middle comma anchor point can be set to 3 characters to avoid short sounds accidentally triggering scrolling recovery. The pre-reading requirement parameter refers to the minimum amount of text remaining in the prompting window that students can pre-read according to the anchor point type. It can be set according to the window size, font size, speech rate and anchor point type. For example, the sentence-end anchor point can be set to 15 characters and the sentence-middle comma anchor point can be set to 10 characters to ensure that students have enough space to pre-read subsequent text when they resume the speech.

[0035] In this embodiment, the allowable pause duration of the sentence-end anchor point is configured to 1200ms, and the initial effective advance zone is configured to be 5 characters wide (i.e., The pre-read requirement parameter is configured to 15 characters; the allowable pause duration for comma anchors in a sentence is configured to 600ms; and the initial effective advance zone is configured to a width of 3 characters (i.e., The pre-read requirement parameter is configured to 10 characters.

[0036] For example, in the manuscript text of this embodiment, the system extracted and generated a total of 18 valid anchor points, including 4 sentence-end anchor points (located at characters 48, 96, 150, and 200 respectively), 8 sentence-intermediate comma anchor points, 3 tone transition anchor points, and 3 paragraph initiation anchor points. Taking the sentence-intermediate comma anchor point located at character 105 as an example... For example, its attribute parameters are: anchor point position = 105, anchor point range = [102, 108] (that is, the pauses within the range of characters 102 to 108 belong to this anchor point), allowable pause duration = 600ms, initial effective advance area = 3 characters, and pre-read requirement parameter = 10 characters.

[0037] The pause detection module is used to acquire the broadcast speech and align it with the text of the script to be broadcast to determine the current broadcast position; when the acoustic feature changes extracted from the broadcast speech meet the preset pause conditions, a pause event is generated and the broadcast position before the pause is recorded.

[0038] It should be noted that this module uses a speech-text alignment algorithm to map the time position in the speech stream to the character coordinate position in the text in real time, thereby determining the current broadcast position. ; Detect pauses during broadcasting by analyzing acoustic features.

[0039] The speech-to-text alignment is based on frame-level alignment information output by the acoustic model, mapping the timestamp of each frame in the speech stream to the corresponding character coordinates in the text. This mapping function is denoted as... ,in, The timestamp of the current audio frame. This is the set of text coordinates for the manuscript. The end-to-end calculation delay of this alignment process needs to be controlled within 50ms to meet the real-time requirements of closed-loop control.

[0040] The preset pause condition is determined using an energy-zero-crossing rate dual-threshold detection mechanism: when the short-time energy of the speech frame... continuously below the energy threshold Furthermore, the short-term zero-crossing rate is continuously lower than the zero-crossing rate threshold. The duration exceeds the minimum pause determination time. When the time frame is set to 200ms, the system determines that a pause event has occurred. The system then locks the text coordinates of the last valid audio frame before the pause and records it as the position that was broadcast before the pause. .

[0041] The energy threshold is used to determine whether the speech energy has dropped to a pause candidate state. It can be set based on the short-term energy distribution of background noise segments and normal broadcast segments in the training environment. For example, the average background noise energy plus a safety margin can be used to reduce the interference of environmental noise on pause detection. The zero-crossing rate threshold is used to help determine whether the current speech frame still has valid speech activity. It can be set based on the statistical distribution of the zero-crossing rate of silent segments, breathing segments, and normal speech segments to avoid false pauses caused solely by low-energy noise or breath sounds. The minimum pause determination duration is used to exclude short-term low-energy segments caused by normal syllable gaps. It can be set based on the difference between the duration of syllable gaps and actual pauses. For example, it can be set to 200ms so that the system only generates a pause event after the low-energy, low-zero-crossing rate state has lasted for a sufficient period of time.

[0042] In this embodiment, the student paused when the speech reached the 105th character. The system detected that the speech energy dropped below the threshold and remained so for more than 200ms, generating a pause event. And record the location that was broadcast before the pause. The timestamp of the pause event is recorded as follows: =32.5s (time starts from the start of the broadcast).

[0043] Anchor point matching module is used to select candidate anchor points from the legal pause anchor points based on the position already broadcast before the pause, and calculate the belonging degree of the candidate anchor points based on the distance between the position already broadcast before the pause and the position of the candidate anchor point, whether the pause event falls within the scope of the candidate anchor point, the matching degree between the pause duration and the allowable pause duration of the candidate anchor point, and whether the remaining effective text volume of the obtained prompting window meets the pre-reading requirements of the candidate anchor point, and the one with the highest belonging degree is the main anchor point;

[0044] It should be noted that the function of this module is to find the anchor point that best matches the current pause event from the pre-generated set of valid pause anchor points after a pause event is detected, and use it as a reference for subsequent scrolling control. Since the pause position may fall exactly near a valid pause anchor point, or it may fall in a non-standard position between two anchor points, it is necessary to use multi-factor comprehensive calculation to determine the anchor point to which the pause event is most likely to belong.

[0045] The calculation of the affiliation degree of candidate anchor points, with the highest affiliation degree serving as the primary anchor point, includes:

[0046] First, the distance between the previously broadcast position and the candidate anchor position, the matching result of whether the pause event falls within the scope of the candidate anchor, the matching degree between the pause duration and the allowable pause duration of the candidate anchor, and the satisfaction degree of the remaining effective text volume in the prompting window and the pre-reading requirement parameters of the candidate anchor are converted into corresponding factor values, and then weighted and summed according to preset weights to obtain the belonging degree of each candidate anchor.

[0047] Specifically, for each candidate anchor point The system calculates four factor values ​​respectively: distance factor The location that was broadcast before the pause With anchor point position The reciprocal of the character distance is normalized; the closer the distance, the higher the factor value. Range matching factor If the broadcast location before the pause falls within the anchor point's effective range, the value is 1; otherwise, the value is 0. Duration matching factor. Calculate the ratio of pause duration to the allowable pause duration at the anchor point. The closer the ratio is to 1 (i.e., the better the pause duration matches the allowable duration), the higher the factor value. The portion exceeding the allowable duration will result in a lower factor value; pre-read satisfies the factor. —Calculate the ratio of the remaining effective text in the prompting window to the anchor pre-reading requirement parameter. If the ratio is greater than or equal to 1, the factor value is taken at its maximum value; otherwise, it is reduced proportionally. The formula for calculating the affiliation degree is:

[0048] ;

[0049] in, For the preset weights, satisfy In this embodiment, the preset weight is set to... =0.15. The distance factor is weighted the highest because the spatial distance between the pause position and the anchor point is the most direct physical basis for judging the attribution relationship; the range matching factor is weighted the second highest, used to distinguish whether the pause occurs within the grammatical jurisdiction of the anchor point; the duration matching factor is weighted the third highest because the pause duration can help distinguish different types of anchor points but is easily affected by the student's broadcast status; the pre-reading satisfaction factor is weighted the lowest because the remaining effective text in the prompting window mainly reflects the subsequent pre-reading conditions, and its direct judgment effect on the anchor point attribution relationship is relatively weak.

[0050] For example, in this embodiment, the student's location was announced before the pause. The system selects the distance from 18 legal anchor points. The three most recent candidate anchor points are used for calculation: anchor point (Comma anchor point in the sentence, located at character 105) Anchor point (Sentence-end anchor point, located at character 96) Anchor point (The comma anchor point in the sentence is located at character 112). After four-factor weighted summation calculation, The degree of belonging is 0.92. The degree of belonging is 0.41. The degree of belonging is 0.35.

[0051] Secondly, it is determined whether the difference in affiliation degree between the first candidate anchor point with the highest affiliation degree and the second candidate anchor point with the second highest affiliation degree is less than a preset difference threshold.

[0052] Finally, if the difference in affiliation degree is less than the preset difference threshold, the first candidate anchor point and the second candidate anchor point are merged together to determine a composite anchor point; in the step of determining the pause nature, the weight matrix corresponding to the candidate anchor points included in the composite anchor point is calculated by parameter fusion according to the affiliation degree ratio.

[0053] The preset difference threshold (e.g., set to 0.15) characterizes the boundary at which the system determines whether the competitive relationship between two candidate anchor points is ambiguous. This threshold is set because when the difference in the degree of belonging between two candidate anchor points is extremely small, simply selecting the one with the highest degree of belonging may lead to misjudgment. By merging the two into a composite anchor point and fusing their respective discrimination weight matrices according to the degree of belonging in the subsequent determination of pause properties, the robustness of the determination can be improved.

[0054] In this embodiment, and The difference in membership degree is 0.92 - 0.41 = 0.51 > 0.15, therefore there is no need to merge them, and they can be directly used as... The main anchor point.

[0055] The repair identification module is used to obtain the resumed broadcast position after the pause ends, calculate its positional offset from the broadcast position before the pause, and generate repair behavior feature data; when the positional offset indicates a re-reading, the re-reading point is recorded.

[0056] It should be noted that the purpose of this module is to calculate the resumption position of the broadcast after the student finishes pausing and resumes speaking. Location announced before the pause Spatial offset between This module accurately identifies the type of corrective action taken by the learner. The sign and magnitude of the positional offset directly reflect the learner's error-correction strategy after a pause: a positive value indicates moving forward, zero or a small positive value indicates repeating in place, and a negative value indicates regressing from the previous context. This module categorizes corrective actions into four subtypes, each corresponding to different physical meanings and subsequent processing strategies, achieving a refined classification of learners' corrective actions after non-standard pauses.

[0057] The generated repair behavior feature data includes:

[0058] If the position offset moves back towards the beginning of the text and exceeds the preset tolerance range, then repair behavior feature data representing the rereading is generated;

[0059] Specifically, when the position offset And| Time (of which) To pre-determine a tolerance range to eliminate character-level positioning errors in speech-text alignment, this range can be set based on the maximum or upper quantile positioning deviation of the alignment algorithm in the verification samples. In this embodiment, it is set to 2 characters to avoid misinterpreting normal coordinate jitter as a repetition. If the system determines that the student has performed a repetition, meaning the student has returned to a previous position and restarted reading, the system will restore the original reading position. Record as the landing point of the retreat This pullback point will be used as the starting point for judgment in the subsequent scrolling control module. Preset tolerance range. The setting is to allow for minor positioning errors at character boundaries in the speech-text alignment algorithm, and to avoid misjudging normal in-place restoration as a reversal behavior due to positioning jitter.

[0060] If the positional offset falls within the preset tolerance range of the position already broadcast before the pause, and the restored broadcast text coincides with the last segment of text before the pause, then repair behavior feature data representing repeated reading is generated; that is, when | Furthermore, when the text content being resumed overlaps with several characters from the last text read before the pause, the system determines that the learner has performed a repetition reading action. Repetition reading typically occurs when learners feel uncertain about their own pauses and choose to repeat reading from near the pause location to re-establish the rhythm of their speech.

[0061] If the restored text contains partial replacements and the positional offset subsequently advances forward, then corrective behavior feature data representing corrective restatement is generated; that is, after restoring the repetition, the learner replaces and corrects individual words that were previously spoken before the pause, and then continues speaking forward. This behavior usually occurs when the learner realizes that there were inappropriate words or errors in the preceding text.

[0062] If the resumed broadcast position directly connects to the previously broadcast position before the pause and there is no rollback, then repair behavior feature data representing no repair behavior is generated. That is, the position offset. ≈0 and the resumed text directly connects to the text before the pause, without any backtracking or replacement. This behavior indicates a normal pause recovery, typically corresponding to a compliant breathing pause or grammatical pause.

[0063] In this embodiment, after the student pauses at the 105th character, the recitation resumes from the 98th character. The position offset at this point is... , | Furthermore, since the offset direction is backwards from the preceding text, the system generates repair behavior feature data representing the retraction and rereading, and records the retraction landing point. =98. The physical meaning of this pullback point is: the student chooses to restart reading from the 98th character, therefore the teleprompter system's scrolling recovery judgment should be based on this position rather than the 105th character position before the pause.

[0064] The advance adjustment module is used to obtain recent pause history data to calculate the broadcast stability; when the broadcast stability is lower than a preset threshold, the initial effective advance area of ​​the main anchor point is extended along the text advance direction according to the difference between the broadcast stability and the preset threshold to obtain a dynamic effective advance area;

[0065] It should be noted that this module dynamically adjusts the length of the advance zone required for scrolling recovery based on the student's recent broadcast stability. When a student frequently exhibits abnormal pauses or rereading behaviors, their broadcast stability decreases, and the system correspondingly increases the length of the effective advance zone. This requires the student to smoothly and continuously traverse a longer text distance after resuming broadcasting before triggering scrolling recovery. Conversely, when the student's broadcasting is stable, the effective advance zone maintains a shorter initial length to avoid redundant delays in normal broadcasting. By dynamically extending the effective advance zone, the system can effectively filter out interference from tentative, short vocalizations on scrolling recovery judgments, only resuming scrolling when the student demonstrates a genuine and continuous intention to advance their broadcast.

[0066] The calculation of broadcast stability by obtaining recent historical data on pauses includes:

[0067] The first step is to extract recent historical pause data containing multiple historical pause events within a preset time window before the current pause event occurred;

[0068] Specifically, the system uses the time of occurrence of the current pause event. As the endpoint, backtrack by a preset time window. (In this embodiment, it is set to 60 seconds), all historical pause events occurring within this time window are extracted as recent pause history data. Preset time window The time limit is set at 60 seconds because this duration can cover the typical broadcasting cycle of a trainee in a medium-length news segment. It can reflect the recent broadcasting trend of the trainee, and will not introduce outdated historical data due to an excessively long window.

[0069] In this embodiment, Within the 60-second window preceding 32.5 seconds (the actual window is 0 to 32.5 seconds since the broadcast started at 32.5 seconds), the system recorded a total of 5 historical pause events, of which 3 were marked as abnormal pauses and 2 were marked as normal pauses. One of the abnormal pauses also included a rollback and reread flag.

[0070] The second step is to calculate the time decay weight of each historical pause event according to its temporal position in the recent pause history data; wherein the value of the time decay weight decreases as the time interval from the occurrence of the historical pause event to the current time increases; the calculation formula is as follows:

[0071] ;

[0072] in, For the first The time decay weight of each historical pause event For the first The moment when a historical pause occurred. This refers to the moment when the current pause event occurred. The time-related forgetting decay constant (dimension 1) In this embodiment, it is set to Time-related forgetting decay constant. It represents the rate at which the influence of historical events diminishes. The larger the value, the faster the impact of historical events decays, and the more the system tends to focus only on recently occurring events. In this embodiment, =0.05 means that an event that occurred 20 seconds ago has its weight decayed to 0.05. That is, only about 36.8% of the influence is retained.

[0073] For example, the times of occurrence of the five historical pause events are as follows: The corresponding time decay weights are as follows: , , , , .

[0074] The third step involves using the time decay weight to perform a weighted summation of the abnormal pause markers in each historical pause event to obtain the weighted abnormal pause rate; using the time decay weight to perform a weighted summation of the rollback and reread markers in each historical pause event to obtain the weighted rollback frequency; and finally, the weighted abnormal pause rate... The calculation formula is:

[0075] ;

[0076] in, It is an indicator function; This indicates the first [number] in the recent pause history data. A historical pause; Represents a predefined set of abnormal pause events; when The condition is met (i.e., the first) When an event is an abnormal pause, the indicator function... The value is 1 if the value is not 0 otherwise. The proportion of abnormal events in the historical pause events is calculated by time decay weighting, and the more recent the abnormal event, the greater its contribution to the current weighted abnormal pause rate.

[0077] Weighted drawdown frequency The calculation formula is:

[0078] ;

[0079] in, The dimensions are This represents the frequency of weighted drawdown behavior per unit time. This indicates the first [number] in the recent pause history data. A historical pause; Represents a predefined set of rollback and reread events; when The condition is met (i.e., the first) When an event belongs to the rollback / reread behavior, the indicator function... The value is 1 if it is set to 1, and 0 otherwise. This indicates the duration of the preset time window used when extracting the recent pause history data.

[0080] In this embodiment, among the five historical pause events, events 1, 3, and 4 are abnormal pauses, events 2 and 5 are normal pauses, and event 4 also includes a rollback / reread flag. Weighted abnormal pause rate. Weighted drawdown frequency .

[0081] The fourth step is to extract the maximum number of consecutive abnormal pauses from the recent historical pause data;

[0082] Specifically, in the recent historical data of pauses, the time sequence of all events is traversed to count the maximum number of consecutive occurrences of abnormal pause markers. In this embodiment, the abnormal pause identifier sequence for the five events is [1, 0, 1, 1, 0], and the maximum number of consecutive abnormal events is... =2 (the 3rd and 4th events are consecutive abnormal pauses).

[0083] Fifth, the weighted abnormal pause rate, the weighted pullback frequency, and the maximum number of consecutive broadcasts are weighted according to preset weight parameters to obtain the broadcast stability; wherein, the value of the broadcast stability decreases as the weighted abnormal pause rate or the weighted pullback frequency increases. Broadcast Stability The calculation formula is:

[0084] ;

[0085] in, The dimensionless gain weighting parameter for the abnormal pause rate (set to in this embodiment) (to make stability significantly decrease when abnormal pauses occur frequently), used to control the intensity of the impact of abnormal pause rate on broadcast stability, which can be set according to the sample relationship between abnormal pause rate and scrolling error trigger probability; The gain weighting parameter for the pullback frequency (with dimensions T, set to [value] in this embodiment) To improve the system's sensitivity to rollback and reread behavior, the rollback frequency is converted into a dimensionless influence term, which can be set according to the length of the statistical time window and the degree of influence of rollback and reread on window misalignment. The dimensionless gain weight parameter for the number of consecutive anomalies (set to in this embodiment) This setting (to reflect the risk of continuous anomalies while avoiding over-amplification of occasional continuous pauses) is used to control the impact of continuous stalls on broadcast stability and can be set based on the impact of continuous abnormal pauses on the probability of failure to resume broadcasting. When there are no recent abnormal pauses and no rollback behavior... ,at this time This represents the ideal steady state; when the abnormal pause rate and pullback frequency continue to increase, A value approaching 0 indicates a severely unstable state.

[0086] Specifically, broadcast instability is captured from three dimensions (abnormality rate, pullback frequency, and number of consecutive abnormalities)—abnormality rate reflects the overall trend, pullback frequency reflects the severity, and the number of consecutive abnormalities reflects whether it has fallen into a stuck loop.

[0087] In this embodiment, The broadcast stability value is approximately 0.31, indicating that the trainee's current broadcast stability is low.

[0088] The dynamically effective propulsion zone includes:

[0089] First, calculate the difference between the broadcast stability and the preset threshold;

[0090] Among them, the preset threshold (In this embodiment, it is set to 0.75 to distinguish between stable and unstable broadcast states; this value can be set based on the broadcast stability distribution of stable and unstable samples.) The characterization system determines that the student's broadcast state is within the lower bound of the safe and stable range. When the broadcast stability... When the system considers the trainee's broadcast status to be within the normal range, there is no need to extend the effective advance zone; when At that time, the system calculates the difference. As a quantitative indicator to measure the degree of instability, the effective propulsion zone is extended accordingly. In this embodiment, =0.75-0.3065=0.4435.

[0091] Secondly, using the difference as input, a piecewise mapping function is used to calculate the effective propulsion zone extension. The piecewise mapping function is configured as follows: when the difference is within a first preset interval, the product of the difference and a first adjustment coefficient is output as the effective propulsion zone extension; when the difference is within a second preset interval greater than the first preset interval, the product of the difference and a second adjustment coefficient is output as the effective propulsion zone extension. The second adjustment coefficient is greater than the first adjustment coefficient, so that the marginal increase in the effective propulsion zone extension is greater when the broadcast stability is lower.

[0092] The specific implementation of the piecewise mapping function is as follows: Let the first preset interval be [0, ... The second preset interval is [ The first adjustment factor is The second adjustment factor is ( The effective extension of the propulsion zone is then determined. The calculation formula is:

[0093] like , ;

[0094] like , ;

[0095] in, The boundary value between the first preset interval and the second preset interval (set to in this embodiment) This setting is used to distinguish between slightly unstable and obviously unstable states, and can be set according to the statistical relationship between the difference in broadcast stability and the scrolling false trigger rate. The first adjustment factor (set to in this embodiment) (number of characters / unit difference) The second adjustment factor (set to in this embodiment) (Character / unit difference). The second adjustment coefficient is set greater than the first because when the broadcast instability is minor (small difference, within the first preset range), the system extends the advance zone with a gentler coefficient to avoid overreacting to minor fluctuations; when the broadcast instability worsens (larger difference, entering the second preset range), the system extends the advance zone with a larger coefficient to ensure sufficient scroll recovery constraints are applied to severely unstable states. This segmented design implements an adaptive control strategy that provides a gentle response to minor fluctuations and strong constraints for severe crashes.

[0096] In this embodiment, The difference is within the second preset interval, therefore =25×0.4435≈11.09, rounded down to 11 characters.

[0097] Finally, the length of the initial effective propulsion zone is added to the extension of the effective propulsion zone to obtain the dynamic effective propulsion zone. The calculation formula is:

[0098] ;

[0099] in, The initial effective propulsion zone length corresponding to the main anchor point. In connection with the preceding steps, the main anchor point in this embodiment... The initial effective propagation zone is the comma anchor point in the sentence. One character, therefore =3 + 11 = 14 characters. This means that after resuming playback, the student needs to smoothly and continuously move across a distance of 14 characters before the system will consider it a valid playback advance and resume normal scrolling. This 14-character advance distance is sufficient to filter out the student's tentative, short vocalizations (tentative vocalizations usually only cover a distance of 1 to 3 characters before stopping), thus effectively avoiding accidental scrolling triggers caused by tentative vocalizations.

[0100] The scrolling control module is used to determine whether the repair behavior feature data indicates a re-reading and re-rolling, and to determine whether the updated current broadcast position continuously crosses the dynamic effective advancement area from the determination starting point, and to output scrolling control parameters according to the determination result.

[0101] It should be noted that this module dynamically switches the judgment benchmark for scrolling recovery to a spatial position that matches the student's actual error correction intention, based on the differences in the types of corrective behavior feature data. When a student retracts and rereads, the system resets the judgment starting point from the main anchor point to the retraction landing point (i.e., the position where the student actually restarts reading), ensuring the consistency of the scrolling recovery judgment benchmark with the student's current actual broadcasting position in physical space.

[0102] Wherein, if the repair behavior feature data represents a re-read, then the re-read point is used as the starting point for judgment; otherwise, the main anchor point position is used as the starting point for judgment, including:

[0103] If the repair behavior feature data indicates a re-reading, then the starting point of the judgment is determined as the re-reading point, the re-reading distance between the re-reading point and the main anchor point is obtained, and the re-reading compensation amount is calculated based on the re-reading distance. The re-reading compensation amount is then superimposed on the dynamic effective advancement zone so that the termination boundary of the dynamic effective advancement zone is offset and adapted relative to the re-reading point.

[0104] Specifically, when repairing behavioral feature data representations and rereading them, the system performs the following operations: First, it sets the starting point of the judgment from the main anchor point position. Reset to the landing point Secondly, calculate the pullback distance. Then, the drawdown compensation amount is calculated based on the drawdown distance. (in The pullback compensation coefficient is set to [value] in this embodiment. This means that 50% of the pullback distance is used as an additional advance zone compensation. This coefficient is set considering that students usually re-establish part of the speech flow rhythm after pullback and rereading, so it is not necessary to perform a complete advance verification for the entire pullback distance. Finally, the pullback compensation is superimposed on the dynamic effective advance zone, making the termination boundary of the dynamic effective advance zone... The corresponding offset is made relative to the withdrawal landing point. The formula for calculating the termination boundary is:

[0105] ;

[0106] Specifically, in the scenario of rewinding and rereading, trainees need to start from the rewind landing point. Starting from there, it smoothly and continuously traverses the initial effective propulsion zone. Effective extension of the promotion area and drawdown compensation amount Only after calculating the distance of the sum of the three factors does the system determine that the broadcast resumption is valid.

[0107] In this embodiment, the retraction landing point Main anchor point location retracement distance Characters, drawdown compensation amount Rounded down to 4 characters. Therefore, the termination boundary. =98+3+11+4=116, meaning the student needs to continuously read from the 98th character to the 116th character before the system resumes normal scrolling.

[0108] If the repair behavior feature data does not characterize a rollback behavior, then the starting point for judgment is determined as the main anchor point position. The main anchor point position... As the starting point for judgment, the termination boundary is at this point. This is because, in non-retreat scenarios, the trainee's resume broadcast position is spatially close to (or directly connected to) the main anchor point position, eliminating the need for starting point reset.

[0109] The scrolling control parameters output based on the judgment result include:

[0110] If it is determined that the updated current broadcast position has crossed the termination boundary of the dynamic effective propulsion area, and no secondary pause event is detected by acoustic feature changes during the crossing period, and no new position offset is obtained, then scrolling control parameters for restoring the normal scrolling speed are generated.

[0111] Specifically, the system continuously monitors and updates the currently broadcast positions. When detected At this time, two additional conditions must be met simultaneously: first, no new pause events are detected during the entire process from the starting point to the termination boundary (i.e., the voice energy does not drop below the pause judgment threshold again); second, no new positional offset is acquired (i.e., no new pullback or repair behavior occurs). Only when the above three conditions are met simultaneously will the system generate scrolling control parameters to restore the normal scrolling speed and output smooth servo scrolling commands. Among them, the pause judgment threshold is composed of the energy threshold, the zero-crossing rate threshold, and the minimum pause judgment duration, and is used to comprehensively judge pause events.

[0112] In this embodiment, the trainees retreat from the landing point After the broadcast resumes, the system continuously monitors the student's current broadcast position. When the student broadcasts continuously to... (i.e., crossing the termination boundary) =116), and if no secondary pause or new rollback occurs during this period, the system determines that the broadcast recovery is effective and generates scrolling control parameters to restore the normal scrolling speed.

[0113] If a second pause event or a new rollback reread is detected before the updated current broadcast position reaches the termination boundary, the current scrolling control parameters are maintained, and the detected event is used as a new pause event to repeatedly trigger the propulsion adjustment module to update the recent pause history data and recalculate the dynamic effective propulsion area.

[0114] Specifically, when a trainee pauses or retraces while crossing the dynamic effective advancement zone, the system does not unlock the scrolling screen. Instead, it incorporates the newly detected event into the recent pause history data and re-triggers the advancement adjustment module to calculate the updated broadcast stability and dynamic effective advancement zone. Since newly added abnormal pause events further reduce broadcast stability, the effective advancement zone lengthens, resulting in a longer newly calculated dynamic effective advancement zone than before, creating an increasing constraint. This closed-loop mechanism ensures that the system will not mistakenly trigger scrolling recovery when the trainee remains in an unstable state, providing multiple layers of safety protection.

[0115] For example, if the student pauses again when reading from the 98th character to the 110th character ( The system adds the second pause event to the recent pause history data and recalculates the broadcast stability. Due to the addition of an abnormal pause event, the broadcast stability further decreases (e.g., from 0.31 to 0.24), the effective propulsion zone lengthens accordingly, and the dynamic effective propulsion zone becomes longer. Trainees need to traverse a longer distance after the second broadcast resumption before triggering the scrolling recovery.

[0116] The system also includes a pause detection module, which performs the following processing after setting the highest degree of belonging as the main anchor point:

[0117] Step 1: Identify the anchor type of the main anchor point and call the discrimination weight matrix corresponding to the anchor type;

[0118] Specifically, different types of anchors have different criteria for determining the nature of pauses. For example, the discriminant weight matrix corresponding to an anchor at the end of a sentence will give a higher tolerance for the duration of the pause (because the pause at the end of a sentence is naturally longer), while the discriminant weight matrix corresponding to an anchor at a comma in the middle of a sentence will have a lower tolerance for the duration of the pause.

[0119] In this embodiment, the main anchor point For comma anchor points in the sentence, the system calls the corresponding discrimination weight matrix. .

[0120] Step 2: Obtain speech flow change data before and after the pause event and extract speech flow continuity;

[0121] Among them, speech continuity reflects whether the transition of speech before and after a pause is natural and smooth. The system calculates speech continuity by comparing the consistency of acoustic features such as speech rate, pitch, and energy in 500ms speech segments before and after a pause. The higher the consistency, the higher the continuity, indicating that the pause is more likely to be a natural and normal pause rather than an abnormal interruption.

[0122] Step 3: Subtract the pre-reading requirement parameter of the main anchor point from the remaining effective text in the prompting window to obtain the pre-reading matching degree;

[0123] Among them, the pre-read matching degree reflects whether the amount of text available for pre-reading in the current prompting window meets the pre-reading requirements of the main anchor point.

[0124] Main anchor point in this embodiment The pre-read requirement parameter is 10 characters. The prompting window is currently displaying the 140th character. Therefore, the remaining effective text is 140-105=35 characters, and the pre-read matching degree is 35-10=25 (a positive value indicates that the requirement is met).

[0125] Step 4: Determine the intensity of the repair behavior based on the subtype of the repair behavior feature data; wherein, no repair behavior corresponds to the first intensity value, repeated reading corresponds to the second intensity value, repair reread corresponds to the third intensity value, and rollback reread corresponds to the fourth intensity value, and the values ​​of the first intensity value, the second intensity value, the third intensity value and the fourth intensity value increase sequentially.

[0126] Specifically, the intensity values ​​of the four repair behavior subtypes are arranged from low to high according to the degree of disruption to the broadcast coherence: no repair behavior (lowest intensity value, for example, set to 0.0) represents a completely normal interruption recovery; repeated reading (intensity value set to 0.3) represents a mild interruption of speech flow; repairing re-speech (intensity value set to 0.6) represents a moderate content correction; and retraction re-read (highest intensity value, set to 1.0) represents a severe broadcast regression.

[0127] In this embodiment, the repair behavior feature data represents the rollback reread, so the repair behavior intensity is set to 1.0.

[0128] Step 5: Using the discrimination weight matrix, perform weighted fusion calculation on the attribution degree of the main anchor point, the speech flow continuity, the repair behavior intensity, and the pre-read matching degree to generate standardized pause characterization value and abnormal pause characterization value.

[0129] Specifically, let the input feature vector be X=[G, Fc, Mr, Pm] (corresponding to the normalized values ​​of attribution degree, speech continuity, repair behavior intensity, and preread matching degree, respectively), and the discriminant weight matrix W be a 2×4 matrix, then the output vector Y=W·X=[ ,in To standardize the values ​​representing disconnection, This is a value representing an abnormal pause.

[0130] Furthermore, the pause discrimination module is also used to compare the standard pause characterization value and the abnormal pause characterization value with the first classification threshold and the second classification threshold respectively, to determine the pause nature of the current pause event, and to store the pause nature in the recent pause history data;

[0131] Among them, the first classification threshold Second classification threshold These are the decision boundaries for determining standard pauses and abnormal pauses, respectively. and When, it is judged as a standard disconnection; when If a pause occurs, it is considered an abnormal pause. The determination result is stored in the recent pause history data, and this pause type identifier will be used as an abnormal pause identifier in the calculation of the weighted abnormal pause rate.

[0132] The first classification threshold can be set based on the statistical distribution of the characteristic values ​​of the standard pause samples. For example, it can be set to a lower limit that covers most standard pause samples to ensure that the standard pause determination has sufficient confidence. The second classification threshold can be set based on the statistical distribution of the characteristic values ​​of the abnormal pause samples. For example, it can be set to a lower limit that covers most abnormal pause samples to improve the reliability of abnormal pause identification.

[0133] The scrolling control module is also used to output corresponding scrolling control parameters based on the pause characteristics and the pre-read matching degree:

[0134] If the pause is a standard pause and the pre-read matching degree is greater than the preset matching threshold, then scrolling control parameters that maintain the current normal scrolling speed are generated; that is, when the pause is determined to be a compliant grammatical pause or breathing pause, and there is sufficient remaining text in the prompting window, the system maintains the current scrolling speed unchanged without any adjustment.

[0135] The preset matching threshold is used to determine whether the remaining text in the prompting window meets the pre-reading requirements. It can be set according to the calculation method of the pre-reading matching degree. When the pre-reading matching degree is the remaining effective text minus the pre-reading requirement parameter, the preset matching threshold can be set to 0 or slightly greater than 0 so as to trigger local buffering when the remaining text is insufficient.

[0136] If the pause is a standard pause and the pre-read matching degree is less than or equal to the preset matching threshold, then scrolling control parameters for performing local buffering are generated to maintain an effective reading area; that is, although the pause itself is standard, if there is insufficient amount of text available for pre-reading in the prompting window, the system slows down the scrolling of the text upward by reducing the scrolling speed (performing local buffering) to ensure that the learner still has enough text available for pre-reading in the window.

[0137] If the pause is deemed an abnormal pause, scrolling control parameters are generated to either lock the current scrolling position or maintain a low scrolling speed. Specifically, when a pause is determined to be abnormal, the system immediately locks the current scrolling position or reduces the scrolling speed to an extremely low value to prevent the teleprompter from continuing to scroll forward before the student resumes effective communication, thus avoiding visual misalignment. This locked state will remain until the scrolling control module's continuous overshoot judgment logic is verified successfully.

[0138] Exemplary computer-readable medium:

[0139] Embodiments of this application may also be computer-readable storage media storing computer program instructions that, when executed by a processor, cause the processor to perform the steps described in the "Exemplary Systems" section above according to the various embodiments of this application.

[0140] Computer-readable storage media may take the form of any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may, for example, include, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: electrical connections having one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0141] The basic principles of this application have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in this application are merely examples and not limitations, and should not be considered as essential features of each embodiment of this application. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not limit the application to the necessity of employing the aforementioned specific details for implementation.

[0142] The block diagrams of devices, apparatuses, devices, and systems involved in this application are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, devices, and systems can be connected, arranged, and configured in any manner. Words such as “comprising,” “including,” “having,” etc., are open-ended terms meaning “including but not limited to,” and are used interchangeably with them. The terms “or” and “and” as used herein refer to the terms “and / or,” and are used interchangeably with them unless the context clearly indicates otherwise. The term “such as” as used herein refers to the phrase “such as but not limited to,” and is used interchangeably with it.

[0143] It should also be noted that in the apparatus, equipment, and methods of this application, the components or steps can be disassembled and / or recombined. These disassemblies and / or recombinations should be considered as equivalent solutions of this application.

[0144] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use this application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of this application. Therefore, this application is not intended to be limited to the aspects shown herein, but rather to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0145] The above description has been given for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of this application to the forms disclosed herein. Although numerous exemplary aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations thereof.

Claims

1. An intelligent prompting and adaptive scrolling system for broadcasting and hosting training, characterized in that, The system includes: The anchor point generation module is used to acquire the text of the script to be broadcast and perform structured processing to generate legal pause anchor points that include anchor point position, anchor point range, initial effective advancement zone, and pre-reading requirement parameters. The anchor point range refers to the text range to which the pause belongs, the initial effective advancement zone refers to a basic length interval along the text advancement direction from the anchor point position, and the pre-reading requirement parameters refer to the minimum amount of text remaining in the prompting window that can be pre-read by the trainee for this anchor point type. The pause detection module is used to acquire the broadcast speech and align it with the text of the script to be broadcast to determine the current broadcast position; when the acoustic feature changes extracted from the broadcast speech meet the preset pause conditions, a pause event is generated and the broadcast position before the pause is recorded. Anchor point matching module is used to select candidate anchor points from the legal pause anchor points based on the position already broadcast before the pause, and calculate the belonging degree of the candidate anchor points based on the distance between the position already broadcast before the pause and the position of the candidate anchor point, whether the pause event falls within the scope of the candidate anchor point, the matching degree between the pause duration and the allowable pause duration of the candidate anchor point, and whether the remaining effective text volume of the obtained prompting window meets the pre-reading requirements of the candidate anchor point, and the one with the highest belonging degree is the main anchor point; The repair identification module is used to obtain the resumed broadcast position after the pause ends, calculate its positional offset from the broadcast position before the pause, and generate repair behavior feature data; when the repair behavior feature data indicates a retraction and reread, the retraction landing point is recorded. The advance adjustment module is used to obtain recent pause history data to calculate the broadcast stability; when the broadcast stability is lower than a preset threshold, the initial effective advance area of ​​the main anchor point is extended along the text advance direction according to the difference between the broadcast stability and the preset threshold to obtain a dynamic effective advance area; The scrolling control module is used to determine whether the repair behavior feature data indicates a re-reading and re-reading, and to determine whether the updated current broadcast position continuously crosses the dynamic effective advancement area from the determination starting point, and to output scrolling control parameters according to the determination result. The step of outputting scrolling control parameters based on the judgment result includes: If it is determined that the updated current broadcast position has crossed the termination boundary of the dynamic effective propulsion area, and no secondary pause event is detected by acoustic feature changes during the crossing period, and no new position offset is obtained, then scrolling control parameters for restoring the normal scrolling speed are generated. If a second pause event or a new rollback reread is detected before the updated current broadcast position reaches the termination boundary, the current scrolling control parameters are maintained, and the detected event is used as a new pause event to repeatedly trigger the propulsion adjustment module to update the recent pause history data and recalculate the dynamic effective propulsion area.

2. The intelligent prompting and adaptive scrolling system for broadcasting and hosting training according to claim 1, characterized in that, After the pause ends, the resumed broadcast position is obtained, and its positional offset from the previously broadcast position before the pause is calculated to generate repair behavior feature data, including: If the position offset moves back towards the beginning of the text and exceeds the preset tolerance range, then repair behavior feature data representing the rereading is generated; If the position offset falls within the preset tolerance range of the position already broadcast before the pause, and the restored broadcast text coincides with the last segment of the text before the pause, then repair behavior feature data representing repeated reading is generated. If the restored broadcast text has a local replacement and the position offset is subsequently advanced, then repair behavior feature data representing the repair retelling is generated; If the resumed broadcast position directly connects to the previously broadcast position before the pause and there is no rollback, then repair behavior feature data representing no repair behavior is generated.

3. The intelligent prompting and adaptive scrolling system for broadcasting and hosting training according to claim 2, characterized in that, If the repair behavior feature data indicates a re-read, then the re-read point is used as the starting point for judgment; otherwise, the main anchor point position is used as the starting point for judgment, including: If the repair behavior feature data indicates a re-reading, then the starting point of the judgment is determined as the re-reading point, the re-reading distance between the re-reading point and the main anchor point is obtained, and the re-reading compensation amount is calculated based on the re-reading distance. The re-reading compensation amount is then superimposed on the dynamic effective advancement zone so that the termination boundary of the dynamic effective advancement zone is offset and adapted relative to the re-reading point. If the repair behavior feature data does not characterize the rollback behavior, then the judgment starting point is determined as the main anchor point position.

4. The intelligent prompting and adaptive scrolling system for broadcasting and hosting training according to claim 1, characterized in that, The process of obtaining recent historical data on pauses to calculate broadcast stability includes: Extract recent historical pause data containing multiple historical pause events within a preset time window prior to the current pause event; Based on the temporal position of each historical pause event in the recent pause history data, the time decay weight of each historical pause event is calculated; wherein, the value of the time decay weight decreases as the time interval from the occurrence of the historical pause event to the current time increases; The abnormal pause markers in each historical pause event are weighted and summed using the time decay weight to obtain the weighted abnormal pause rate; the rollback and reread markers in each historical pause event are weighted and summed using the time decay weight to obtain the weighted rollback frequency. Extract the maximum number of consecutive abnormal pauses from the recent historical pause data; The broadcast stability is obtained by weighting the weighted abnormal pause rate, the weighted pullback frequency, and the maximum number of consecutive broadcasts according to preset weight parameters; wherein the value of the broadcast stability decreases as the weighted abnormal pause rate or the weighted pullback frequency increases.

5. The intelligent prompting and adaptive scrolling system for broadcasting and hosting training according to claim 4, characterized in that, The obtained dynamic and effective propulsion zone includes: Calculate the difference between the broadcast stability and the preset threshold; Using the difference as input, a piecewise mapping function is used to calculate the effective propulsion zone extension. The piecewise mapping function is configured as follows: when the difference is within a first preset interval, the product of the difference and a first adjustment coefficient is output as the effective propulsion zone extension; when the difference is within a second preset interval greater than the first preset interval, the product of the difference and a second adjustment coefficient is output as the effective propulsion zone extension. The second adjustment coefficient is greater than the first adjustment coefficient, so that the marginal increase in the effective propulsion zone extension is greater when the broadcast stability is lower. The dynamic effective propulsion zone is obtained by adding the length of the initial effective propulsion zone to the extension of the effective propulsion zone.

6. The intelligent prompting and adaptive scrolling system for broadcasting and hosting training according to claim 1, characterized in that, The system also includes a pause detection module, which performs the following processing after setting the highest degree of belonging as the main anchor point: Identify the anchor point type of the main anchor point and call the discrimination weight matrix corresponding to the anchor point type; Acquire speech flow change data before and after the pause event and extract speech flow continuity; Subtract the pre-reading requirement parameter of the main anchor point from the remaining effective text in the prompting window to obtain the pre-reading matching degree; The intensity of the repair behavior is determined based on the subtype of the repair behavior feature data; wherein, no repair behavior corresponds to the first intensity value, repeated reading corresponds to the second intensity value, repair reread corresponds to the third intensity value, and rollback reread corresponds to the fourth intensity value, and the values ​​of the first intensity value, the second intensity value, the third intensity value and the fourth intensity value increase sequentially. The discrimination weight matrix is ​​used to perform weighted fusion calculations on the attribution degree of the main anchor point, the continuity of the speech flow, the intensity of the repair behavior, and the pre-read matching degree to generate standardized pause characterization values ​​and abnormal pause characterization values.

7. The intelligent prompting and adaptive scrolling system for broadcasting and hosting training according to claim 6, characterized in that, The pause discrimination module is also used to compare the standard pause characterization value and the abnormal pause characterization value with the first classification threshold and the second classification threshold respectively, to determine the pause nature of the current pause event, and to store the pause nature in the recent pause history data; The scrolling control module is also used to output corresponding scrolling control parameters based on the pause characteristics and the pre-read matching degree: If the pause is a normal pause and the pre-read matching degree is greater than the preset matching threshold, then scrolling control parameters to maintain the current normal scrolling speed are generated. If the pause is a normal pause and the pre-read matching degree is less than or equal to the preset matching threshold, then scrolling control parameters for performing local buffering are generated to maintain an effective reading area. If the pause is an abnormal pause, then scrolling control parameters are generated to either lock the current position or maintain a low speed.

8. The intelligent prompting and adaptive scrolling system for broadcasting and hosting training according to claim 1, characterized in that, The generation of the legal disconnection anchor point includes: Punctuation marks are extracted based on their features, and non-punctuation marks, sentence group boundaries, logical transitions, and sentence initiation points are extracted by combining syntax tree sequence features and segment semantic features. The extracted punctuation pause positions and non-punctuation pause positions are marked according to the stylistic features to generate legal pause anchors of different anchor types; The permissible pause duration, initial effective advance zone, and pre-reading requirement parameters are pre-configured for the legal anchor points to match their corresponding anchor point types.

9. The intelligent prompting and adaptive scrolling system for broadcasting and hosting training according to claim 1, characterized in that, When calculating the affiliation degree of candidate anchors, the anchor matching module converts the distance between the previously broadcast position and the candidate anchor position, the matching result of whether the pause event falls within the scope of the candidate anchor's effect, the matching degree between the pause duration and the allowable pause duration of the candidate anchor, and the satisfaction degree of the remaining effective text volume in the prompting window and the pre-reading requirement parameters of the candidate anchor into corresponding factor values, and performs weighted summation according to preset weights to obtain the affiliation degree of each candidate anchor. The term "using the highest degree of belonging as the primary anchor point" includes: Determine whether the difference in affiliation between the first candidate anchor point with the highest affiliation degree and the second candidate anchor point with the second highest affiliation degree is less than a preset difference threshold. If the difference in affiliation degree is less than the preset difference threshold, the first candidate anchor point and the second candidate anchor point are merged together to determine a composite anchor point; in the step of determining the pause nature, the weight matrix corresponding to the candidate anchor points included in the composite anchor point is calculated by parameter fusion according to the affiliation degree ratio.

Images