Dynamic music rhythm system based on nonlinear dynamics and generation method

Abstract

The invention relates to the technical field of music generation and human-computer interaction, and discloses a dynamic music rhythm system of nonlinear dynamics and a generation method. The dynamic music rhythm system based on nonlinear dynamics comprises an input sensing module, a nonlinear oscillator network core module, a parameter adaptive mapping module, a rhythm event extraction and synthesis module, an interaction feedback module and a storage and learning module. Dynamic rhythms are generated in real time through a nonlinear dynamic model, and system parameters are optimized through deep reinforcement learning, so that highly personalized music rhythm output is realized; the method has stronger dynamics, real-time performance and interactivity, is particularly suitable for various application scenes such as music performance, psychotherapy and healing, public art and the like, and has the advantages of important popularization value, application prospect and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of music generation and human-computer interaction, and in particular to a dynamic music rhythm system and a generation method based on nonlinear dynamics. Background Art

[0002] With the continuous development of music generation and human-computer interaction technologies, traditional rhythm generation systems often rely on preset beat templates or linear rhythm models. These methods can meet basic rhythm generation requirements to a certain extent, but they have significant limitations in handling complex and dynamically changing musical rhythms, making it difficult to achieve natural evolution and personalized expression of musical rhythms.

[0003] In recent years, researchers have attempted to introduce nonlinear dynamic models to simulate and analyze the complex structures in audio signals. For example, patent publication number CN102947883A proposes a method for analyzing audio signals using nonlinear oscillator networks. This method converts an acoustic input signal into a response in a nonlinear oscillator network to identify the structural characteristics of the audio signal.

[0004] However, this technology primarily focuses on the structural analysis of audio signals, aiming to simulate the human ear's perception of sound. It does not involve the generation or real-time control of musical rhythms. Its application scope is primarily limited to the recognition and analysis of audio signals, lacking support for the dynamic and interactive generation of musical rhythms.

[0005] Therefore, the existing technology has not yet provided a system that can use nonlinear dynamic models to achieve dynamic generation and real-time regulation of musical rhythm and has high personalization and interactivity. Summary of the Invention

[0006] The main purpose of the present invention is to solve the problem that most existing rhythm generation systems rely on preset beat templates or linear rhythm models. These methods have significant limitations in processing complex and dynamically changing musical rhythms, making it difficult to achieve natural evolution, personalized expression and real-time regulation of rhythms, and lack the ability to adapt to user interaction and feedback.

[0007] In a first aspect, the present invention provides a dynamic music rhythm system based on nonlinear dynamics, the dynamic music rhythm system based on nonlinear dynamics comprising:

[0008] An input sensing module, configured to collect at least one user input signal, wherein the user input signal includes a physical control signal, an emotional signal, a physiological signal, or an environmental beat signal;

[0009] The nonlinear oscillator network core module is used to receive the signal output by the input sensing module and perform calculations based on the following complex state equation, which is specifically:

[0010]

[0011] Among them, z i is the state of the i-th oscillator, α, β, κ ij , γ is an adjustable parameter, s(t) is the input signal coupling term;

[0012] A parameter adaptive mapping module, configured to update the adjustable parameters in real time based on the signal output by the input sensing module;

[0013] a rhythmic event extraction and synthesis module, configured to detect phase synchronization nodes in the oscillator network to generate a rhythmic event sequence, and drive a synthesis engine to output an audio signal accordingly;

[0014] Interactive feedback module, used to provide visual or auditory feedback to the user and receive secondary adjustment instructions from the user to achieve closed-loop control;

[0015] A storage and learning module, configured to record user historical preferences and train the adaptive mapping strategy of the parameter adaptive mapping module to improve the personalization and predictability of rhythm generation;

[0016] Among them, the nonlinear oscillator network core module and the parameter adaptive mapping module are coupled through high-speed shared memory or in-process calling to ensure that the system end-to-end delay is less than or equal to five milliseconds, thereby realizing the real-time performance of the dynamic music rhythm system.

[0017] Optionally, the input perception module includes a brain-computer interface subunit, which is used to map the EEG rhythm signal in the range of one to forty hertz into a driving force signal in the oscillator network core module.

[0018] Optionally, the parameter adaptive mapping module optimizes the reward function based on the Qianghu learning algorithm as follows:

[0019] R=λ1C sync +λ2C var -λ3C lat

[0020] Among them, C sync Indicates rhythm synchronization, C var Indicates rhythm complexity, C lat represents the system delay, and λ1, λ2, and λ3 are weight coefficients.

[0021] Optionally, the rhythm event extraction and synthesis module detects oscillator phase zero-crossing events based on a wavelet packet decomposition algorithm, and generates at least four-part rhythm tracks through a multi-channel sample synthesizer or a physical modeling synthesizer.

[0022] Optionally, the interactive feedback module provides a parameter visualization interface, which displays the phase plane diagram and Lyapunov exponent of the oscillator network in real time to indicate rhythm stability.

[0023] In a second aspect, the present application provides a method for generating a dynamic music rhythm based on nonlinear dynamics, which is used in the dynamic music rhythm system of nonlinear dynamics as described above. The method for generating a dynamic music rhythm based on nonlinear dynamics includes:

[0024] S1. Input acquisition, obtaining user input signals;

[0025] S2. Model initialization: establish N models with natural frequency ω according to the user input signal. i oscillator network.

[0026] S3. Adaptive parameter adjustment: calling mapping strategy f to update adjustable parameters in real time;

[0027] S4. Solve the dynamics using a fourth-order Runge-Kutta method for numerical integration with a step size of ≤ 2 milliseconds;

[0028] S5. Event detection: when the oscillator phase satisfies 2πk (k is an integer) and the synchronization threshold is lower than the set value, a rhythm trigger event is output;

[0029] S6. Rhythm rendering, mapping rhythm trigger events to MIDI or OSC commands to generate audio signals;

[0030] S7. Interactive feedback: output audio to the user and receive adjustment signals, and return to step S3.

[0031] Optionally, the mapping strategy f described in step S3 uses a bidirectional gated recurrent network (Bi-GRU) to encode the input signal and output parameter increments to update the oscillator network.

[0032] Optionally, during the calculation process of step S4, the system solves the oscillator equations in parallel on a graphics processing unit (GPU) and uses the Compute Unified Device Architecture (CUDA) streaming batch processing technology to control the delay to less than or equal to one millisecond.

[0033] Optionally, the synchronization threshold in step S5 is obtained by Euclidean distance calculation, and the specific calculation formula is:

[0034]

[0035] And dynamically adjusted to between 0.05 and 0.15 according to the rhythm stability.

[0036] Optionally, the rhythm trigger event detected in step S6 is subdivided into time values ​​according to a rhythm template set by the user, and a Poisson-Gaussian jitter model is used to enhance the human voice performance for the micro time difference.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] 1. The present invention addresses the problem that traditional rhythm generation systems rely on preset templates or linear models and are difficult to cope with complex dynamic rhythms. By introducing nonlinear dynamic models and adaptive control strategies, the present invention realizes a highly personalized, dynamically changing, and real-time controllable music rhythm generation system, significantly improving the naturalness, complexity, and artistic expression of the rhythm output.

[0039] 2. The present invention adopts technologies such as multimodal input perception, reinforcement learning parameter adjustment, and GPU parallel computing, which not only ensures that the system delay is less than five milliseconds, but also supports large-scale, multi-part rhythm real-time rendering and interactive control, greatly enhancing the system's scalability and adaptability to actual application scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a block diagram of the overall structure of the system of the present invention.

[0041] Figure 2 It is a system workflow diagram of the present invention.

[0042] Figure 3 This is a diagram of the hardware and software implementation architecture of the present invention. DETAILED DESCRIPTION

[0043] The present invention provides a dynamic music rhythm system based on nonlinear dynamics, comprising: an input perception module for collecting at least one user input signal, wherein the user input signal includes a physical control signal, an emotional signal, a physiological signal, or an environmental beat signal; and a nonlinear oscillator network core module for receiving the signal output by the input perception module and performing calculations based on the following complex state equation, wherein the complex state equation is specifically:

[0044]

[0045] Among them, z i is the state of the i-th oscillator, α, β, κ ij, γ is an adjustable parameter, s(t) is an input signal coupling term; a parameter adaptive mapping module is used to update the adjustable parameter in real time based on the signal output by the input perception module; a rhythm event extraction and synthesis module is used to detect the phase synchronization nodes in the oscillator network to generate a rhythm event sequence, and accordingly drive the synthesis engine to output an audio signal; an interactive feedback module is used to provide visual or auditory feedback to the user and receive the user's secondary adjustment instructions to achieve closed-loop control; a storage and learning module is used to record user historical preferences and train the adaptive mapping strategy of the parameter adaptive mapping module to improve the personalization and predictability of rhythm generation; wherein the nonlinear oscillator network core module and the parameter adaptive mapping module are coupled via high-speed shared memory or in-process calling to ensure that the system end-to-end delay is less than or equal to five milliseconds, thereby achieving the real-time performance of the dynamic music rhythm system. The main purpose of the present invention is to solve the problem that most existing rhythm generation systems rely on preset beat templates or linear rhythm models. These methods have significant limitations in processing complex and dynamically changing music rhythms, making it difficult to achieve natural rhythm evolution, personalized expression and real-time regulation, and lack high adaptability to user interaction and feedback.

[0046] The terms "first," "second," "third," "fourth," and the like (if any) in the description and claims of the present invention and in the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a particular order or precedence. It should be understood that the terms used in this manner are interchangeable where appropriate so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units that are not explicitly listed or that are inherent to these processes, methods, products, or apparatus.

[0047] For ease of understanding, the specific process of an embodiment of the present invention is described below. An embodiment of a dynamic music rhythm system based on nonlinear dynamics provided by the present invention includes:

[0048] The input perception module, nonlinear oscillator network core module, parameter adaptive mapping module, rhythm event extraction and synthesis module, interactive feedback module, and storage and learning module are interconnected via a high-speed bus or shared memory within the same process to ensure end-to-end latency of less than five milliseconds, enabling real-time rhythm generation and control.

[0049] The input sensing module is used to collect at least one user input signal, which can include physical control signals (such as MIDI controllers or touch screen signals), emotional and physiological signals (such as EEG and heart rate), or environmental beat signals (such as microphone input). This module uses asynchronous sampling in a multi-threaded buffer and uses a normalization operator to uniformly map various heterogeneous signals to a standard amplitude range of [-1, 1].

[0050] The core module of the nonlinear oscillator network consists of multiple (N) nonlinear oscillator units, each of which is described by the following complex state equation:

[0051]

[0052] Among them, α controls the amplitude convergence, ω i is the natural frequency of the i-th oscillator, β is the amplitude limit parameter, κ ij is the coupling coefficient, γ is the coupling strength of the external driving force, and s(t) is the input driving force. This module uses graphics processing units (GPUs) for parallel computing to ensure real-time and high efficiency.

[0053] The parameter adaptive mapping module is based on the deep reinforcement learning strategy π and optimizes the parameters of the nonlinear oscillator network in real time according to the following reward function:

[0054] R=λ1C sync +λ2C var -λ3C lat

[0055] Among them C sync Indicates the degree of synchronization, C var Indicates complexity, C lat Denotes delay, λ1, λ2, and λ3 are weight coefficients. This module can preferably adopt a Bi-GRU-DDPG network with an attention mechanism, and continuously optimize the mapping strategy based on user feedback.

[0056] The rhythm event extraction and synthesis module performs multi-scale analysis of the oscillator phase trajectory based on the wavelet packet decomposition algorithm. When the synchronization threshold σ satisfies:

[0057]

[0058] When a phase zero crossing event occurs within a preset range (e.g., 0.05 to 0.15), a rhythm trigger event is output. The module then converts the trigger event into MIDI or OSC commands and renders a multi-part audio track through a multi-channel sample synthesizer, a physical modeling synthesizer, or a frequency modulation / phase modulation (FM / PM) synthesis chain.

[0059] The interactive feedback module provides a graphical user interface that visualizes information such as the oscillator phase plane, Lyapunov exponential curve, and Groove pattern in real time, allowing users to adjust and optimize system parameters. Furthermore, the module supports multimodal operation, including sliding, clicking, and knob-based human-computer interaction methods, ensuring a positive interactive experience.

[0060] The storage and learning module is used to persistently store user preference profiles P, policy network parameters W, rhythm event sequences E, and system logs L. This module supports the rapid recovery of user feature vectors based on P to shorten the system learning convergence time and continuously optimizes the adaptive mapping module through big data analysis.

[0061] Based on the above system description, in large-scale live music performances, this embodiment uses a MIDI drum pad as the physical input port of the input perception module to collect the drummer's striking force and beat speed in real time. The nonlinear oscillator network core module dynamically adjusts the coupling parameters and natural frequencies according to the input signal to generate a multi-level rhythm that matches the performer's emotions and stage atmosphere. The rhythm trigger events detected by the rhythm event extraction and synthesis module are mapped to multi-part electronic drum tracks and played in real time through the main sound reinforcement system, effectively enhancing the on-site interactivity and audience immersion. During the performance, the system can also display phase diagrams and rhythm stability information through the interactive feedback module for the drummer's reference and adjustment, thereby achieving personalized dynamic rhythm accompaniment.

[0062] Based on the above system description, in the scenarios of psychological healing and meditation music generation, this embodiment uses a dry electrode brain-computer interface as the input perception module to collect the user's alpha and theta wave EEG signals in real time. The parameter adaptive mapping module uses a bidirectional gated recurrent network (Bi-GRU) to analyze the changes in the input signal and adjust the parameters of the nonlinear oscillator network core module online to synchronize the generated rhythm with the user's current emotional state. The rhythm event extraction and synthesis module uses physical modeling of wooden fish, bells and other timbres to render soothing rhythm outputs, which are played through high-fidelity audio equipment to provide users with a progressive and personalized meditation experience. According to subjective feedback from users, the system can effectively enhance relaxation and concentration.

[0063] Based on the above system description, in this embodiment, in an interactive public art installation, the environmental microphone collects environmental audio signals such as crowd density and background noise in real time through the input perception module, and the system converts it into a driving force signal and inputs it into the nonlinear oscillator network core module. As the environment changes, the system adjusts the oscillator parameters in real time to generate a dynamically evolving rhythm sequence. The rhythm output by the rhythm event extraction and synthesis module is not only played through the audio, but also synchronously drives the LED light array to form a linked display of music and vision. The interactive feedback module supports the audience to participate in real-time control through movement, gestures or touch, making the device not only a passive display, but also an art experience platform co-created by the audience and the system.

[0064] The above describes the dynamic music rhythm system based on nonlinear dynamics in the embodiment of the present invention. The following describes the dynamic music rhythm generation method based on nonlinear dynamics in the embodiment of the present invention. Figure 2 , an embodiment of the method for generating dynamic music rhythm based on nonlinear dynamics in an embodiment of the present invention includes:

[0065] S1. Input acquisition, obtaining user input signals;

[0066] S2. Model initialization: establish N models with natural frequency ω according to the user input signal. i oscillator network.

[0067] S3. Adaptive parameter adjustment, calling mapping strategy f to update adjustable parameters in real time; wherein the mapping strategy f uses a bidirectional gated recurrent network (Bi-GRU) to encode the input signal and output parameter increments to update the oscillator network

[0068] S4. Dynamics are solved using a fourth-order Runge-Kutta method with numerical integration in steps of 2 milliseconds or less. The system solves the oscillator equations in parallel on the graphics processing unit (GPU) and uses Compute Unified Device Architecture (CUDA) streaming batching to minimize latency to 1 millisecond or less.

[0069] S5. Event detection: When the oscillator phase satisfies 2πk (k is an integer) and the synchronization threshold is lower than the set value, a rhythm trigger event is output. The synchronization threshold is obtained by Euclidean distance calculation, and the specific calculation formula is:

[0070]

[0071] and dynamically adjusted to between 0.05 and 0.15 according to the rhythm stability;

[0072] S6. Rhythm rendering: Mapping rhythm trigger events to MIDI or OSC commands to generate audio signals. The detected rhythm trigger events are subdivided into durations according to the user-defined rhythm template, and a Poisson-Gaussian jitter model is used to refine the micro-time differences and enhance the humanized performance.

[0073] S7. Interactive feedback: output audio to the user and receive adjustment signals, and return to step S3.

[0074] The above describes the dynamic music rhythm generation method based on nonlinear dynamics in the embodiment of the present invention. The hardware and software implementation of the dynamic music rhythm generation method based on nonlinear dynamics in the embodiment of the present invention are described below. Figure 3 ,include:

[0075] In this embodiment of the present invention, the system is preferably implemented on an x86_64 architecture-based central processing unit (CPU) and a CUDA-supported GPU platform. The GPU video memory is at least four gigabytes (GB), and the PCIe bandwidth between the main memory and the video memory is at least eight gigabytes per second (GB / s) to meet the requirements of large-scale parallel computing. In terms of software, it is recommended to use a Linux-based real-time kernel, the audio driver uses Jack or ASIO low-latency interfaces, the algorithm module is compiled as a shared object (.so) and scheduled by a mixed Python or C++ script. The deep learning component can use the PyTorch or TensorFlow2.x framework.

[0076] In summary, this invention utilizes a nonlinear oscillator network as the core of rhythm generation, significantly improving the system's dynamics, complexity, and personalization. Compared to existing rhythm generation methods based on templates or linear models, this invention can dynamically adjust rhythm output based on real-time input with a delay of less than or equal to five milliseconds. Combined with a deep reinforcement learning adaptive strategy, this method achieves personalized rhythm control for different users, offering greater real-time performance, interactivity, and artistic expression.

[0077] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments can still be modified, or some of the technical features thereof can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A dynamic music rhythm system of nonlinear dynamics, characterized in that: The nonlinear dynamics dynamic music rhythm system includes: An input sensing module, configured to collect at least one user input signal, wherein the user input signal includes a physical control signal, an emotional signal, a physiological signal, or an environmental beat signal; The nonlinear oscillator network core module is used to receive the signal output by the input sensing module and perform calculations based on the following complex state equation, which is specifically: Among them, z i is the state of the i-th oscillator, α, β, κ ij , γ is an adjustable parameter, s(t) is the input signal coupling term; A parameter adaptive mapping module, configured to update the adjustable parameters in real time based on the signal output by the input sensing module; a rhythmic event extraction and synthesis module, configured to detect phase synchronization nodes in the oscillator network to generate a rhythmic event sequence, and drive a synthesis engine to output an audio signal accordingly; Interactive feedback module, used to provide visual or auditory feedback to the user and receive secondary adjustment instructions from the user to achieve closed-loop control; A storage and learning module, configured to record user historical preferences and train the adaptive mapping strategy of the parameter adaptive mapping module to improve the personalization and predictability of rhythm generation; Among them, the nonlinear oscillator network core module and the parameter adaptive mapping module are coupled through high-speed shared memory or in-process calling to ensure that the system end-to-end delay is less than or equal to five milliseconds, thereby realizing the real-time performance of the dynamic music rhythm system.

2. The nonlinear dynamics dynamic music rhythm system according to claim 1, characterized in that: The input perception module includes a brain-computer interface subunit, which is used to map the EEG rhythm signal in the range of one to forty hertz into a driving force signal in the oscillator network core module.

3. The nonlinear dynamics music rhythm system according to claim 1, characterized in that: The parameter adaptive mapping module optimizes the reward function based on the Qianghu learning algorithm as follows: R=λ1C sync +λ2C var -λ3C lat Among them, C sync Indicates rhythm synchronization, C var Indicates rhythm complexity, C lat represents the system delay, and λ1, λ2, and λ3 are weight coefficients.

4. The nonlinear dynamics music rhythm system according to claim 1, characterized in that: The rhythm event extraction and synthesis module detects oscillator phase zero-crossing events based on a wavelet packet decomposition algorithm, and generates at least four-part rhythm tracks through a multi-channel sample synthesizer or a physical modeling synthesizer.

5. The nonlinear dynamics music rhythm system according to claim 1, characterized in that: The interactive feedback module provides a parameter visualization interface, which displays the phase plane diagram and Lyapunov exponent of the oscillator network in real time to indicate rhythm stability.

6. A method for generating dynamic music rhythm based on nonlinear dynamics, used in the dynamic music rhythm system based on nonlinear dynamics as claimed in any one of claims 1 to 5, characterized in that: include: S1. Input acquisition, obtaining user input signals; S2. Model initialization: establish N models with natural frequency ω according to the user input signal. i oscillator network. S3. Adaptive parameter adjustment: calling mapping strategy f to update adjustable parameters in real time; S4. Solve the dynamics using a fourth-order Runge-Kutta method for numerical integration with a step size of ≤ 2 milliseconds; S5. Event detection: when the oscillator phase satisfies 2πk (k is an integer) and the synchronization threshold is lower than the set value, a rhythm trigger event is output; S6. Rhythm rendering, mapping rhythm trigger events to MIDI or OSC commands to generate audio signals; S7. Interactive feedback: output audio to the user and receive adjustment signals, and return to step S3.

7. The method for generating dynamic music rhythm based on nonlinear dynamics according to claim 6, characterized in that: The mapping strategy f described in step S3 uses a bidirectional gated recurrent network (Bi-GRU) to encode the input signal and output parameter increments to update the oscillator network.

8. The method for generating dynamic music rhythm based on nonlinear dynamics according to claim 6, characterized in that: During the calculation process of step S4, the system solves the oscillator equations in parallel on the graphics processing unit (GPU) and uses the Compute Unified Device Architecture (CUDA) streaming batch processing technology to control the latency to less than or equal to one millisecond.

9. The method for generating dynamic music rhythm based on nonlinear dynamics according to claim 6, characterized in that: The synchronization threshold in step S5 is obtained by Euclidean distance calculation, and the specific calculation formula is: And dynamically adjusted to between 0.05 and 0.15 according to the rhythm stability.

10. The method for generating dynamic music rhythm based on nonlinear dynamics according to claim 6, characterized in that: The rhythm trigger event detected in step S6 is subdivided into time values ​​according to the rhythm template set by the user, and the Poisson-Gaussian jitter model is used to enhance the human voice performance for the micro time difference.

Images