Artificial intelligence-based audio intelligent regulation method and system, and storage medium

Abstract

The present application relates to the field of artificial intelligence and audio signal processing technology, and provides an audio intelligent regulation method and system based on artificial intelligence and a storage medium, comprising: obtaining usage habit data and basic environment data of a current device; identifying a software that the device is clicking into and determining a basic volume, determining a pre-impact factor according to whether there is other audio playing before entering the software; determining an environment adjustment factor based on the basic environment data, adjusting the basic volume to obtain a preliminary predicted volume in combination with the pre-impact factor and the environment adjustment factor; calculating a personalized bias rate based on the usage habit data and compensating the preliminary predicted volume to obtain a final predicted volume; setting the device volume to the final predicted volume before entering the software; the present application realizes active prediction and non-susceptible regulation before audio playing, and eliminates volume adjustment delay and auditory abruptness.

Description

Technical Field

[0001] This application relates to the fields of artificial intelligence and audio signal processing, and in particular to artificial intelligence-based intelligent audio control methods, systems, and storage media. Background Technology

[0002] Currently, in the field of audio playback control for smart terminal devices, volume adjustment mechanisms generally adhere to a passive response paradigm of "problem occurs—user perception—manual correction." Under this paradigm, the system has no ability to predict the suitability of the audio output. Only when the audio directly impacts the user's hearing with its raw, unadapted intensity, causing jarring disturbances, missing content, or environmental interference, does it passively wait for the user to make compensatory adjustments via physical buttons or screen touch. Within this delayed response window, the user is not only forced to endure the complete chain of discomfort from unsuitable volume to ideal loudness, but also experiences momentary tension and irritation due to the abrupt auditory impact. In continuous use scenarios, this repeated interruption fragments the immersive experience. Even more seriously, when users quickly switch between different multimedia applications, such as exiting a game with high volume and immediately entering a voice call, or jumping from a quiet reading interface to video playback, the system lacks the recognition of the audio transition scene at the moment of application switching. It continues to use the fixed volume value or only relies on slow environmental mapping, resulting in a large jump in audio intensity before and after. The loudness memory of the previous application is brutally cleared, and the new application suddenly releases a volume that is completely disconnected from the current auditory state, creating a terrible experience of "hearing stop" or "sudden explosion".

[0003] Existing automatic volume control solutions, despite incorporating environmental noise monitoring, are almost entirely built upon a single, rigid environmental noise-volume mapping curve. This curve, often calibrated in an ideal laboratory environment, provides only a uniform linear compensation, completely failing to perceive the personalized preferences and dynamically evolving auditory habits that users develop over long-term use. It cannot differentiate between the varying volume adjustment needs of different time contexts, such as day and night, public spaces and bedrooms, nor can it understand that when switching from continuous audio playback to a new task, the user's auditory perception is in a state of inertia, requiring a smooth transition rather than a reset. More importantly, these solutions completely ignore the in-depth analysis of users' manual adjustment behavior—every button press driven by user dissatisfaction carries rich information about the direction, magnitude, and duration of preference bias. Existing systems discard this valuable feedback, preventing the formation of a closed loop. The volume remains at a coarse level of the group average, increasingly distant from the individual's comfort zone. Users are forced to accept the awkward reality of "automatic adjustment that doesn't suit my needs," still requiring frequent manual intervention, rendering the so-called intelligence useless.

[0004] Ultimately, the problem with existing volume control mechanisms lies in the fundamental lag in decision-making timing and the limited and impoverished system of influencing factors. Not only can they not predict and deploy in advance before the sound is played, thus concealing the adjustment action into the user's imperceptible startup interval, but they also lack the ability to proactively model the audio continuity in the application context and the self-evolutionary ability to detect the preference drift hidden in the user's long-term behavior.

[0005] Therefore, the present invention provides an audio intelligent control method, system and storage medium based on artificial intelligence to solve the above technical problems. Summary of the Invention

[0006] This application provides an artificial intelligence-based audio intelligent control method, system, and storage medium, which solves the technical problem that the existing volume control mechanism is fundamentally lagging behind, unable to complete prediction and deployment in advance before the sound is played, and hides the adjustment action into the start-up gap that the user is unaware of.

[0007] To achieve the above objectives, this application adopts the following technical solution: Firstly, it provides an AI-based intelligent audio control method, including: S1: Obtain current device usage habit data and basic environmental data; among which, basic environmental data includes time point and ambient noise value; usage habit data includes the number of manual adjustments and average adjustment range of the device after the final predicted volume adjustment; S2: Identify the software that the device is currently clicking to enter, obtain the historical volume of the software, and determine the base volume of the current software based on the historical volume; determine whether there is other audio playing within n seconds before the device enters the software; if yes, determine the pre-influence factor based on the volume of the playing audio; if no, set the pre-influence factor to 0. S3: Determine the environmental adjustment factor based on the basic environmental data, adjust the basic volume based on the preceding influencing factor and the environmental adjustment factor to obtain the preliminary predicted volume, calculate the personalized bias rate based on the usage habit data, and compensate the preliminary predicted volume to obtain the final predicted volume. S4: Set the device volume to the final predicted volume before entering the software.

[0008] In conjunction with the first aspect mentioned above, one possible implementation involves acquiring current device usage habit data and basic environmental data, including: The system calls the audio acquisition interface of the terminal device to acquire the ambient audio stream in the current environment; the audio acquisition interface includes a microphone. The ambient audio stream is preprocessed to calculate the root mean square (RMS) energy value within a preset time window, and the energy value is converted into a decibel value to obtain the ambient noise value. The preprocessing includes DC component removal and filtering. The preset time window is usually 1 second, but can also be set to other values. Call the system clock interface of the terminal device to obtain the current time. Access the locally stored historical interaction database, which stores audio playback logs of the device within a preset period in the past; wherein the preset period is generally the last 60 days. In the historical interaction database, target log records containing manual adjustment tags are filtered out; wherein, the manual adjustment tags are used to distinguish between volume changes actively triggered by the device and volume changes automatically triggered by the system. Key feature data is extracted from the target log records. This key feature data includes at least: the number of adjustments and the adjustment magnitude; the final volume level: the stable volume value maintained during audio playback after the device stops manual adjustment; the number of adjustments: the cumulative number of times the device triggers the volume adjustment button or gesture within the first few seconds of a single audio playback; and the average adjustment magnitude: the average value calculated by subtracting the final predicted volume from the manually adjusted volume value within a single audio playback, dividing by the final predicted volume, and then averaging the result. This average value characterizes the device's dissatisfaction with the current volume. The first few seconds are manually set, typically 3 seconds.

[0009] In conjunction with the first aspect above, in one possible implementation, determining the base volume of the current software based on the historical volume levels includes: Extract the historical volume within the reference range of the current time, and input the historical volume into the basic volume recognition model to obtain the basic volume of the current software; wherein, the reference range of the current time is a manually set time range centered on the current time, for example, the time range within one hour before and after the current time is marked as the reference range of the current time. The basic volume recognition model is as follows: Extract historical volume values ​​within several reference ranges from the historical volume database, as well as the base volume corresponding to each reference range; wherein, the historical volume database includes historical volume values ​​within several reference ranges, and the base volume set by experts based on the historical volume values; The historical volume levels and baseline volume levels within the reference range are integrated into several sets of training and testing data. The training data is used to train the artificial intelligence model, and the testing data is used to test the trained artificial intelligence model. The artificial intelligence model is then adjusted based on the testing results. Finally, a baseline volume recognition model is obtained, with the historical volume levels within the reference range as the input and the baseline volume as the output. The artificial intelligence model is a nonlinear regression model based on a neural network, implemented using a BP neural network structure and / or an RBF neural network structure.

[0010] In conjunction with the first aspect above, in one possible implementation, determining the pre-influence factor based on the volume of the played audio includes: Extract the volume of the device within the first n seconds before entering the software, and label the volume values ​​from front to back according to time. And the rate of change of volume between adjacent volumes is determined based on the calculation formula (1). ;in, Volume The number, and The value range of is [1, n], where n is obtained by manual setting; The rate of change of volume within the range of change constraints The preceding influence factor is obtained by calculating the average value. The variation constraint range is used to eliminate situations where the sound suddenly increases or decreases excessively. The variation constraint range is set based on how the human ear adapts to different sound changes in the experiment. The calculation formula (1) is: ; In the formula, when At that time, the rate of change of volume The value is 0; This is the decay rate constant, used to control the rate of decay. The value is obtained by conducting A / B testing on a number of test users and statistically analyzing their performance in different... The empirically optimal value obtained after manually adjusting the frequency at the given value, and The value range is [0,1]; Number The corresponding time point, The time point closest to the current time within n seconds before the device enters the software.

[0011] In conjunction with the first aspect mentioned above, one possible implementation involves determining environmental regulation factors based on basic environmental data, including: A1: Extract time points from the basic environmental data within the last m seconds and determine whether the current time falls within a quiet period; if yes, adjust the time influence coefficient corresponding to the time points within the last m seconds. Set to 1; otherwise, set the time influence coefficient corresponding to the most recent m seconds to 1. Set to 0; where m is obtained through manual setting, and is generally set to 5 seconds; the silent time is determined based on the historical ambient noise value of the device. When the ambient noise value of the device is frequently lower than the preset noise threshold for a certain period of time, this period of time is set as the quiet period. A2: Determine if the current time falls within the late night period; if yes, apply the time influence coefficient 2 corresponding to the time point within the most recent m seconds. Set to 1; otherwise, set the time influence coefficient corresponding to the most recent m seconds to 2. Set to 0; the late-night period is determined manually and is generally set to 0:00-6:00. A3: Extract the ambient noise value from the basic environmental data within the last m seconds and obtain the average value of the ambient noise value. Based on the average value Time Influence Coefficient 1 And time influence coefficient 2 Determine environmental regulating factors Among them, environmental regulatory factors Satisfy the calculation formula: , The standard value for environmental noise can be the average value of environmental noise values ​​in the historical data of the current equipment.

[0012] In conjunction with the first aspect mentioned above, one possible implementation involves adjusting the base volume based on antecedent factors and environmental adjustment factors to obtain a preliminary predicted volume, including: Extracting the preceding impact factor Environmental regulating factors and base volume And the preceding influence factors and environmental regulators Normalization is performed based on the aforementioned prior influence factors. Environmental regulating factors and base volume The preliminary predicted volume is determined by formula (2). ; The calculation formula (2) is: ; In the formula, and The proportional adjustment coefficient is determined based on the current software type identifier, for example, , When the software is detected as video software, set... , Because short videos are more sensitive to environmental noise and need to adapt quickly to the environment; when the app is detected as a game, settings are set... , Because the continuity of game sound effects is important, the residual sound effects from the front have a significant impact; This is the maximum value of the standard fluctuation range. This is the minimum value of the standard fluctuation range, which is used to limit the initial forecast volume. The limit for excessive deviation is based on the historical preceding influencing factors. and environmental regulators The overall ratio value is set, for example, based on the current device's historical data. The values ​​are sorted from smallest to largest, and the value at the 30th percentile is taken as the minimum value of the standard fluctuation range. The value at the 70% mark is taken as the maximum value of the standard fluctuation range. .

[0013] In conjunction with the first aspect mentioned above, one possible implementation involves calculating a personalized bias rate based on usage habit data, including: Extract the final predicted volume of each device. The corresponding number of manual adjustments to the equipment and average adjustment amplitude And the time interval between the user entering the software and the first manual adjustment. ;in, Is the final predicted volume The corresponding number, and The range of values ​​is , For the final predicted volume The maximum value corresponding to the number; When the number of adjustments Greater than the number of times threshold The number of adjustments is determined by formula (3). Corresponding corrected confidence coefficient ; When the number of adjustments Not greater than the number of times threshold At that time, the number of adjustments will be... Corresponding corrected confidence coefficient The value is 1; Based on modified confidence coefficient Average adjustment range and time interval The personalized bias rate is determined by formula (4). The threshold number of adjustments is determined by the number of times the volume was adjusted by several experimenters who were dissatisfied with the volume during the experiment, and can be the average value of the number of adjustments. The calculation formula (3) is: ; In the formula, It involves adjusting the number of times in history using the current device. The average value determines the amplitude adjustment coefficient, and The value range is [0, 0.3]. This indicates the number of times the device has been adjusted in the history. If the quantity is less than the preset analysis threshold, the amplitude adjustment coefficient will be adjusted. The value is a fixed value of 0.15; The calculation formula (4) is: ; In the formula, The standard time interval is the time interval between several devices. The average value.

[0014] In conjunction with the first aspect above, in one possible implementation, the preliminary predicted volume is compensated to obtain the final predicted volume, including: Extracting personalized bias rates and preliminary volume prediction The personalized bias rate and preliminary volume prediction Perform a product operation to obtain the final predicted volume. Among them, the final predicted volume Satisfy the calculation formula: .

[0015] Secondly, it provides an AI-based intelligent audio control system, including a communication unit and a processing unit; The communication unit is used to acquire current device usage habit data and basic environmental data; wherein, the basic environmental data includes time points and ambient noise values; the usage habit data includes the number of manual adjustments made by the device after the final predicted volume adjustment and the average adjustment range; The processing unit is used to identify the software that the device is currently clicking to enter, obtain the historical volume of the software, determine the base volume of the current software based on the historical volume, and determine whether there is other audio playing within n seconds before the device enters the software. If yes, it determines the pre-influence factor based on the volume of the playing audio; otherwise, it sets the pre-influence factor to 0. Based on basic environmental data, environmental adjustment factors are determined. Based on the preceding influencing factors and environmental adjustment factors, the basic volume is adjusted to obtain the preliminary predicted volume. Based on usage habit data, the personalized bias rate is calculated and the preliminary predicted volume is compensated to obtain the final predicted volume. Set the device volume to the final predicted volume before entering the software.

[0016] Thirdly, this application provides a storage medium storing instructions that, when executed on an AI-based audio intelligent control system, cause the AI-based audio intelligent control system to perform the methods described in the first aspect and any possible implementation thereof.

[0017] This application provides an audio intelligent control method, system, and storage medium based on artificial intelligence, with the following beneficial effects: 1. This invention fundamentally eliminates the auditory delay and abruptness of traditional passive adjustment modes by shifting the volume decision-making time from after audio playback to before software startup and integrating pre-influence factors and environmental adjustment factors. In particular, the introduction of pre-influence factors utilizes a time-attenuation weighted mechanism to accurately capture the auditory inertia during cross-application switching, making the volume transition between old and new scenes as smooth and continuous as a natural sound field, avoiding the startling feeling and experience disruption caused by drastic volume jumps. Secondly, through multi-dimensional environmental perception and dynamic weight allocation mechanisms, it achieves refined control that balances scene characteristics and safety boundaries. The environmental adjustment factor, combined with time contextual cognition and non-linear buffering algorithms, enables the device to have a higher tolerance threshold for sudden noise at night or during habitually quiet periods, while maintaining agile response in noisy environments. Simultaneously, dynamic weighting based on software type and historical data limiting mechanisms ensure that the predicted volume meets the sound effect requirements of specific applications while always remaining within the safe range of the user's auditory comfort zone. Finally, through a closed-loop feedback learning mechanism to calculate personalized bias rates, it achieves a leap from "group standard" to "individually exclusive" experience. The system continuously quantifies the intensity of dissatisfaction and time urgency behind users' manual correction behavior, and dynamically tracks the evolution of user preferences in a time decay manner. This allows the predicted volume to become more and more accurate in matching the user's current real auditory needs as the usage period increases, truly achieving the goal of intelligent control that understands you better the more you use it and evolves silently.

[0018] 2. Traditional volume adjustment mechanisms always follow a passive cycle of "problem occurs—user perception—manual correction." That is, only when the sound is played at an inappropriate intensity, startling the user due to excessive loudness or causing them to miss crucial information due to excessive softness, or even after experiencing harsh shocks during frequent high-volume transitions, do users have to interrupt their current task to find physical buttons or on-screen sliders for adjustment. This delayed processing not only interrupts the immersive experience but also causes irreparable auditory damage and psychological frustration in the seconds before the adjustment is complete. The core innovation of this invention lies in subverting this logic. By constructing a volume prediction system with forward-looking perception and individual self-evolution capabilities, the device pre-sets the optimal volume even before the sound is emitted, allowing users to enjoy consistently comfortable listening without even noticing. This elevates adjustment from "passive remediation" to "active prediction," truly achieving imperceptible, extremely smooth, and seamless control.

[0019] It should be understood that the descriptions of technical features, technical solutions, beneficial effects, or similar language in this application do not imply that all features and advantages can be achieved in any single embodiment. Rather, it is understood that the description of a feature or beneficial effect means that a specific technical feature, technical solution, or beneficial effect is included in at least one embodiment. Therefore, the descriptions of technical features, technical solutions, or beneficial effects in this specification do not necessarily refer to the same embodiment. Furthermore, the technical features, technical solutions, and beneficial effects described in this embodiment can be combined in any suitable manner. Those skilled in the art will understand that embodiments can be implemented without one or more specific technical features, technical solutions, or beneficial effects of a particular embodiment. In other embodiments, additional technical features and beneficial effects may be identified in specific embodiments that do not embody all embodiments. Attached Figure Description

[0020] Figure 1 A schematic diagram illustrating the steps of the AI-based intelligent audio control method provided in this application embodiment; Figure 2 This is a schematic diagram of the structure of an AI-based intelligent audio control system provided in an embodiment of this application. Detailed Implementation

[0021] In the description of this application, unless otherwise stated, " / " means "or," for example, A / B can mean A or B. The "and / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone. Furthermore, "at least one" means one or more, and "multiple" means two or more. The terms "first," "second," etc., do not limit the quantity or order of execution, and "first," "second," etc., do not necessarily imply differences.

[0022] It should be noted that, in this application, the terms "exemplary" or "for example" are used to indicate that something is being described as an example, illustration, or illustration. Any embodiment or design described as "exemplary" or "for example" in this application should not be construed as being more preferred or advantageous than other embodiments or design solutions. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0023] like Figure 1 As shown, the first aspect of this application provides an artificial intelligence-based intelligent audio control method, including: S1: Based on user authorization, obtain current device usage habit data and basic environmental data; among which, basic environmental data includes time point and ambient noise value; usage habit data includes the number of manual adjustments and average adjustment range of the device after the final predicted volume adjustment; S2: Identify the software the device is currently clicking to enter, obtain the historical volume of the software, and determine the base volume of the current software based on the historical volume. Determine whether there is other audio playing within n seconds before the device enters the software. If yes, determine the preceding influence factor based on the volume of the playing audio; otherwise, set the preceding influence factor to 0. Specifically, identifying the software the device is currently clicking to enter in this step means capturing the moment the user actively triggers the application launch. This is different from the state where the application is running silently in the background or is automatically woken up by the system, ensuring accurate synchronization between the adjustment behavior and the user's display intention. The base volume is the initial reference value determined by the volume performance of the software under similar historical periods. It reflects the sound effect attributes of the specific application and the user's usual choices in similar scenarios. The preceding influence factor is the key mechanism of this invention to solve the abruptness of switching between applications. When it is detected that there is other audio playing within n seconds before entering the target software, the system quantifies the volume state of the preceding audio as an influence factor, so that the predicted volume of the new software can inherit the auditory inertia of the previous moment and achieve a smooth transition. If there is no preceding audio, the factor is set to zero to avoid introducing irrelevant noise. This design makes volume prediction no longer a static calculation isolated to a single application, but rather has dynamic coherence in the time dimension.

[0024] S3: Based on basic environmental data, environmental adjustment factors are determined. The base volume is adjusted using both the preceding influence factors and the environmental adjustment factors to obtain a preliminary predicted volume. A personalized bias rate is calculated based on usage habit data, and the preliminary predicted volume is compensated to obtain the final predicted volume. Specifically, this step completes the decision fusion process from multi-dimensional input to a single output value. The environmental adjustment factor converts the time point and environmental noise value into a correction coefficient for the base volume, allowing the prediction result to adaptively fluctuate with changes in the external sound field. The preliminary predicted volume is an intermediate result superimposed on the base volume, incorporating the continuity inertia from the preceding influence factors and the scene adaptation from the environmental adjustment factors; it balances auditory continuity and environmental adaptability. The personalized bias rate is a unique compensation amount extracted based on user usage habit data, used to correct the residual error between the general group model and individual true preferences. The final predicted volume is the result of the preliminary predicted volume after compensation by the personalized bias rate; it integrates information from three dimensions: environmental perception, scene continuity, and individual learning. It should be understood that the specific calculation methods and fusion logic of each factor in this step will be further detailed in subsequent embodiments. In this embodiment, it is only necessary to clarify its logical function as a multi-source information aggregation node, that is, to converge complex context information into an executable volume control command through hierarchical processing.

[0025] S4: Set the device volume to the final predicted volume before entering the software.

[0026] It's worth noting that traditional volume adjustment mechanisms always follow a passive cycle of "problem occurs—user perception—manual correction." This means that only when the sound is played at an inappropriate intensity, startling the user due to excessive loudness or causing them to miss crucial information due to excessive softness, or when frequently experiencing harsh jarring effects during high-volume transitions, do users have to interrupt their current task to find physical buttons or on-screen sliders for adjustment. This delayed processing not only disrupts the immersive experience but also causes irreparable auditory damage and psychological frustration in the seconds before the adjustment is complete. The core innovation of this invention lies in completely overturning this logic. By constructing a volume prediction system with forward-looking perception and individual self-evolution capabilities, the device pre-sets the optimal volume even before the sound is emitted, allowing users to enjoy consistently comfortable listening without even noticing. This elevates adjustment from "passive remediation" to "active prediction," truly achieving seamless, imperceptible control with no noticeable delay.

[0027] This groundbreaking pre-regulation capability is rooted in a deep understanding of the user's operational context. The system doesn't just monitor volume in the background; it identifies the software the user is clicking on and extracts its historical volume settings. This ensures that music players, video conferencing, or reading applications have a base volume that matches their content attributes milliseconds before launch. Crucially, it calculates a pre-influence factor, keenly detecting whether other audio is playing within n seconds before entering the target software. If traces of preceding audio are detected, this connection is quantified as an influence factor. This allows for a smooth, emotional transition in volume between new and old software, rather than abrupt changes. When a user switches from live streaming to music, or from music to a sudden incoming call, the pre-influence factor ensures the new software's volume maintains the continuity of the auditory scene, eliminating the "delay" and "startle" caused by drastic volume fluctuations during the switch. For independent launches without preceding audio, this factor is set to zero, ensuring a clean launch without any historical baggage, while maintaining accurate predictions.

[0028] The integration of environment and time further solidifies the reliability of advance prediction. The time points in the basic environmental data endow the system with human-like awareness of daily routines, enabling it to proactively reduce volume at night, even if the ambient noise level is no different from daytime, preventing disturbance to rest. Real-time ambient noise levels ensure that in different sound fields such as streets, subways, or libraries, the volume is subtly adjusted to a clearly identifiable yet non-intrusive level without any user intervention. Subsequently, the personalized bias rate calculated based on the number of manual adjustments and the average adjustment range is the core incremental factor that makes this advance adjustment increasingly refined with use. Traditional devices, even with automatic adjustment, often push a "standard" value to all users using a fixed algorithm, failing to address the diverse auditory preferences of individuals. This invention, however, continuously learns from every manual adjustment made by the user after receiving the predicted volume, precisely quantifying their dissatisfaction tendency and converting this tendency into a compensation value injected into the final predicted volume. This means that the volume setting before each playback is not an isolated calculation, but a customized prediction that incorporates the user's long-term behavioral profile. The system can adjust the slight deviations in the user's preferences before the user is even aware of it, achieving a reversal where the device proactively embraces the user's habits rather than the user adapting to the device.

[0029] This mechanism, which executes the final volume prediction before the software even starts, eliminates the painful chain of "waiting—playing—discomfort—adjustment—re-adaptation" at its root. The user taps the icon, the screen appears, and the sound flows out—all without lifting a finger to press a button. The ideal loudness, perfectly suited to the current environment, the specific scene, and the user's long-standing habits, immediately fills the ears, as if the device understands the user's intentions. There's not a millisecond of delay, nor a single unnecessary touch, making the entire multimedia interaction natural and smooth. Ultimately, this invention achieves far more than just automatic volume adjustment; it provides a completely immersive audio experience that eliminates interruptions. It moves the decision-making time of artificial intelligence from after the sound is emitted to before playback, ensuring that every listening experience begins with precise prediction and ends with a smooth enjoyment. It imbues all audio-enabled devices—smartphones, tablets, in-car entertainment systems, and even smart homes—with a proactive and gentle touch, truly allowing technology to take a backseat and enabling a seamless experience.

[0030] It should be noted that the software that the device is currently clicking to enter is the software that the device marks as the software that the user is currently clicking to enter when the user opens the software.

[0031] In one possible implementation of this application embodiment, the above-mentioned S1 can be implemented by the following S101, which will be described in detail below: S101: Obtain current device usage habit data and basic environmental data, including: The system calls the audio acquisition interface of the terminal device to acquire the ambient audio stream in the current environment; the audio acquisition interface includes a microphone. The ambient audio stream is preprocessed to calculate the root mean square (RMS) energy value within a preset time window, and the energy value is converted into a decibel value to obtain the ambient noise value. The preprocessing includes DC component removal and filtering. The preset time window is usually 1 second, but can also be set to other values. Call the system clock interface of the terminal device to obtain the current time. Access the locally stored historical interaction database, which stores audio playback logs of the device within a preset period in the past; wherein the preset period is generally the last 60 days. In the historical interaction database, target log records containing manual adjustment tags are filtered out; wherein, the manual adjustment tags are used to distinguish between volume changes actively triggered by the device and volume changes automatically triggered by the system. Key feature data is extracted from the target log records. This key feature data includes at least: the number of adjustments and the adjustment magnitude; the final volume level: the stable volume value maintained during audio playback after the device stops manual adjustment; the number of adjustments: the cumulative number of times the device triggers the volume adjustment button or gesture within the first few seconds of a single audio playback; and the average adjustment magnitude: the average value calculated by subtracting the final predicted volume from the manually adjusted volume value within a single audio playback, dividing by the final predicted volume, and then averaging the result. This average value characterizes the device's dissatisfaction with the current volume. The first few seconds are manually set, typically 3 seconds.

[0032] In one possible implementation of this application embodiment, the above-mentioned S2 can be implemented by the following S201 and S202, which are described in detail below: S201: Determine the base volume of the current software based on the historical volume levels, including: The historical volume levels within the reference range of the current time are extracted and input into the basic volume recognition model to obtain the basic volume of the current software. The reference range of the current time is a manually set time range centered on the current time. For example, the time range within one hour before and after the current time is marked as the reference range of the current time. This time window setting fully considers the time-dependent nature of users' listening habits. For example, users' volume expectations for the same software during lunch breaks and evening entertainment periods often differ significantly. By limiting the reference range, the model can accurately capture the contextual features of such dynamic changes. The basic volume recognition model is as follows: Extract historical volume values ​​within several reference ranges from the historical volume database, as well as the base volume corresponding to each reference range; wherein, the historical volume database includes historical volume values ​​within several reference ranges, and the base volume set by experts based on the historical volume values; The historical volume levels and baseline volume levels within the reference range are integrated into several sets of training and testing data. The training data is used to train the artificial intelligence model, and the testing data is used to test the trained artificial intelligence model. The artificial intelligence model is then adjusted based on the testing results. The final result is a baseline volume recognition model with the historical volume levels within the reference range as input and the baseline volume as output. The artificial intelligence model is a nonlinear regression model based on a neural network, implemented using a BP neural network structure and / or an RBF neural network structure. BP networks are good at handling complex nonlinear mapping relationships, while RBF networks have better local approximation capabilities and faster convergence speeds. Both can effectively fit the implicit correlation between user volume preferences and multidimensional factors such as environment and time.

[0033] Specifically, the steps for testing the trained AI model using validation data and adjusting the AI ​​model based on the validation results are as follows: The historical volume values ​​within the reference range from the test data are input into the trained AI model to obtain the corresponding baseline volume. This baseline volume is then compared with the corresponding baseline volume in the test data. If the difference is within a threshold (obtained empirically), no parameter adjustment is needed, and the next set of test data is tested. If the difference is outside the threshold, the corresponding parameters are adjusted until the difference in baseline volume between the test data and the baseline volume is within the threshold. The next set of test data is then tested. When the number of test data points where the difference in baseline volume is within the threshold accounts for 95% or more of the total test data, the historical volume values ​​within the reference range are obtained, and the baseline volume recognition model is output. It should be understood that although this embodiment uses a BP or RBF network as an example, other machine learning algorithms such as Support Vector Regression (SVR) and Random Forest Regression can also be used in other implementations, as long as they can meet the above requirements for nonlinear fitting based on historical data within the reference range and pass the predetermined convergence criterion test.

[0034] S202: Determine the preceding influence factor based on the volume of the played audio, including: Extract the volume of the device within the first n seconds before entering the software, and label the volume values ​​from front to back according to time. And the rate of change of volume between adjacent volumes is determined based on the calculation formula (1). ;in, Volume The number, and The value of is in the range of [1, n], where n is manually set. Here, n seconds constitutes a sliding observation window, the length of which determines the system's memory span of historical auditory context. In practice, n is generally taken as 3 to 8 seconds. This range is sufficient to cover the audio fade-in / fade-out process during most application startups while avoiding the introduction of irrelevant scene noise from excessively long historical data. It should be understood that although this embodiment uses a fixed duration n as an example, in other implementations, n can also be dynamically adjusted according to the type of preceding audio. For example, when a long video is detected, the observation window can be automatically extended, while when a short notification sound is detected, the window can be shortened, as long as it can represent the audio state before entering the software.

[0035] The rate of change of volume within the range of change constraints The preceding influence factor is obtained by calculating the average value. The variation constraint range is used to eliminate situations where the sound suddenly increases or decreases excessively. This range is set based on the human ear's adaptation to different sound changes observed in experiments. This step constitutes a key anti-avoidance design at the algorithm level. In real-world scenarios, users may experience sudden system notification sounds, game explosion sound effects, or instantaneous muting before switching software. These extreme events can lead to… Abnormal peaks or troughs appear. If these extreme values ​​are included in the averaging calculation without discrimination, the preceding impact factor... This will severely skew the predicted volume of the new software, causing it to be either excessively high and piercing or too low and indistinguishable, completely defeating the original purpose of a smooth transition. To more clearly illustrate the necessity of the variation constraint range, let's take a counter-scenario as an example: Suppose a user suddenly receives a high-decibel notification sound while watching a video and then immediately switches to a music app. Without setting a variation constraint range, the huge positive rate of change caused by the notification sound will dominate the average calculation, causing the preceding influence factor to... An abnormally high volume caused the music software's initial predicted volume to be incorrectly and significantly increased, startling the user. In this embodiment, because the rate of change caused by the prompt sound exceeded the constraint range set based on human ear adaptation experiments, the system automatically discarded it, retaining only the steady rate of change during normal video playback for calculation. This resulted in... It accurately reflects the background auditory tone before the user switches, ensuring that the music software can start with a smooth and consistent volume. It should be understood that the specific values ​​of the variation constraint range are not fixed. They can be fixed empirical thresholds or adaptive ranges that are dynamically learned based on the user's historical auditory tolerance, as long as their function is to filter out transient abnormal changes that exceed the human ear's comfortable adaptation ability.

[0036] The calculation formula (1) is: ; In the formula, when At that time, the rate of change of volume The value is 0; This is the decay rate constant, used to control the rate of decay. The value is obtained by conducting A / B testing on a number of test users and statistically analyzing their performance in different... The empirically optimal value obtained after manually adjusting the frequency at the given value, and The value range is [0,1]; Number The corresponding time point, The time point closest to the current time within n seconds before the device enters the software.

[0037] It is worth noting that the pre-influence factor obtained in this step solves the problem of auditory abruptness and experience break caused by drastic volume fluctuations when switching between different software or connecting audio playback, achieving an extremely smooth and natural cross-scene volume transition. Its core intelligence lies in the precise frame-by-frame tracing and dynamic analysis of the volume of the audio played in the previous n seconds. By constructing the rate of change between adjacent volumes and creatively introducing a weighted mechanism that decays over time, the calculation is no longer a simple averaging of historical data, but an intelligent abstraction that highly matches the characteristics of human auditory short-term memory. Specifically, the exponential decay term introduced in the calculation formula uses the distance between the current time point and each historical time point to adjust the weight. The farther away the volume change is from the current time, the lower its contribution to the pre-influence factor. Conversely, the volume change closer to the moment the user clicks to enter the software is given a higher influence weight. This time-proximity-based bias design accurately simulates the human brain's strongest physiological instinct to rely on the latest perceptual state in a continuous auditory scene. This allows the system to predict a seamless predictive value that inherits the auditory context of the previous second before the new software sound is emitted, without any abrupt drops or sudden increases. Meanwhile, the calculation formula includes a variation constraint range to intelligently remove abnormally large or small changes caused by sudden noises, brief pops, or momentary silences. If these abrupt changes are not removed, they will severely skew the final average prediction value, resulting in a prediction that not only fails to transition smoothly but also retains harshness. The variation constraint range is derived from experiments on the actual physiological adaptation limits of the human ear to different sound changes, ensuring that the filtering action is not blind, eliminating abnormal abrupt changes that exceed the comfortable perception boundaries of the human ear, while preserving the natural fluctuations that the human ear is accustomed to. This ensures that the data retained for calculation is within the perceptual channel that allows for auditory smoothness. The final pre-influence factor is calculated using the average value, which condenses and smooths the effective variation trend after screening. The result is not an isolated extreme value but a comprehensive index representing the overall volume evolution tendency. This smoothing feature effectively suppresses the jitter of a single calculation, providing an extremely stable and continuous sound foundation for subsequent volume prediction. In summary, this method enables the system to extract key factors carrying the auditory inertia of the moment when a device suddenly clicks on a voice call while playing a high-volume video, or switches from a voice call to start music. It allows environmental adjustment and personalized bias to work on this basis, so that the final predicted volume is seamlessly connected to the loudness that the human ear was just experiencing or enjoying. The user hears what sounds like an uninterrupted, unified sound field, fundamentally avoiding the experience of being forced to readjust to the human ear due to a sudden change in volume, which leads to a sense of delay and discomfort. It truly forges a smooth and silky intelligent audio control capability that is uninterrupted and requires no manual correction, greatly improving the comfort and premium feel of multimedia devices in complex multi-application switching scenarios.

[0038] It should be noted that when hour, The value is 0.

[0039] In one possible implementation of this application embodiment, the above-mentioned S3 can be implemented by the following S301, S302, S303 and S304, which are described in detail below: S301: Determine environmental regulation factors based on basic environmental data, including: A1: Extract time points from the basic environmental data within the last m seconds and determine whether the current time falls within a quiet period; if yes, adjust the time influence coefficient corresponding to the time points within the last m seconds. Set to 1; otherwise, set the time influence coefficient corresponding to the most recent m seconds to 1. Set to 0; where m is manually set, typically 5 seconds; the silent time is determined based on the device's historical ambient noise levels. When the ambient noise level frequently falls below a preset threshold for a certain period, this period is designated as the quiet period. This adaptive recognition mechanism based on historical data allows the system to accurately capture users' unique lifestyle patterns, such as their habitual quiet periods when entering a library or meeting room during the day, even if it's not traditionally nighttime. Triggering corresponding adjustment strategies, thereby distinguishing it from the objective nighttime period, and achieving accurate perception of the subjective intention to remain silent; A2: Determine if the current time falls within the late night period; if yes, apply the time influence coefficient 2 corresponding to the time point within the most recent m seconds. Set to 1; otherwise, set the time influence coefficient corresponding to the most recent m seconds to 2. The setting is 0; the late-night time period is manually set, typically from 0:00 to 6:00. This step incorporates social consensus-based silence guarantees as an auxiliary anchor. Even if the device lacks sufficient historical data to learn the user's personalized quiet time periods, or if the user is in an unusual environment, the system can still provide a basic volume suppression baseline based on this fixed late-night time period, ensuring the robustness of the technical solution under cold start or abnormal scenarios. It should be understood that although this embodiment uses 0:00-6:00 as an example, in other implementations, the late-night time period can also be adjusted according to the user's time zone or a custom schedule, as long as it can provide basic time context constraints.

[0040] A3: Extract the ambient noise value from the basic environmental data within the last m seconds and obtain the average value of the ambient noise value. Based on the average value Time Influence Coefficient 1 And time influence coefficient 2 Determine environmental regulating factors Among them, environmental regulatory factors Satisfy the calculation formula: , The standard value for environmental noise can be the average value of environmental noise values ​​in the historical data of the current equipment.

[0041] It is worth noting that the multi-dimensional decision-making system constructed in this step when determining the environmental adjustment factor elevates the adjustment of ambient volume from a cold, mechanical linear compensation to an intelligent judgment that recognizes the time context and empathizes with the user's life rhythm. This prevents the awkward situation of erroneous adjustments caused by slight fluctuations in ambient noise at inappropriate times, thus disrupting the tranquility. First, in step A1, the system no longer relies on rigidly set fixed quiet periods. Instead, it learns by backtracking the device's historical ambient noise values, accurately capturing the regular low-noise range formed by the device owner over long-term use. This personalized quiet period is used as the threshold for trigger time influence coefficient one. This allows the environmental adjustment factor to distinguish between subjective nighttime and objective silence. Even if the user enters a library or meeting room during the day, the device is subject to a high degree of constraint because it is familiar with the habitual quietness of this place at this time. The late-night period introduced in step A2 serves as an auxiliary anchor point, providing a social consensus-level guarantee of silence through the manually calibrated early morning period, ensuring a safe volume suppression baseline even when historical data is insufficient. The key lies in the design of the environmental adjustment factor calculation formula in A3. It uses the deviation between the average environmental noise value and the standard noise value as the basis for adjustment, but creatively nests a logarithmic function with the base of the natural constant e superimposed on the maximum time influence coefficient in the denominator. This structure produces a highly physically intuitive and physiologically adaptable nonlinear buffering effect: when both time influence coefficients are 0, meaning the current time is neither the user's usual quiet time nor late at night, the denominator simplifies to a logarithmic unit with base e. At this point, the environmental deviation information is fully transmitted to the predictive adjustment at a proportion of normal sensitivity. The system responds swiftly to real-world conditions, ensuring clear audibility even in noisy outdoor environments. However, during quiet periods or late at night, the maximum time-impact coefficient is set to 1. The denominator instantly transforms into an amplified form of the logarithmic value, forcibly increasing the base of the division. This significantly compresses the final value of the environmental adjustment factor under the same increase in ambient noise, making volume adjustment more conservative. Even sudden, brief noise spikes like barking dogs or car alarms from outside are deemed unworthy of disturbing the user's habitual quiet atmosphere due to the strong time constraint, and are not recklessly amplified by the system. This mechanism perfectly simulates the psychological resistance and tolerance threshold of humans to sudden sounds in late night or known quiet environments using the compression characteristics of logarithmic functions. It achieves a dual regulatory personality: highly sensitive to environmental changes during the day and highly insensitive to noise at night and in habitually quiet scenarios. This transforms device volume control from a black-and-white mechanical reaction into a lifelike experience reminiscent of circadian rhythms, deeply respecting the user's emotional space. Ultimately, it presents a seamless control experience that always resonates harmoniously with the environmental atmosphere, time perception, and individual habits.

[0042] S302: Based on the preceding influence factor and environmental adjustment factor, the baseline volume is adjusted to obtain the preliminary predicted volume, including: Extracting the preceding impact factor Environmental regulating factors and base volume And the preceding influence factors and environmental regulators Normalization is performed based on the aforementioned prior influence factors. Environmental regulating factors and base volume The preliminary predicted volume is determined by formula (2). ; The calculation formula (2) is: ; In the formula, and The proportional adjustment coefficient is determined based on the current software type identifier, for example, , When the software is detected as video software, set... , Because short videos are more sensitive to environmental noise and need to adapt quickly to the environment; when the app is detected as a game, settings are set... , Because the continuity of game sound effects is important, the lingering effect of preceding sound effects has a significant impact; it should be understood that the above values ​​are only illustrative examples, and in actual implementation, and It can also be configured differently based on other software types such as music, calls, and navigation, and can even be dynamically fine-tuned based on the user's real-time interaction status in specific software, such as whether voice chat is enabled in games, as long as it follows the core idea of ​​"scenario determines weight".

[0043] This is the maximum value of the standard fluctuation range. This is the minimum value of the standard fluctuation range, which is used to limit the initial forecast volume. The limit for excessive deviation is based on the historical preceding influencing factors. and environmental regulators The overall ratio value is set, for example, based on the current device's historical data. The values ​​are sorted from smallest to largest, and the value at the 30th percentile is taken as the minimum value of the standard fluctuation range. The value at the 70% mark is taken as the maximum value of the standard fluctuation range. This generation method, based on the percentile of a user's historical data, achieves a personalized adaptive limit boundary. For users accustomed to large volume fluctuations, their 30%-70% percentile range is naturally wider, giving the system greater adjustment freedom; while for users who prefer a stable listening experience, this range automatically narrows, providing stricter smoothing constraints. Compared to a globally uniform hard-coded threshold, this data-driven soft limiting mechanism not only effectively eliminates the interference of extreme outliers in historical data on the current calculation, but also ensures that the limiting boundary itself conforms to the user's long-term auditory comfort zone, avoiding the disjointed experience caused by a "one-size-fits-all" cutoff. Through the synergistic effect of the above dual mechanisms, the preliminary predicted volume output in this embodiment... It not only accurately adapts to the current software's scenario characteristics, but is also firmly constrained within the user's personalized safety and comfort zone, laying an extremely robust and scenario-appropriate foundation for subsequent personalized bias compensation.

[0044] It is worth noting that in this step, when integrating the pre-influence factors and environmental adjustment factors to generate the preliminary predicted volume, the software type-aware dynamic weight allocation and the intelligent amplitude limiting mechanism based on historical data enable fine-grained control of the basic volume that takes into account both scene characteristics and safety boundaries. This completely eliminates the experience mismatch caused by treating different application types the same under the traditional fixed parameter adjustment mode. Meanwhile, the calculation formula also includes a limiting layer consisting of the minimum and maximum values ​​of the standard fluctuation range based on the percentile of historical data. Through the double nesting of the min and max functions, the fluctuation of the initially predicted volume is strictly constrained within the reasonable historical range. This mechanism acts like a buffer valve with self-learning capabilities, allowing the volume to fluctuate flexibly according to the preceding influences and the environment, while absolutely preventing the volume from being distortedly amplified or excessively suppressed to an unacceptable state due to an extreme outlier in a certain factor in the calculation process. Moreover, the threshold of the standard fluctuation range is not determined arbitrarily, but is taken from the cumulative proportional value distribution in the device's own historical operation, realizing the adaptive limitation boundary that is personalized to each individual, so that the limitation itself is in line with personal usage habits, rather than a one-size-fits-all drastic cutoff. Ultimately, this step allows the base volume to be determined by the combined effects of auditory inertia and real-time environmental response, through intelligent weighting of software features and flexible convergence of historical safety boundaries. This results in a preliminary predicted volume that maintains a smooth and consistent feel during game transitions, adapts quickly to the environment during video playback, and never deviates from the user's auditory comfort zone under any circumstances. This provides an extremely stable and highly scene-appropriate foundation for subsequent personalized bias rate fine-tuning, laying the key foundation for the entire system's accurate predictions and smooth, comfortable listening experience.

[0045] S303: Calculate personalized bias rates based on usage habit data, including: Extract the final predicted volume of each device. The corresponding number of manual adjustments to the equipment and average adjustment amplitude And the time interval between the user entering the software and the first manual adjustment. ;in, Is the final predicted volume The corresponding number, and The range of values ​​is , For the final predicted volume The maximum value corresponding to the number; When the number of adjustments Greater than the number of times threshold The number of adjustments is determined by formula (3). Corresponding corrected confidence coefficient ; When the number of adjustments Not greater than the number of times threshold At that time, the number of adjustments will be... Corresponding corrected confidence coefficient The value is set to 1; in practice, when the number of adjustments does not exceed the threshold... When the number of adjustments significantly exceeds a threshold, the system treats it as a normal expression of preference or an occasional fine-tuning, assigning a modified confidence coefficient of 1 and preserving its original influence. At this time, this often no longer represents a rational auditory preference, but is more likely an emotional continuous key press caused by extreme frustration, accidental touches by children, or a jitter signal caused by a physical malfunction of the volume keys. If these high-frequency operations are regarded as strong preferences without discrimination, the model will overfit to noisy data. The calculation formula (3) utilizes the growth characteristics of the logarithmic function, so that when Far exceeding hour, The growth rate slows down dramatically, and even the offset design within absolute values ​​has a suppressive effect, effectively reducing the weight of such extreme outliers in the overall learning. This "too much of a good thing" confidence evaluation mechanism ensures that the learning of personalized bias rates is always based on rational and stable user feedback, enhancing the system's robustness in complex real-world interactions.

[0046] Based on modified confidence coefficient Average adjustment range and time interval The personalized bias rate is determined by formula (4). The threshold number of adjustments is determined by the number of times the volume was adjusted by several experimenters who were dissatisfied with the volume during the experiment, and can be the average value of the number of adjustments. The calculation formula (3) is: ; In the formula, It involves adjusting the number of times in history using the current device. The average value determines the amplitude adjustment coefficient, and The value range is [0, 0.3]. This indicates the number of times the device has been adjusted in the history. If the quantity is less than the preset analysis threshold, the amplitude adjustment coefficient will be adjusted. The value is a fixed value of 0.15; The calculation formula (4) is: ; In the formula, The standard time interval is the time interval between several devices. The average value.

[0047] It is worth noting that the closed-loop feedback learning mechanism constructed in this step when calculating the personalized bias rate breaks through the traditional volume adjustment system's disregard for individual user preferences. By finely quantifying the intensity of dissatisfaction, the direction of correction, and the degree of hesitation behind each manual correction behavior of the user, and dynamically distilling the compensation increment that is highly consistent with the individual's auditory characteristics in a time decay weighted manner, it truly achieves a qualitative leap from "system forcibly pushing fixed values" to "device actively adapting to people". The solution first extracts user feedback data from three dimensions for each final predicted volume: the number of adjustments, the average adjustment magnitude, and the time interval between the user entering the software and the first manual adjustment. These three factors together paint a complete picture of user dissatisfaction: the number of adjustments reflects the frequency of user resistance to a certain type of volume prediction result, whether it is habitual dissatisfaction or occasional fluctuations; the average adjustment magnitude directly indicates the direction and magnitude of the deviation between the user's expectation and the system's prediction, whether it is insufficient positive compensation or excessive negative compensation; the time interval implies the length of time the user hesitates and tolerates, with extremely short intervals indicating extremely poor predictions and the user's eagerness to make corrections, while longer intervals may only indicate minor adjustments or passive adjustments due to environmental changes. At the data processing level, the solution creatively sets a threshold for the number of adjustments to distinguish between abnormal and random behavior. When the number of adjustments does not exceed the threshold, a correction confidence coefficient of 1 is assigned, which is equivalent to treating these occasional behaviors as statistically insignificant noise and retaining them without additional reinforcement, avoiding contamination of the overall learning results due to one or two accidental touches or temporary environmental changes. The final personalized bias rate calculation employs a weighted average architecture with time intervals as the exponential decay weight. Each manual adjustment is assigned a distinct impact based on its timing: older adjustments have lower weights or are even naturally forgotten, while more recent adjustments are highly focused. This design perfectly replicates the long-term process of human habit drift and preference evolution, allowing the system to automatically shed outdated biases and keenly pursue the user's current auditory needs. Simultaneously, using standard time intervals as the decay benchmark further eliminates inconsistencies in dimensions caused by differences in operating rhythms across different devices or users, ensuring the decay rate has cross-user comparability and generalization capabilities. In summary, this solution results in a final personalized bias rate that is a comprehensive compensation value that has undergone frequency confidence filtering, precise directional amplitude quantification, and intelligent weighting based on time. It allows the volume predicted by artificial intelligence to leap from a "group standard comfort value" to an "individually exclusive ideal value," and automatically tracks subtle shifts in user preferences over time, truly achieving an ultimate intelligent control experience that becomes increasingly intuitive and self-evolving with use.

[0048] It should be noted that when the number of adjustments is mentioned... If the number of attempts exceeds the threshold, it indicates that the volume adjustment may be unstable due to user emotions or unresponsive volume keys, and therefore the reliability of the volume adjustment needs to be reduced.

[0049] It should be noted that when the final predicted volume... When the number of manual adjustments to the corresponding device is 0, the number of manual adjustments and the average adjustment range corresponding to the final predicted volume, as well as the time interval from when the user enters the software to when manual adjustment is performed, are not extracted.

[0050] S304: Compensate the preliminary predicted volume to obtain the final predicted volume, including: Extracting personalized bias rates and preliminary volume prediction The personalized bias rate and preliminary volume prediction Perform a product operation to obtain the final predicted volume. Among them, the final predicted volume Satisfy the calculation formula: .

[0051] The above primarily describes the solutions of the embodiments of this application from the perspective of device implementation. It is understood that each device, such as an AI-based intelligent audio control system, includes at least one of the hardware structures and software modules corresponding to the execution of each function in order to achieve the above-mentioned functions. Those skilled in the art should readily recognize that, based on the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein, this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed by hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0052] This application embodiment can divide the AI-based intelligent audio control system into functional units according to the above method example. For example, each function can be divided into separate functional units, or two or more functions can be integrated into one processing unit. The integrated unit can be implemented in hardware or as a software functional unit. It should be noted that the unit division in this application embodiment is illustrative and only represents one logical functional division; other division methods may be used in actual implementation.

[0053] When using integrated units, Figure 2 The diagram shows a possible structure of the AI-based intelligent audio control system (referred to as communication system 30) involved in the above embodiments. The communication system 30 includes a processing unit 301 and a communication unit 302, and may also include a storage medium (referred to as storage unit 303). Figure 2The schematic diagram shown can be used to illustrate the structure of the AI-based intelligent audio control system involved in the above embodiments.

[0054] when Figure 2 The schematic diagram shown illustrates the structure of the AI-based intelligent audio control system involved in the above embodiments. The processing unit 301 is used to control and manage the actions of the AI-based intelligent audio control system, the communication unit 302 is used for the AI-based intelligent audio control system to communicate with other devices, and the storage unit 303 is used to store the program code and data of the AI-based intelligent audio control system.

[0055] For example, communication unit 302 is used to acquire current device usage habit data and basic environmental data; wherein, the basic environmental data includes time points and ambient noise values; the usage habit data includes the number of times the device manually adjusts and the average adjustment range after the final predicted volume adjustment; Processing unit 301: Used to identify the software that the device is currently clicking to enter, obtain the historical volume of the software, determine the base volume of the current software based on the historical volume; determine whether there is other audio playing within n seconds before the device enters the software, if yes, determine the pre-influence factor based on the volume of the playing audio, if no, set the pre-influence factor to 0; Based on basic environmental data, environmental adjustment factors are determined. Based on the preceding influencing factors and environmental adjustment factors, the basic volume is adjusted to obtain the preliminary predicted volume. Based on usage habit data, the personalized bias rate is calculated and the preliminary predicted volume is compensated to obtain the final predicted volume. Set the device volume to the final predicted volume before entering the software.

[0056] Although this application has been described herein in conjunction with various embodiments, those skilled in the art, by reviewing the accompanying drawings, disclosure, and appended claims, will understand and implement other variations of the disclosed embodiments in carrying out the claimed application. In the claims, the word "comprising" does not exclude other components or steps, and "a" or "an" does not exclude multiple instances. A single processor or other unit can implement several functions listed in the claims. While different dependent claims may recite certain measures, this does not mean that these measures cannot be combined to produce good results.

[0057] Although this application has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made thereto without departing from the spirit and scope of this application. Accordingly, this specification and drawings are merely illustrative descriptions of the application as defined by the appended claims, and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from the spirit and scope of this application. Thus, if such modifications and modifications fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and modifications.

[0058] Some of the data in the above calculation formula are obtained by removing dimensions and taking their numerical values. The calculation formula is a calculation formula that is closest to the real situation, obtained by software simulation of a large amount of collected data. The preset parameters and preset thresholds in the calculation formula are set by those skilled in the art according to the actual situation or obtained through simulation of a large amount of data.

Claims

1. An audio intelligent regulation method based on artificial intelligence, characterized in that, include: S1: Obtain current device usage habit data and basic environmental data; among which, basic environmental data includes time point and ambient noise value; usage habit data includes the number of manual adjustments and average adjustment range of the device after the final predicted volume adjustment; S2: Identify the software that the device is currently clicking to enter, obtain the historical volume of the software, and determine the base volume of the current software based on the historical volume; determine whether there is other audio playing within n seconds before the device enters the software; if yes, determine the pre-influence factor based on the volume of the playing audio; if no, set the pre-influence factor to 0. S3: Determine the environmental adjustment factor based on the basic environmental data, adjust the basic volume based on the preceding influencing factor and the environmental adjustment factor to obtain the preliminary predicted volume, calculate the personalized bias rate based on the usage habit data, and compensate the preliminary predicted volume to obtain the final predicted volume. S4: Set the device volume to the final predicted volume before entering the software.

2. The audio intelligent control method based on artificial intelligence according to claim 1, characterized in that, The acquisition of current device usage habit data and basic environmental data includes: The system calls the audio acquisition interface of the terminal device to acquire the ambient audio stream in the current environment; the audio acquisition interface includes a microphone. The ambient audio stream is preprocessed to calculate the root mean square energy value within a preset time window, and the energy value is converted into a decibel value to obtain the ambient noise value; wherein, the preprocessing includes DC component removal and filtering. Call the system clock interface of the terminal device to obtain the current time. Access the locally stored historical interaction database, which stores audio playback logs of the device within a preset period in the past; wherein the preset period is generally the last 60 days. In the historical interaction database, target log records containing manual adjustment tags are filtered out; wherein, the manual adjustment tags are used to distinguish between volume changes actively triggered by the device and volume changes automatically triggered by the system. Key feature data is extracted from the target log records. The key feature data includes at least: the number of adjustments and the adjustment range; the final volume level: the stable volume value maintained by the device during audio playback after the device stops manual adjustment; the number of adjustments: the cumulative number of times the device triggers the volume adjustment button or gesture within a few seconds before a single audio playback; and the average adjustment range: the average value obtained by subtracting the final predicted volume from the volume value manually adjusted by the device during a single audio playback, dividing by the final predicted volume, and then averaging the result.

3. The audio intelligent control method based on artificial intelligence according to claim 1, characterized in that, Determining the base volume of the current software based on the historical volume levels includes: Extract the historical volume within the reference range of the current time, and input the historical volume into the basic volume recognition model to obtain the basic volume of the current software; wherein, the reference range of the current time is a time range centered on the current time; The basic volume recognition model is as follows: Extract historical volume values ​​within several reference ranges from the historical volume database, as well as the base volume corresponding to each reference range; wherein, the historical volume database includes historical volume values ​​within several reference ranges, and the base volume set by experts based on the historical volume values; The historical volume levels and baseline volume levels within the reference range are integrated into several sets of training and testing data. The training data is used to train the artificial intelligence model, and the testing data is used to test the trained artificial intelligence model. The artificial intelligence model is then adjusted based on the testing results. Finally, a baseline volume recognition model is obtained, with the historical volume levels within the reference range as the input and the baseline volume as the output. The artificial intelligence model is a nonlinear regression model based on a neural network, implemented using a BP neural network structure and / or an RBF neural network structure.

4. The audio intelligent control method based on artificial intelligence according to claim 1, characterized in that, The determination of the preceding influence factor based on the volume of the played audio includes: Extract the volume of the device within the first n seconds before entering the software, and label the volume values ​​from front to back according to time. And the rate of change of volume between adjacent volumes is determined based on the calculation formula (1). ;in, Volume The number, and The range of values ​​for is [1, n]. The rate of change of volume within the range of change constraints The preceding influence factor is obtained by calculating the average value. The variation constraint range is used to eliminate situations where the sound suddenly increases or decreases excessively. The variation constraint range is set based on how the human ear adapts to different sound changes in the experiment. The calculation formula (1) is: ； In the formula, This is the decay rate constant, used to control the rate of decay. The value range is [0,1]; Number The corresponding time point, The time point closest to the current time within n seconds before the device enters the software.

5. The audio intelligent control method based on artificial intelligence according to claim 4, characterized in that, The determination of environmental regulation factors based on basic environmental data includes: A1: Extract time points from the basic environmental data within the last m seconds and determine whether the current time falls within a quiet period; if yes, adjust the time influence coefficient corresponding to the time points within the last m seconds. Set to 1; otherwise, set the time influence coefficient corresponding to the most recent m seconds to 1. Set to 0; the silent time is determined based on the device's historical ambient noise levels. A2: Determine if the current time falls within the late night period; if yes, apply the time influence coefficient 2 corresponding to the time point within the most recent m seconds. Set to 1; otherwise, set the time influence coefficient corresponding to the most recent m seconds to 2. Set to 0; A3: Extract the ambient noise value from the basic environmental data within the last m seconds and obtain the average value of the ambient noise value. Based on the average value Time Influence Coefficient 1 And time influence coefficient 2 Determine environmental regulating factors Among them, environmental regulatory factors Satisfy the calculation formula: , The standard value for environmental noise can be the average value of environmental noise values ​​in the historical data of the current equipment.

6. The audio intelligent control method based on artificial intelligence according to claim 5, characterized in that, The process of adjusting the base volume based on the preceding influence factor and the environmental adjustment factor to obtain the preliminary predicted volume includes: Extracting the preceding impact factor Environmental regulating factors and base volume And the preceding influence factors and environmental regulators Normalization is performed based on the aforementioned prior influence factors. Environmental regulating factors and base volume The preliminary predicted volume is determined by formula (2). ; The calculation formula (2) is: ； In the formula, and The proportional adjustment coefficient is determined based on the current software type identifier; This is the maximum value of the standard fluctuation range. This is the minimum value of the standard fluctuation range, which is used to limit the initial forecast volume. The limit for excessive deviation is based on the historical preceding influencing factors. and environmental regulators The overall ratio value is obtained by setting it.

7. The audio intelligent control method based on artificial intelligence according to claim 1, characterized in that, The calculation of personalized bias rate based on usage habit data includes: Extract the final predicted volume of each device. The corresponding number of manual adjustments to the equipment and average adjustment amplitude And the time interval between the user entering the software and the first manual adjustment. ;in, Is the final predicted volume The corresponding number, and The range of values ​​is , For the final predicted volume The maximum value corresponding to the number; When the number of adjustments Greater than the number of times threshold The number of adjustments is determined by formula (3). Corresponding corrected confidence coefficient ; When the number of adjustments Not greater than the number of times threshold At that time, the number of adjustments will be... Corresponding corrected confidence coefficient The value is 1; Based on modified confidence coefficient Average adjustment range and time interval The personalized bias rate is determined by formula (4). The threshold number of adjustments is determined by the number of times the volume was adjusted by several experimenters who were dissatisfied with the volume during the experiment, and can be the average value of the number of adjustments. The calculation formula (3) is: ； In the formula, It involves adjusting the number of times in history using the current device. The average value determines the amplitude adjustment coefficient, and The value range is [0, 0.3]; The calculation formula (4) is: ； In the formula, The standard time interval is the time interval between several devices. The average value.

8. The audio intelligent control method based on artificial intelligence according to claim 7, characterized in that, The process of compensating the initial predicted volume to obtain the final predicted volume includes: Extracting personalized bias rates and preliminary volume prediction The personalized bias rate and preliminary volume prediction Perform a product operation to obtain the final predicted volume. Among them, the final predicted volume Satisfy the calculation formula: .

9. An AI-based intelligent audio control system, used to run the AI-based intelligent audio control method according to any one of claims 1 to 8, characterized in that, include: Communication unit and processing unit; The communication unit is used to acquire current device usage habit data and basic environmental data; wherein, the basic environmental data includes time points and ambient noise values; the usage habit data includes the number of manual adjustments made by the device after the final predicted volume adjustment and the average adjustment range; The processing unit is used to identify the software that the device is currently clicking to enter, obtain the historical volume of the software, determine the base volume of the current software based on the historical volume, and determine whether there is other audio playing within n seconds before the device enters the software. If yes, it determines the pre-influence factor based on the volume of the playing audio; otherwise, it sets the pre-influence factor to 0. Based on basic environmental data, environmental adjustment factors are determined. Based on the preceding influencing factors and environmental adjustment factors, the basic volume is adjusted to obtain the preliminary predicted volume. Based on usage habit data, the personalized bias rate is calculated and the preliminary predicted volume is compensated to obtain the final predicted volume. Set the device volume to the final predicted volume before entering the software.

10. A storage medium storing instructions that, when executed on an AI-based audio intelligent control system, cause the AI-based audio intelligent control system to perform the AI-based audio intelligent control method according to any one of claims 1 to 8.

Images