Artificial intelligence-based digital guqin tone color optimization method and system

Abstract

The application relates to the field of artificial intelligence, and discloses a digital qin timbre optimization method and system based on artificial intelligence and a storage medium. The method comprises the following steps: performing weighted synthesis processing on analog audio signals and original audio signals to generate an initial synthesized audio signal; determining target timbre envelope parameters based on a physical modeling engine, preset timbre deviation parameters and preset fingering rules; and generating a target audio signal according to the target timbre envelope parameters and the initial synthesized audio signal. In the manner, the original audio signals of the digital qin are collected and subjected to layered sampling processing, the timbre characteristics of qin playing under different dynamics are captured, the weighted synthesis manner is used to reasonably distribute different signals according to the characteristics of the different signals, the target timbre envelope parameters define the dynamic change range and characteristics of the timbre, the fullness and level of the timbre are enhanced, and the authenticity of the timbre of the digital qin is improved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a method and system for optimizing the tone of a digital guqin based on artificial intelligence. Background Technology

[0002] The guqin, a gem of traditional Chinese music, possesses a profound historical heritage and unique artistic charm. In modern society, the digital guqin, as an innovative instrument combining the traditional guqin with modern electronic technology, aims to inject new vitality into the inheritance and development of guqin art, while also providing convenience for more people to understand and learn the guqin. However, in the development of the digital guqin, tone quality has always been a key and urgent issue to be addressed. The traditional guqin, due to its complex manufacturing process, wood characteristics, and playing techniques, produces a rich, delicate, and melodious tone. However, when simulating the guqin's tone, the digital guqin is often limited by electronic components, audio processing algorithms, and sampling technology, resulting in a significant difference in tone compared to the traditional guqin, making it difficult to meet the high demands of professional performers and guqin enthusiasts.

[0003] With the rapid development of artificial intelligence (AI) technology, its application in audio processing has gradually attracted attention. AI algorithms, especially neural network technology in deep learning, have demonstrated powerful capabilities in pattern recognition, feature extraction, and data fitting. In music audio processing, numerous successful cases have shown that AI can effectively analyze, generate, and optimize audio signals, such as in speech synthesis, music style conversion, and audio enhancement.

[0004] Currently, existing methods for optimizing the tone of digital guqin mainly focus on physical modeling and traditional digital signal processing. For example, by physically modeling the sound production principle of the guqin, attempts are made to reproduce its vibration patterns and acoustic characteristics in the digital realm, thereby generating a more realistic tone. However, this method often requires a precise mathematical description of the guqin's physical structure, and in actual calculations and simulations, due to high computational complexity and the existence of simplifying assumptions, it is difficult to completely and accurately reproduce the actual tone of the guqin. Therefore, how to optimize the tone of digital guqin through artificial intelligence technology and improve the realism of its tone has become an urgent technical problem to be solved. Summary of the Invention

[0005] This application provides a method, system, and storage medium for optimizing the tone of a digital guqin based on artificial intelligence, in order to improve the realism of the tone of the digital guqin.

[0006] Firstly, this application provides a method for optimizing the tone of a digital guqin based on artificial intelligence, the method comprising:

[0007] The original audio signal of the digital guqin was acquired, and the original audio signal was subjected to layered sampling processing to determine the audio samples of each dynamic range.

[0008] The analog audio signal of the digital guqin is generated by a preset string vibration algorithm, and the analog audio signal and the original audio signal are weighted and synthesized to generate an initial synthesized audio signal.

[0009] The target timbre envelope parameters of the digital guqin are determined based on a preset physical modeling engine, preset timbre deviation parameters, and preset fingering rules.

[0010] The target audio signal of the digital guqin is generated based on the target timbre envelope parameters and the initial synthesized audio signal.

[0011] Secondly, this application also provides a digital guqin tone optimization system based on artificial intelligence, the system comprising:

[0012] The dynamic audio sample determination module is used to acquire the original audio signal of the digital guqin and perform layered sampling processing on the original audio signal to determine each dynamic audio sample;

[0013] The initial synthesized audio signal generation module is used to generate an analog audio signal of the digital guqin through a preset string vibration algorithm, and to perform weighted synthesis processing on the analog audio signal and the original audio signal to generate an initial synthesized audio signal;

[0014] The target timbre envelope parameter determination module is used to determine the target timbre envelope parameters of the digital guqin based on a preset physical modeling engine, preset timbre deviation parameters, and preset fingering rules.

[0015] The target audio signal generation module is used to generate the target audio signal of the digital guqin based on the target timbre envelope parameters and the initial synthesized audio signal.

[0016] Thirdly, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to implement the aforementioned artificial intelligence-based digital guqin tone optimization method.

[0017] This application discloses a method, system, and storage medium for optimizing the tone of a digital guqin (a seven-stringed zither) based on artificial intelligence. The method includes: acquiring the original audio signal of the digital guqin; performing layered sampling processing on the original audio signal to determine audio samples at different dynamic levels; generating a simulated audio signal of the digital guqin using a preset string vibration algorithm; and performing weighted synthesis processing on the simulated audio signal and the original audio signal to generate an initial synthesized audio signal; determining the target tone envelope parameters of the digital guqin based on a preset physical modeling engine, preset tone deviation parameters, and preset fingering rules; and generating the target audio signal of the digital guqin based on the target tone envelope parameters and the initial synthesized audio signal. Through this method, this application accurately captures the tone characteristics of guqin performance at different dynamic levels by acquiring the original audio signal of the digital guqin and performing layered sampling processing. The algorithm uses a preset string vibration algorithm to generate analog audio signals. The weighted synthesis method is reasonably adjusted according to the characteristics of different signals, so that the synthesized audio signals are more natural and harmonious in terms of timbre blending and overall performance, enhancing the fullness and layering of the timbre. The target timbre envelope parameter defines the dynamic range and characteristics of the timbre, improving the realism of the digital guqin timbre. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a schematic flowchart of a digital guqin timbre optimization method based on artificial intelligence, provided in the first embodiment of this application;

[0020] Figure 2 This is a schematic flowchart of a digital guqin timbre optimization method based on artificial intelligence, provided in the second embodiment of this application;

[0021] Figure 3 This is a schematic flowchart of a digital guqin timbre optimization method based on artificial intelligence provided in the third embodiment of this application;

[0022] Figure 4 A schematic block diagram of an artificial intelligence-based digital guqin tone optimization system provided for embodiments of this application;

[0023] Figure 5 A schematic block diagram of the structure of a computer device provided for an embodiment of this application. Detailed Implementation

[0024] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0025] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content and operations / steps, nor does it necessarily have to be performed in the described order. For example, some operations / steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.

[0026] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of the application. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0027] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0028] This application provides an AI-based method, system, and storage medium for optimizing the tone of a digital guqin (a seven-stringed zither). The AI-based method can be applied to an AI-based digital guqin tone optimization system. By acquiring the original audio signal of the digital guqin and performing layered sampling processing, it accurately captures the tonal characteristics of the guqin performance at different dynamic levels. A preset string vibration algorithm is used to generate simulated audio signals, and a weighted synthesis method is rationally adjusted according to the characteristics of different signals. This results in a more natural and harmonious synthesized audio signal in terms of tone blending and overall performance, enhancing the fullness and layering of the tone. The target tone envelope parameter defines the dynamic range and characteristics of the tone, improving the realism of the digital guqin tone.

[0029] The following detailed description of some embodiments of this application is provided in conjunction with the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0030] Please see Figure 1 , Figure 1This is a schematic flowchart illustrating an AI-based digital guqin timbre optimization method provided in the first embodiment of this application. This AI-based digital guqin timbre optimization method can be applied to an AI-based digital guqin timbre optimization system. It is used to accurately capture the timbre characteristics of guqin performance at different dynamic levels by acquiring the original audio signal of the digital guqin and performing layered sampling processing. A preset string vibration algorithm is used to generate simulated audio signals, and a weighted synthesis method is rationally adjusted according to the characteristics of different signals. This makes the synthesized audio signal more natural and harmonious in terms of timbre blending and overall performance, enhancing the fullness and layering of the timbre. The target timbre envelope parameter defines the dynamic range and characteristics of the timbre, improving the realism of the digital guqin timbre.

[0031] like Figure 1 As shown, the AI-based digital guqin tone optimization method specifically includes steps S10 to S40.

[0032] Step S10: Acquire the original audio signal of the digital guqin, and perform layered sampling processing on the original audio signal to determine the audio samples of each dynamic range;

[0033] Specifically, high-precision audio acquisition equipment, such as professional audio interfaces and audio recording software, is used to ensure accurate capture of the original audio signal emitted by the digital guqin. Appropriate sampling frequencies (e.g., 44.1kHz, 48kHz, 96kHz) and bit depths (e.g., 16-bit, 24-bit) are set to ensure the acquired audio signal has sufficiently high quality and resolution. The performer plays the digital guqin according to specific performance standards and dynamic requirements. The repertoire or note sequences should cover performance segments with different dynamic levels, including soft, medium, and loud dynamics.

[0034] The acquired raw audio signal is imported into professional audio processing software, such as a digital audio workstation or other audio analysis and processing tools, to perform amplitude analysis on the pre-processed audio signal, determining the signal's amplitude range and dynamic variation characteristics. Based on the analysis results, different velocity thresholds are set to divide the audio signal into weak, medium, and strong velocity audio samples. For example, these three levels can be divided based on a certain volume range or amplitude magnitude, ensuring that the audio samples at each velocity level can represent the performance characteristics at the corresponding velocity. Further refinement can be performed on the audio samples at each velocity level, such as appropriately adjusting the gain of weak velocity samples to increase their volume and make them more clearly distinguishable in subsequent use; and appropriately compressing strong velocity samples to avoid signal overload, etc.

[0035] After stratified sampling, the audio samples at each velocity level are labeled and classified to ensure that each sample can be accurately classified into the corresponding velocity category. The audio samples at each velocity level are then checked for quality, and samples with obvious defects or that do not meet the requirements are removed.

[0036] Step S20: Generate analog audio signals for the digital guqin using a preset string vibration algorithm, and perform weighted synthesis processing on the analog audio signals and the original audio signals to generate an initial synthesized audio signal;

[0037] Specifically, the preset string vibration algorithm can be the Karplus-Strong algorithm, which simulates the vibration of the guqin string. The core parameters include:

[0038] String length (corresponding to pitch, such as C4=65cm), tension, damping coefficient (controlling the decay rate).

[0039] Excitation model: The initial excitation signal that simulates string plucking (such as an exponentially decaying pulse) or string brushing (noise modulation);

[0040] Frequency domain correction: The harmonic attenuation curve (such as the 1 / f slope) is adjusted by simulating the resonance of the guqin cavity through an all-pole filter (IIR).

[0041] Add vibrato and beat effects to enhance realism.

[0042] The performance parameters (dynamics, pitch, technique) are input into the algorithm, and the output is dynamically adjusted. Strong dynamics increase high-frequency harmonic components, weak dynamics shorten decay time, glissando is achieved by continuously adjusting the string length, and overtones suppress the fundamental frequency and enhance specific harmonics.

[0043] Ensure that the original audio signals of the digital guqin have been acquired and processed through layered sampling, identifying audio samples for weak, medium, and strong dynamics. Use these original audio samples as one input source for subsequent weighted synthesis. Based on actual timbre optimization requirements, select corresponding analog audio signal samples and original audio signal samples for pairing. For example, for a specific note and dynamic level, select the corresponding analog and original audio samples as the two base signals for synthesis.

[0044] Weighting coefficients are set based on the characteristics of the target timbre and the degree of emphasis placed on the analog and original signals. These weighting coefficients are typically values ​​between 0 and 1, corresponding to the relative weights of the analog and original audio signals in the synthesis process. For example, if you want the synthesized timbre to lean more towards the characteristics of the analog signal, you can set the weighting coefficient for the analog audio signal to 0.7 and the weighting coefficient for the original audio signal to 0.3; conversely, if you want to retain more of the original timbre's characteristics, adjust the corresponding coefficients.

[0045] The selected analog audio signal and the original audio signal are added point-by-point according to the set weighting coefficients. That is, for the sample values ​​of the two signals at the same time point, they are multiplied by the corresponding weighting coefficients and then added together to obtain the sample values ​​of the initial synthesized audio signal. This operation is performed sequentially over the entire time range of the audio signal to finally generate the complete initial synthesized audio signal.

[0046] Step S30: Determine the target timbre envelope parameters of the digital guqin based on the preset physical modeling engine, preset timbre deviation parameters, and preset fingering rules;

[0047] Specifically, based on the physical structure and sound production principle of the digital guqin, a preliminary physical modeling engine is configured. This includes inputting physical parameters such as the material, length, diameter, and tension of the guqin strings, as well as data on the wood properties and resonance cavity structure of the instrument, enabling the engine to basically simulate the sound production characteristics of the guqin. Simultaneously, a pre-set timbre deviation parameter database is prepared. This database contains characteristic data showing potential deviations in timbre from the ideal state under different performance conditions, such as spectral shifts and dynamic range changes in timbre under different dynamics, pitches, and performance environments. Furthermore, a pre-set set of fingering rules is compiled, detailing the characteristics of various guqin fingering techniques (such as stroking, plucking, hooking, flicking, strumming, and supporting), their impact on string vibration, and corresponding timbre characteristics. The previously acquired raw audio signals and the initial synthesized audio signals obtained through weighted synthesis are input into audio analysis software to extract their timbre characteristic parameters, including frequency response curves, harmonic component distribution, envelope shape (characteristics of attack, sustain, decay, etc.), dynamic range, and other key information. By comparing the differences between these parameters and the ideal guqin timbre characteristics, we can clarify the direction and degree of optimization needed, providing a reference for determining the target timbre envelope using a physical modeling engine and relevant rule parameters.

[0048] Based on the extracted audio feature differences, parameter adjustments and simulation calculations are performed in a preset physical modeling engine. For example, if the brightness of the initial synthesized audio in the high-pitched range is insufficient, the frequency response and harmonic changes of the high-pitched timbre under different parameter combinations can be simulated by adjusting the string tension, density, and other parameters in the physical model. This allows observation of which parameter adjustments can make the brightness of the high-pitched range closer to the ideal guqin timbre characteristics. Simultaneously, by combining the vibration modes and acoustic response data output by the physical modeling engine, the vibration amplitude, frequency change patterns, and attenuation characteristics of the string under different playing strengths are analyzed. This simulates the corresponding dynamic range and envelope shape changes of the timbre, and initially generates a set of candidate timbre envelope parameters based on physical modeling.

[0049] Referring to a pre-defined database of timbre deviation parameters, we identify timbre deviations that match the current digital guqin performance scenario and audio characteristics. For example, under specific environmental humidity and temperature conditions, the guqin timbre may exhibit a certain degree of spectral shift and dynamic compression. We retrieve the corresponding correction parameter ranges from the database for such deviations. These correction parameters are then applied to the timbre envelope candidate parameters output by the physical modeling engine. This appropriately compensates for the frequency response and adjusts the dynamic range by expanding or compressing it to counteract the impact of timbre deviation factors on the timbre envelope, making the timbre envelope parameters closer to the actual ideal guqin timbre under standard conditions.

[0050] By invoking a preset set of fingering rules, the timbre envelope parameters are further refined and adjusted for different guqin fingering techniques. For example, for the "mo" fingering, the initial vibration of the string is relatively gentle, the onset time is relatively long, and the high-frequency components of the timbre are fewer; while the "tiao" fingering has a faster onset and more high-frequency overtones. Based on these fingering rules, targeted adjustments are made to the onset, decay, and harmonic component changes of the timbre envelope, so that the target timbre envelope can accurately reflect the subtle differences in guqin timbre under different fingering techniques, enhancing the realism and expressiveness of the timbre. By traversing various common fingerings and adjusting the corresponding envelope parameters, a comprehensive and detailed set of digital guqin target timbre envelope parameters that meet the requirements of different fingering techniques is finally obtained.

[0051] Step S40: Generate the target audio signal for the digital guqin based on the target timbre envelope parameters and the initial synthesized audio signal.

[0052] In one embodiment, an envelope modulator is used to adjust the envelope of the initial synthesized audio signal based on information such as the attack time, decay time, hold level, and release time in the target timbre envelope parameters. For example, if the attack time of the target timbre envelope is long, the initial stage of the initial synthesized audio signal can be gradually increased to the target volume to simulate the natural attack process during guqin performance; regarding the decay time, the rate at which the signal gradually decreases after reaching its peak is set according to the target parameters to conform to the natural decay characteristics of the guqin timbre.

[0053] Analyze the adjusted audio signal to check for timbre distortion or unnaturalness. If distortion or a mismatch with the target timbre is found in certain frequency ranges after envelope adjustment, digital filters can be used to process the signal. For example, a low-pass filter can be used to remove high-frequency noise, or an equalizer can be used to fine-tune specific frequencies to optimize the frequency response of the timbre, making it closer to the timbre characteristics defined by the target timbre envelope.

[0054] The generated audio signal undergoes both subjective listening tests and objective measurements. Music producers, guqin players, and other professionals are invited to conduct listening evaluations, and their feedback on the timbre is collected. Simultaneously, equipment such as a spectrum analyzer and distortion meter are used to objectively measure the sound quality of the audio signal. Based on the evaluation results, the envelope parameter adjustment and filtering processes in the first two steps are further optimized until the timbre of the audio signal achieves the expected effect—that is, a high degree of matching with the target timbre envelope, and a pure, natural sound quality that accurately reflects the ideal timbre of the digital guqin.

[0055] After optimizing the sound quality and confirming that the audio signal meets the target timbre requirements, the processed audio signal is exported as the final target audio signal file. A suitable audio format is selected to save the file to ensure no loss of sound quality. This target audio signal can then be used in actual playback, recording, or other music production scenarios for the digital guqin, providing users with a more authentic and beautiful sound experience.

[0056] This embodiment discloses an AI-based method for optimizing the timbre of a digital guqin. The method includes: acquiring the original audio signal of the digital guqin; performing layered sampling processing on the original audio signal to determine audio samples at different dynamic levels; generating a simulated audio signal of the digital guqin using a preset string vibration algorithm; and performing weighted synthesis processing on the simulated audio signal and the original audio signal to generate an initial synthesized audio signal; determining the target timbre envelope parameters of the digital guqin based on a preset physical modeling engine, preset timbre deviation parameters, and preset fingering rules; and generating the target audio signal of the digital guqin based on the target timbre envelope parameters and the initial synthesized audio signal. Through this method, this application accurately captures the timbre characteristics of guqin performance at different dynamic levels by acquiring the original audio signal of the digital guqin and performing layered sampling processing. The use of a preset string vibration algorithm to generate a simulated audio signal, and the weighted synthesis method rationally adjusting the composition according to the characteristics of different signals, makes the synthesized audio signal more natural and harmonious in terms of timbre integration and overall performance, enhancing the fullness and layering of the timbre. The target timbre envelope parameters define the dynamic range and characteristics of the timbre, improving the realism of the digital guqin timbre.

[0057] Please see Figure 2 , Figure 2 This is a schematic flowchart illustrating an AI-based digital guqin tone optimization method provided in the second embodiment of this application. This AI-based digital guqin tone optimization method can be applied to an AI-based digital guqin tone optimization system. It is used to accurately capture the tone characteristics of the guqin at different playing intensities by acquiring raw audio signals at weak, medium, and strong dynamic ranges, solving the problem of insufficient dynamic range in traditional sampling sound sources and improving the realism of the digital guqin tone.

[0058] based on Figure 1 The illustrated embodiment, for example Figure 2 As shown, step S20 includes steps S201 to S202.

[0059] Step S201: Determine the current weight coefficients that match each dynamic audio sample through the distributed dynamic database, wherein the current weight coefficients include the weight coefficients of weak dynamic samples, the weight coefficients of medium dynamic samples, and the weight coefficients of strong dynamic samples.

[0060] Step S202: Determine the initial synthesized audio signal based on each current weight coefficient, each force audio sample, the simulated audio signal, and the preset string vibration algorithm.

[0061] In one embodiment, a communication connection is established with a distributed dynamics database to ensure access to a large amount of dynamics audio sample data and corresponding weight coefficient information stored therein. The dynamics audio sample data usually comes from guqin performance samples generated by many different performers in various performance scenarios, covering performance situations with various dynamics such as soft, medium, and loud.

[0062] The process involves acquiring key feature information about the current digital guqin performance scene, such as the style of the piece being played (e.g., classical, modern), tempo, note density, and the performer's overall control of dynamics. This information serves as the basis for filtering matching data from a distributed dynamics database. Based on the extracted features of the current performance scene, queries and matching operations are performed in the distributed dynamics database. The process then involves identifying the most similar historical performance samples to the current performance scene and obtaining the corresponding weight coefficients for weak, medium, and strong dynamics samples. This requires data retrieval algorithms and similarity matching techniques, such as cosine similarity calculation based on feature vectors, to accurately find the matching weight coefficients.

[0063] Prepare the weak, medium, and strong audio samples obtained from previous acquisition and layered processing, as well as the simulated audio signal generated by the preset string vibration algorithm. Ensure that these audio data are consistent in parameters such as duration and sampling rate to facilitate subsequent synthesis processing.

[0064] This embodiment discloses an AI-based method for optimizing the timbre of a digital guqin. The method includes: acquiring the original audio signal of the digital guqin; performing layered sampling processing on the original audio signal to determine audio samples at different dynamic levels; determining current weight coefficients matching each dynamic level audio sample using a distributed dynamic level database, wherein the current weight coefficients include weight coefficients for weak, medium, and strong dynamic level samples; determining an initial synthesized audio signal based on the current weight coefficients, the audio samples at different dynamic levels, the simulated audio signal, and a preset string vibration algorithm; determining the target timbre envelope parameters of the digital guqin based on a preset physical modeling engine, preset timbre deviation parameters, and preset fingering rules; and generating the target audio signal of the digital guqin based on the target timbre envelope parameters and the initial synthesized audio signal. Through this method, this application accurately captures the timbre characteristics of the guqin at different playing intensities by acquiring original audio signals at weak, medium, and strong dynamic levels, solving the problem of insufficient dynamic range in traditional sampling sound sources and improving the realism of the digital guqin's timbre.

[0065] based on Figure 2 In the illustrated embodiment, step S202 includes:

[0066] Calculate the corresponding sub-signals of each dynamic audio sample based on each dynamic audio sample and each current weighting coefficient.

[0067] The initial synthesized audio signal is generated by superimposing the various force audio sub-signals and the analog audio signal through a preset string vibration algorithm.

[0068] in, Where S is the initial synthesized audio signal, For weak-weighted samples, The sample weighting coefficients are for medium strength. For strong sample weighting coefficients, This is a weak-force audio sample. This is a medium-intensity audio sample. For high-intensity audio samples, To simulate audio signals, This is the preset mixing ratio for the preset string vibration algorithm.

[0069] Specifically, the matched current weight coefficients are applied to the corresponding dynamic audio samples and analog audio signals.

[0070] Specifically, for the audio signal value at each time point, the weak-intensity audio sample value is multiplied by a weak-intensity sample weighting coefficient, the medium-intensity audio sample value by a medium-intensity sample weighting coefficient, the strong-intensity audio sample value by a strong-intensity sample weighting coefficient, and the analog audio signal value by its corresponding weighting coefficient (this coefficient can be determined individually according to actual needs or associated with other intensity coefficients). Then, these weighted signal values ​​are added together to obtain the signal value of the initial synthesized audio signal at each time point, ultimately forming the complete initial synthesized audio signal.

[0071] Please see Figure 3 , Figure 3 This is a schematic flowchart illustrating an AI-based digital guqin tone optimization method provided in the third embodiment of this application. This AI-based digital guqin tone optimization method can be applied to an AI-based digital guqin tone optimization system. It is used to acquire raw audio signals and perform layered sampling to determine audio samples of various dynamic levels, providing basic materials for subsequent tone optimization. A distributed network is used to acquire preset audio samples and establish a distributed dynamics database. An initial audio signal is synthesized by combining the current weighting coefficients. Then, based on a physical modeling engine, the target tone envelope parameters are determined, ultimately generating the target audio signal. This method considers the influence of different dynamics, techniques, and musical styles on the tone, making the digital guqin tone more personalized and artistically expressive, better meeting the needs of different performers and musical styles. A preset physical modeling engine is used to analyze fingering rules and tone deviation parameters to determine the target tone envelope parameters, further refining the adjustment and optimization of the tone.

[0072] based on Figure 1 The illustrated embodiment, for example Figure 2 As shown, steps S2011 to S2014 precede step S201.

[0073] Step S2011: Obtain each preset audio sample through a distributed network, determine each preset velocity value range based on each preset audio sample, and associate basic weights with each preset velocity value range.

[0074] Specifically, distributed network technology is used to connect multiple servers or nodes storing digital guqin audio samples. These samples are generated by different performers under various playing conditions, covering a wide range of dynamics, techniques, and styles. A large number of preset audio samples are downloaded from these network resources to ensure the diversity and richness of the samples, providing a sufficient data foundation for subsequent analysis and processing.

[0075] Dynamics analysis is performed on each pre-defined audio sample. Using audio analysis software or algorithms, dynamics features such as peak volume and dynamic range are extracted from each sample. Based on these features, the samples are divided into different dynamic levels, determining pre-defined dynamic ranges, such as weak (0-40 V / L), medium (41-80 V / L), and strong (81-120 V / L). Simultaneously, based on the general performance and importance of each dynamic range in guqin playing, an initial base weight is assigned to them. This base weight can be set based on common performance statistics; for example, a relatively high base weight is initially given to the medium dynamic range because it is likely to be more common in most performance scenarios.

[0076] Step S2012: Extract the preset technique tags and preset music style tags of each preset audio sample, and generate preset art tags based on the preset technique tags and preset music style tags;

[0077] Specifically, for each preset audio sample, audio content analysis algorithms and metadata extraction techniques are used to identify and extract preset technique tags (such as guqin playing techniques like "mo," "tiao," "gou," and "ti") and preset musical style tags (such as classical, modern, and folk music styles). Then, based on these extracted technique and musical style tags, a comprehensive preset artistic tag is generated using preset tag association rules or machine learning models. For example, a sample containing a large number of "mo" and "tiao" techniques and with a "classical" musical style might generate a preset artistic tag of "classical mo and tiao style."

[0078] Step S2013: Correct the basic weights by using preset art labels and determine the preset weight coefficients for each preset audio sample;

[0079] Specifically, the generated preset artistic labels are combined with the previously set basic weights. Different artistic labels may correspond to different dynamic ranges and artistic requirements. For example, samples labeled "intense modern style" may be more valuable in strong dynamic ranges, while samples labeled "delicate classical style" are more important in weak dynamic ranges. Based on the characteristics of these artistic labels and the expected performance in different performance scenarios, the basic weights are adjusted and corrected. For example, for "intense modern style" samples, the weight coefficient in the strong dynamic range is increased; for "delicate classical style" samples, the weight coefficient in the weak dynamic range is increased. In this way, the preset weight coefficients corresponding to each preset audio sample in different dynamic ranges are determined, so that the weight coefficients can more accurately reflect the actual value and applicability of the samples under different artistic styles and performance techniques.

[0080] Step S2014: Establish a distributed intensity database based on each preset weight coefficient, each preset art label, and each preset intensity value range.

[0081] In one embodiment, each preset audio sample, along with its corresponding preset weight coefficients, preset artistic labels, and preset velocity ranges, is integrated. A distributed database management system is used to create the database structure across multiple servers or storage nodes. A database record is created for each audio sample, containing information such as the sample's audio data, preset artistic labels, preset velocity ranges, and corresponding preset weight coefficients. Appropriate indexing and query mechanisms are established to enable rapid retrieval and use of these audio samples and their related data based on conditions such as velocity ranges and artistic labels, thereby completing the establishment of the distributed velocity database.

[0082] This embodiment discloses an AI-based method for optimizing the tone of a digital guqin. The method includes: acquiring the original audio signal of the digital guqin; performing layered sampling processing on the original audio signal to determine audio samples of various dynamic ranges; acquiring preset audio samples through a distributed network; determining preset dynamic ranges based on the preset audio samples; associating basic weights with each preset dynamic range; extracting preset technique tags and preset style tags from each preset audio sample; generating preset art tags based on the preset technique tags and preset style tags; correcting the basic weights using the preset art tags to determine the preset weight coefficients for each preset audio sample; and further optimizing the tone of each preset audio sample based on the preset art tags. A distributed velocity database is established by pre-setting weight coefficients, pre-setting artistic tags, and pre-setting velocity value ranges. The distributed velocity database is used to determine the current weight coefficients that match each velocity audio sample. These current weight coefficients include those for weak, medium, and strong velocity samples. Based on these current weight coefficients, the velocity audio samples, the simulated audio signal, and a pre-set string vibration algorithm, an initial synthesized audio signal is determined. The target timbre envelope parameters for the digital guqin are determined based on a pre-set physical modeling engine, pre-set timbre deviation parameters, and pre-set fingering rules. The target audio signal for the digital guqin is generated based on the target timbre envelope parameters and the initial synthesized audio signal. Through this method, this application acquires raw audio signals and performs layered sampling to determine the velocity audio samples, providing foundational materials for subsequent timbre optimization. By utilizing a distributed network to acquire preset audio samples and establish a distributed velocity database, and combining the current weight coefficients to synthesize an initial audio signal, the target timbre envelope parameters are determined based on a physical modeling engine, and finally the target audio signal is generated. The influence of different velocities, techniques and musical styles on timbre is considered, making the timbre of the digital guqin more personalized and artistically expressive, and better able to meet the needs of different performers and musical styles. The preset physical modeling engine is used to analyze fingering rules and timbre deviation parameters to determine the target timbre envelope parameters, further refining the adjustment and optimization of the timbre.

[0083] In a specific embodiment, step S201 includes:

[0084] If a preset weight coefficient exists in the distributed strength database that matches each strength audio sample, the preset weight coefficient will be determined as the current weight coefficient.

[0085] If there is no preset weight coefficient matching each velocity audio sample in the distributed velocity database, the current weight coefficient is determined from the preset distributed velocity database using the K nearest neighbor algorithm.

[0086] In one embodiment, if there are no preset weight coefficients matching each velocity audio sample in the distributed velocity database, the current weight coefficients are determined from the preset distributed velocity database using the K-nearest neighbor algorithm, including:

[0087] Determine the current velocity value and current art label of the velocity audio sample, and convert the current art label into an art label feature vector;

[0088] Broadcast an art tag query request in the distributed intensity database to determine the art tag to be determined, wherein the difference between the intensity value of the art tag to be determined and the current intensity value is less than a preset threshold.

[0089] Based on a preset distance formula, a preset number of neighboring audio samples are determined from each art tag to be determined, and the target similarity between each neighboring audio sample and the force audio sample is calculated.

[0090] The current weight coefficient is determined by performing a weighted average of each neighboring audio sample using a preset distance weight, where the preset distance weight is the reciprocal of the similarity of each target.

[0091] Specifically, when the distributed dynamics database does not contain preset weight coefficients that precisely match the current performance dynamics, technique, and style, the system will activate the K-Nearest Neighbors (KNN) algorithm to dynamically calculate the optimal weights. First, the current performance parameters (dynamic value, technique type, and style label) are transformed into feature vectors, with dynamic values ​​normalized and categorical variables encoded using one-hot encoding. Then, the system broadcasts a query request to all shard nodes, retrieving candidate samples within each shard that fall within a dynamics range of ±20. The similarity between each candidate sample and the current feature is calculated using a hybrid distance formula (Euclidean distance for dynamics and Hamming distance for categorical features). After the coordinating node collects all candidate samples returned from all shards, it selects K nearest neighbor samples (default K=5) in ascending distance order. These neighbors' preset weight coefficients are then weighted and averaged using the reciprocal of the distance, ultimately generating a weight coefficient combination (W_weak, W_medium, W_strong) suitable for the current performance. The calculation results are verified by user ratings. If the rating is below a threshold, the K value is automatically increased and the calculation is recalculated. Simultaneously, the feedback data is written to the blockchain for subsequent model optimization. To improve real-time performance, each segment uses a KD-tree to accelerate nearest neighbor search and caches high-frequency query results using Redis. This solution achieves a 10ms dynamic response while maintaining the naturalness of the timbre, and the weight interpolation error is controlled within 10% of the expert annotation.

[0092] based on Figure 1 In the illustrated embodiment, step S30 includes:

[0093] The preset fingering rules are analyzed by a preset physical modeling engine to determine the parameters of the low-frequency oscillator and the envelope amplitude parameters;

[0094] The preset timbre deviation parameters are analyzed by the preset physical modeling engine to generate pitch fine-tuning parameters, harmonic enhancement parameters and damping attenuation parameters.

[0095] The low-frequency oscillator parameters, envelope amplitude parameters, pitch fine-tuning parameters, harmonic enhancement parameters, and damping attenuation parameters are determined as the target timbre envelope parameters.

[0096] Specifically, the preset physical modeling engine is launched, which has been pre-configured with the physical model parameters of the digital guqin, such as string parameters (length, linear density, tension, etc.), instrument body parameters (wood density, resonant cavity size, etc.), and air damping parameters. At the same time, the preset set of fingering rules and set of tone deviation parameters are loaded. These rules and parameters are predefined according to the actual playing characteristics and tone requirements of the guqin.

[0097] Preset fingering rules are input into the physics modeling engine. The engine analyzes the impact of the playing actions defined in the fingering rules, such as left-hand fingering and glissando, and right-hand plucking force and angle, on string vibration. By simulating the vibration patterns of the strings under different fingerings, the engine calculates the corresponding low-frequency oscillator parameters, including the frequency, amplitude, and waveform of the low-frequency oscillation, to simulate subtle vibrato or periodic string changes during performance. Simultaneously, based on information such as striking force and duration in the fingering rules, envelope amplitude parameters, including attack time, decay time, sustain level, and release time, are determined to simulate the dynamic changes in the guqin's timbre.

[0098] Preset tone deviation parameters are input into the physical modeling engine. Based on these deviation parameters, such as pitch offset, harmonic distribution changes, and decay rate differences, the engine calculates corresponding pitch fine-tuning parameters to finely adjust the basic pitch, making the tone more consistent with pitch changes in actual performance; harmonic enhancement parameters are used to highlight or weaken specific harmonic components, thereby changing the brightness or softness of the tone; and damping attenuation parameters are used to adjust the tone decay rate, simulating the sound damping effect under different instrument body and performance conditions.

[0099] The low-frequency oscillator parameters, envelope amplitude parameters, pitch fine-tuning parameters, harmonic enhancement parameters, and damping attenuation parameters obtained above are integrated. Through the internal algorithm of the physical modeling engine, these parameters are correlated and coordinated to ensure they work together in the timbre synthesis process to produce the desired timbre effect. Finally, the integrated parameter set is determined as the target timbre envelope parameters. These parameters will guide subsequent timbre synthesis and processing, enabling the digital guqin to generate audio signals with ideal timbre characteristics.

[0100] In a specific embodiment, step S40 includes:

[0101] The time interval of the initial synthesized audio signal is divided by the envelope amplitude parameter;

[0102] The cutoff frequency of the initial synthesized audio signal is adjusted by using low-frequency oscillator parameters, harmonic enhancement parameters, and damping attenuation parameters.

[0103] The dynamic changes in pitch of the initial synthesized audio signal are determined by the pitch fine-tuning parameter;

[0104] The target audio signal is generated by dynamically adjusting the initial synthesized audio signal through time intervals, cutoff frequencies, and pitch changes.

[0105] Specifically, based on the attack time, decay time, hold time, and release time in the envelope amplitude parameters, the initial synthesized audio signal is divided into multiple time intervals. For example, during the attack time, the signal gradually increases; during the decay time, the signal weakens from its peak; during the hold time, the signal maintains a certain amplitude; and during the release time, the signal gradually decays until it ends. The cutoff frequency of the filter is determined by combining the low-frequency oscillator parameters, harmonic enhancement parameters, and damping attenuation parameters. The low-frequency oscillator parameters affect the periodic change of the cutoff frequency, producing a vibrato effect; the harmonic enhancement parameters determine which frequency components are emphasized, thus changing the brightness of the timbre; and the damping attenuation parameters determine the decay rate of high-frequency components, making the sound more natural.

[0106] By utilizing pitch fine-tuning parameters, the pitch offset of the initial synthesized audio signal at each time point is calculated, thus determining the dynamic changes in pitch. This simulates pitch fluctuations in real performances, enhancing the vividness of the audio. The time intervals, cutoff frequencies, and dynamic pitch changes obtained in the above steps are applied to the initial synthesized audio signal. The signal is then precisely adjusted using audio processing software or audio processing libraries in a programming language to ensure that the signal in each time interval meets the target timbre requirements. Finally, the generated target audio signal should possess a more natural and vivid timbre and dynamic changes, meeting the standards for high-quality audio synthesis.

[0107] Please see Figure 4 , Figure 4 This application provides a schematic block diagram of an AI-based digital guqin tone optimization system, which is used to execute the aforementioned AI-based digital guqin tone optimization method. This AI-based digital guqin tone optimization system can be configured on a server.

[0108] like Figure 4 As shown, the AI-based digital guqin tone optimization system 400 includes:

[0109] The dynamic audio sample determination module 410 is used to acquire the original audio signal of the digital guqin and perform layered sampling processing on the original audio signal to determine each dynamic audio sample.

[0110] The initial synthesized audio signal generation module 420 is used to generate an analog audio signal of the digital guqin through a preset string vibration algorithm, and to perform weighted synthesis processing on the analog audio signal and the original audio signal to generate an initial synthesized audio signal;

[0111] The target timbre envelope parameter determination module 430 is used to determine the target timbre envelope parameters of the digital guqin based on a preset physical modeling engine, preset timbre deviation parameters, and preset fingering rules.

[0112] The target audio signal generation module 440 is used to generate the target audio signal of the digital guqin based on the target timbre envelope parameters and the initial synthesized audio signal.

[0113] Furthermore, the initial synthesized audio signal generation module 420 includes:

[0114] The current weight coefficient determination submodule is used to determine the current weight coefficients that match each intensity audio sample through a distributed intensity database. The current weight coefficients include the weight coefficients of weak intensity samples, medium intensity samples, and strong intensity samples.

[0115] The initial synthesized audio signal generation submodule is used to determine the initial synthesized audio signal based on each current weight coefficient, each dynamic audio sample, the analog audio signal, and the preset string vibration algorithm.

[0116] Furthermore, the initial synthesized audio signal generation submodule includes:

[0117] The dynamic audio sub-signal calculation unit is used to calculate the dynamic audio sub-signal corresponding to each dynamic audio sample based on each dynamic audio sample and each current weight coefficient.

[0118] The initial synthesized audio signal generation unit is used to superimpose the various force audio sub-signals and analog audio signals through a preset string vibration algorithm to generate the initial synthesized audio signal.

[0119] Furthermore, the initial synthesized audio signal generation submodule includes:

[0120] The basic weight association unit is used to obtain each preset audio sample through a distributed network, determine each preset velocity value range based on each preset audio sample, and associate basic weights with each preset velocity value range.

[0121] The preset art tag generation unit is used to extract the preset technique tags and preset music style tags of each preset audio sample, and generate preset art tags based on the preset technique tags and preset music style tags;

[0122] The preset weight coefficient determination unit is used to correct each basic weight through preset art labels and determine the preset weight coefficient of each preset audio sample.

[0123] The distributed intensity database establishment unit is used to establish a distributed intensity database based on each preset weight coefficient, each preset art label, and each preset intensity value range.

[0124] Furthermore, the current weight coefficient determination submodule includes:

[0125] The first current weight coefficient determination unit is used to determine the preset weight coefficient as the current weight coefficient if there is a preset weight coefficient in the distributed strength database that matches each strength audio sample.

[0126] The second current weight coefficient determination unit is used to determine each current weight coefficient from the preset distributed velocity database by means of the K nearest neighbor algorithm if there is no preset weight coefficient that matches each velocity audio sample in the distributed velocity database.

[0127] Furthermore, the second current weight coefficient determination unit includes:

[0128] The art label feature vector transformation subunit is used to determine the current velocity value and current art label of the velocity audio sample, and convert the current art label into an art label feature vector;

[0129] The art tag determination subunit is used to broadcast art tag query requests in the distributed intensity database to determine the art tag to be determined, wherein the difference between the intensity value of the art tag to be determined and the current intensity value is less than a preset threshold.

[0130] The target similarity calculation subunit is used to determine a preset number of neighboring audio samples from each art tag to be determined based on a preset distance formula, and to calculate the target similarity between each neighboring audio sample and the velocity audio sample.

[0131] The current weight coefficient determination subunit is used to perform weighted averaging of each neighboring audio sample using a preset distance weight to determine the current weight coefficient, where the preset distance weight is the reciprocal of the similarity of each target.

[0132] Furthermore, the target timbre envelope parameter determination module 430 includes:

[0133] The first type of parameter parsing submodule is used to parse the preset fingering rules through the preset physical modeling engine to determine the low-frequency oscillator parameters and envelope amplitude parameters;

[0134] The second type of parameter parsing submodule is used to parse preset timbre deviation parameters through a preset physical modeling engine, and generate pitch fine-tuning parameters, harmonic enhancement parameters and damping attenuation parameters.

[0135] The target timbre envelope parameter determination submodule is used to determine the low-frequency oscillator parameters, envelope amplitude parameters, pitch fine-tuning parameters, harmonic enhancement parameters, and damping attenuation parameters as the target timbre envelope parameters.

[0136] Furthermore, the target audio signal generation module 440 includes:

[0137] The time interval determination submodule is used to divide the time interval of the initial synthesized audio signal using the envelope amplitude parameter;

[0138] The cutoff frequency determination submodule is used to adjust the cutoff frequency of the initial synthesized audio signal by using low-frequency oscillator parameters, harmonic enhancement parameters, and damping attenuation parameters.

[0139] The pitch dynamic change determination submodule is used to determine the pitch dynamic change of the initial synthesized audio signal through pitch fine-tuning parameters;

[0140] The target audio signal generation submodule is used to dynamically adjust the initial synthesized audio signal by changing the time interval, cutoff frequency, and pitch to generate the target audio signal.

[0141] It should be noted that those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the system and each module described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0142] The aforementioned apparatus can be implemented as a computer program, which can be used in, for example... Figure 5 It runs on the computer device shown.

[0143] Please see Figure 5 , Figure 5 This is a schematic block diagram illustrating the structure of a computer device according to an embodiment of this application. The computer device may be a server.

[0144] See Figure 5 The computer device includes a processor, memory, and network interface connected via a system bus, wherein the memory may include non-volatile storage media and internal memory.

[0145] The non-volatile storage medium can store the operating system and computer program. The computer program includes program instructions that, when executed, cause the processor to perform any artificial intelligence-based digital guqin tone optimization method.

[0146] The processor provides computing and control capabilities, supporting the operation of the entire computer device.

[0147] The internal memory provides an environment for the execution of computer programs in non-volatile storage media. When the computer program is executed by the processor, it can enable the processor to execute any artificial intelligence-based digital guqin tone optimization method.

[0148] This network interface is used for network communication, such as sending assigned tasks. Those skilled in the art will understand that... Figure 5The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0149] It should be understood that the processor can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among these, a general-purpose processor can be a microprocessor or any conventional processor.

[0150] In one embodiment, the processor is configured to run a computer program stored in memory to perform the following steps:

[0151] The original audio signal of the digital guqin was acquired, and the original audio signal was processed by layered sampling to determine the audio samples of each dynamic range.

[0152] The analog audio signal of the digital guqin is generated by a preset string vibration algorithm, and the analog audio signal and the original audio signal are weighted and synthesized to generate the initial synthesized audio signal.

[0153] The target timbre envelope parameters of the digital guqin are determined based on a preset physical modeling engine, preset timbre deviation parameters, and preset fingering rules.

[0154] The target audio signal for the digital guqin is generated based on the target timbre envelope parameters and the initial synthesized audio signal.

[0155] In one embodiment, the analog audio signal and the original audio signal are weighted and synthesized to generate an initial synthesized audio signal, which is used to achieve:

[0156] The current weight coefficients that match each intensity audio sample are determined by a distributed intensity database. The current weight coefficients include the weight coefficients of weak intensity samples, medium intensity samples, and strong intensity samples.

[0157] The initial synthesized audio signal is determined based on the current weight coefficients, the audio samples of each force, the simulated audio signal, and the preset string vibration algorithm.

[0158] In one embodiment, an initial synthesized audio signal is determined based on each current weight coefficient, each dynamic audio sample, the analog audio signal, and a preset string vibration algorithm, for the purpose of:

[0159] Calculate the corresponding sub-signals of each dynamic audio sample based on each dynamic audio sample and each current weighting coefficient.

[0160] The initial synthesized audio signal is generated by superimposing the various force audio sub-signals and the analog audio signal through a preset string vibration algorithm.

[0161] in, Where S is the initial synthesized audio signal, For weak-weighted samples, The sample weighting coefficients are for medium strength. For strong sample weighting coefficients, This is a weak-force audio sample. This is a medium-intensity audio sample. For high-intensity audio samples, To simulate audio signals, This is the preset mixing ratio for the preset string vibration algorithm.

[0162] In one embodiment, before determining the current weight coefficients that match each velocity audio sample using a distributed velocity database, the following is implemented:

[0163] Each preset audio sample is obtained through a distributed network, each preset velocity value range is determined based on each preset audio sample, and basic weights are associated with each preset velocity value range.

[0164] Extract the preset technique tags and preset music style tags from each preset audio sample, and generate preset art tags based on the preset technique tags and preset music style tags;

[0165] By adjusting the basic weights using preset art labels, the preset weight coefficients for each preset audio sample are determined.

[0166] A distributed intensity database is established based on each preset weight coefficient, each preset art label, and each preset intensity value range.

[0167] In one embodiment, a distributed velocity database is used to determine the current weight coefficients that match each velocity audio sample, for the purpose of:

[0168] If a preset weight coefficient exists in the distributed strength database that matches each strength audio sample, the preset weight coefficient will be determined as the current weight coefficient.

[0169] If there is no preset weight coefficient matching each velocity audio sample in the distributed velocity database, the current weight coefficient is determined from the preset distributed velocity database using the K nearest neighbor algorithm.

[0170] In one embodiment, if there are no preset weight coefficients matching each velocity audio sample in the distributed velocity database, the K-nearest neighbor algorithm is used to determine each current weight coefficient from the preset distributed velocity database, for the following purpose:

[0171] Determine the current velocity value and current art label of the velocity audio sample, and convert the current art label into an art label feature vector;

[0172] Broadcast an art tag query request in the distributed intensity database to determine the art tag to be determined, wherein the difference between the intensity value of the art tag to be determined and the current intensity value is less than a preset threshold.

[0173] Based on a preset distance formula, a preset number of neighboring audio samples are determined from each art tag to be determined, and the target similarity between each neighboring audio sample and the force audio sample is calculated.

[0174] The current weight coefficient is determined by performing a weighted average of each neighboring audio sample using a preset distance weight, where the preset distance weight is the reciprocal of the similarity of each target.

[0175] In one embodiment, the target timbre envelope parameters of the digital guqin are determined based on a preset physical modeling engine, preset timbre deviation parameters, and preset fingering rules, for the purpose of:

[0176] The preset fingering rules are analyzed by a preset physical modeling engine to determine the parameters of the low-frequency oscillator and the envelope amplitude parameters;

[0177] The preset timbre deviation parameters are analyzed by the preset physical modeling engine to generate pitch fine-tuning parameters, harmonic enhancement parameters and damping attenuation parameters.

[0178] The low-frequency oscillator parameters, envelope amplitude parameters, pitch fine-tuning parameters, harmonic enhancement parameters, and damping attenuation parameters are determined as the target timbre envelope parameters.

[0179] In one embodiment, a target audio signal for the digital guqin is generated based on the target timbre envelope parameters and the initial synthesized audio signal, for the purpose of:

[0180] The time interval of the initial synthesized audio signal is divided by the envelope amplitude parameter;

[0181] The cutoff frequency of the initial synthesized audio signal is adjusted by using low-frequency oscillator parameters, harmonic enhancement parameters, and damping attenuation parameters.

[0182] The dynamic changes in pitch of the initial synthesized audio signal are determined by the pitch fine-tuning parameter;

[0183] The target audio signal is generated by dynamically adjusting the initial synthesized audio signal through time intervals, cutoff frequencies, and pitch changes.

[0184] The embodiments of this application also provide a computer-readable storage medium storing a computer program, which includes program instructions. The processor executes the program instructions to implement any of the artificial intelligence-based digital guqin tone optimization methods provided in the embodiments of this application.

[0185] The computer-readable storage medium can be an internal storage unit of the computer device described in the foregoing embodiments, such as a hard disk or memory of the computer device. Alternatively, it can be an external storage device of the computer device, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc., provided on the computer device.

[0186] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for optimizing the tone of a digital guqin based on artificial intelligence, characterized in that, include: The original audio signal of the digital guqin was acquired, and the original audio signal was subjected to layered sampling processing to determine the audio samples of each dynamic range. The analog audio signal of the digital guqin is generated by a preset string vibration algorithm, and the analog audio signal and the original audio signal are weighted and synthesized to generate an initial synthesized audio signal. The target timbre envelope parameters of the digital guqin are determined based on a preset physical modeling engine, preset timbre deviation parameters, and preset fingering rules. The target audio signal of the digital guqin is generated based on the target timbre envelope parameters and the initial synthesized audio signal. The determination of the target timbre envelope parameters of the digital guqin based on a preset physical modeling engine, preset timbre deviation parameters, and preset fingering rules includes: The preset physical modeling engine parses the preset fingering rules to determine the low-frequency oscillator parameters and envelope amplitude parameters. The preset timbre deviation parameters are analyzed by the preset physical modeling engine to generate pitch fine-tuning parameters, harmonic enhancement parameters, and damping attenuation parameters. The low-frequency oscillator parameters, the envelope amplitude parameters, the pitch fine-tuning parameters, the harmonic enhancement parameters, and the damping attenuation parameters are determined as the target timbre envelope parameters; The step of generating the target audio signal for the digital guqin based on the target timbre envelope parameters and the initial synthesized audio signal includes: The time interval of the initial synthesized audio signal is divided by the envelope amplitude parameter; The cutoff frequency of the initial synthesized audio signal is adjusted by the low-frequency oscillator parameters, the harmonic enhancement parameters, and the damping attenuation parameters. The dynamic pitch change of the initial synthesized audio signal is determined by the pitch fine-tuning parameter; The target audio signal is generated by adjusting the initial synthesized audio signal through the time interval, the cutoff frequency, and the dynamic change of the pitch. The step of weighted combining the analog audio signal and the original audio signal to generate an initial synthesized audio signal includes: The current weight coefficients that match each of the said weight audio samples are determined by a distributed weight database, wherein the current weight coefficients include the weight coefficients of weak weight samples, medium weight samples, and strong weight samples. The initial synthesized audio signal is determined based on the current weight coefficients, the force audio samples, the analog audio signal, and the preset string vibration algorithm.

2. The method for optimizing the tone of a digital guqin based on artificial intelligence according to claim 1, characterized in that, The dynamic audio samples include weak dynamic audio samples, medium dynamic audio samples, and strong dynamic audio samples. The process of determining the initial synthesized audio signal based on the current weighting coefficients, the dynamic audio samples, the analog audio signal, and the preset string vibration algorithm includes: Calculate each dynamic audio sub-signal corresponding to each dynamic audio sample based on each dynamic audio sample and each current weighting coefficient; The initial synthesized audio signal is generated by superimposing the force audio sub-signals and the analog audio signal using the preset string vibration algorithm. in, Where S is the initial synthesized audio signal, The weighting coefficients for the weak force samples are... The weighting coefficients for the medium-strength samples are... The weighting coefficients for the strong samples are... For the weak force audio sample, The medium-intensity audio sample, The high-intensity audio sample, The analog audio signal, This is the preset mixing ratio for the preset string vibration algorithm.

3. The method for optimizing the tone of a digital guqin based on artificial intelligence according to claim 1, characterized in that, Before determining the current weight coefficients that match each of the said velocity audio samples through a distributed velocity database, the process includes: Each preset audio sample is obtained through a distributed network, each preset velocity value range is determined based on each preset audio sample, and a basic weight is associated with each preset velocity value range. Extract the preset technique tags and preset music style tags from each preset audio sample, and generate preset art tags based on the preset technique tags and preset music style tags; By correcting the basic weights using the preset art tags, the preset weight coefficients of each preset audio sample are determined. The distributed intensity database is established based on the preset weight coefficients, preset art labels, and preset intensity value ranges.

4. The method for optimizing the tone of a digital guqin based on artificial intelligence according to claim 3, characterized in that, The step of determining the current weight coefficients that match each of the said velocity audio samples through a distributed velocity database includes: If the distributed strength database contains a preset weight coefficient that matches each of the strength audio samples, the preset weight coefficient is determined as the current weight coefficient. If there is no preset weight coefficient in the distributed velocity database that matches each of the velocity audio samples, the current weight coefficient is determined from the preset distributed velocity database using the K nearest neighbor algorithm.

5. The method for optimizing the tone of a digital guqin based on artificial intelligence according to claim 4, characterized in that, If no preset weight coefficient matching each of the velocity audio samples exists in the distributed velocity database, the current weight coefficients are determined from the preset distributed velocity database using the K-nearest neighbor algorithm, including: Determine the current velocity value and current art label of the velocity audio sample, and convert the current art label into an art label feature vector; Broadcast an art tag query request in the distributed intensity database to determine the art tag to be determined, wherein the difference between the intensity value of the art tag to be determined and the current intensity value is less than a preset threshold. Based on a preset distance formula, a preset number of neighboring audio samples are determined from each of the art tags to be determined, and the target similarity between each of the neighboring audio samples and the dynamic audio samples is calculated. The neighboring audio samples are weighted and averaged using a preset distance weight to determine the current weight coefficient, wherein the preset distance weight is the reciprocal of the similarity of each target.

6. A digital guqin tone optimization system based on artificial intelligence, characterized in that, include: The dynamic audio sample determination module is used to acquire the original audio signal of the digital guqin and perform layered sampling processing on the original audio signal to determine each dynamic audio sample; An initial synthesized audio signal generation module is used to generate an analog audio signal for the digital guqin using a preset string vibration algorithm, and to perform weighted synthesis processing on the analog audio signal and the original audio signal to generate an initial synthesized audio signal, including: The current weight coefficients that match each of the said weight audio samples are determined by a distributed weight database, wherein the current weight coefficients include the weight coefficients of weak weight samples, medium weight samples, and strong weight samples. The initial synthesized audio signal is determined based on the current weight coefficients, the force audio samples, the analog audio signal, and the preset string vibration algorithm. The target timbre envelope parameter determination module is used to determine the target timbre envelope parameters of the digital guqin based on a preset physical modeling engine, preset timbre deviation parameters, and preset fingering rules; and to determine the low-frequency oscillator parameters and envelope amplitude parameters by parsing the preset fingering rules through the preset physical modeling engine. The preset timbre deviation parameters are analyzed by the preset physical modeling engine to generate pitch fine-tuning parameters, harmonic enhancement parameters, and damping attenuation parameters, including: The preset physical modeling engine parses the preset fingering rules to determine the low-frequency oscillator parameters and envelope amplitude parameters. The preset timbre deviation parameters are analyzed by the preset physical modeling engine to generate pitch fine-tuning parameters, harmonic enhancement parameters, and damping attenuation parameters. The low-frequency oscillator parameters, the envelope amplitude parameters, the pitch fine-tuning parameters, the harmonic enhancement parameters, and the damping attenuation parameters are determined as the target timbre envelope parameters; The target audio signal generation module is used to generate the target audio signal of the digital guqin based on the target timbre envelope parameters and the initial synthesized audio signal, including: The time interval of the initial synthesized audio signal is divided by the envelope amplitude parameter; The cutoff frequency of the initial synthesized audio signal is adjusted by the low-frequency oscillator parameters, the harmonic enhancement parameters, and the damping attenuation parameters. The dynamic pitch change of the initial synthesized audio signal is determined by the pitch fine-tuning parameter; The target audio signal is generated by adjusting the initial synthesized audio signal through the time interval, the cutoff frequency, and the dynamic changes in pitch.

Images