Multi-voice intelligent arrangement and voice balance optimization method for folk music band rehearsal

Abstract

The application discloses a multi-voice intelligent arrangement and voice balance optimization method for folk music band rehearsal, and relates to the technical field of audio signal processing. The method acquires a score, identifies melody backbone, sub melody, harmony filling and rhythm support function roles, and constructs an instrument timbre image. According to a matching degree score, target penetration force weighted energy proportion of each voice part is obtained. A mixed sound field signal is collected by a multi-point pickup array above a band four corners and a conducting seat. A time-frequency fingerprint is decoupled with a prior constraint of a score active note to obtain a measured penetration force weighted energy proportion. The target and measured proportions are compared frame by frame. The voice part deviating from the target interval is prompted by a spectrum table light output intensity and is iteratively picked up until the standard is reached. The penetration force weighted proportion is used to replace sound energy as a balance physical quantity. In a short reverberation rehearsal hall without sound amplification equipment and full-time sound engineers, the multi-voice listening balance of the folk music band is continuously improved with rehearsal iteration.

Description

Technical Field

[0001] This invention relates to the field of audio signal processing technology, specifically to a method for intelligent multi-part arrangement and part balance optimization for rehearsals of folk music orchestras. Background Technology

[0002] The Chinese folk music orchestra is composed of four main categories of instruments: wind instruments, bowed string instruments, plucked string instruments, and percussion instruments. In rehearsal settings, it has long faced the practical difficulty of automatically arranging and balancing multiple parts in real time. This difficulty stems from three main sources: First, the four instrument categories have significant overlap in their registers. The timbre of the flute, xiao, sheng, and guan is far more penetrating than that of the plucked and bowed strings. If the scores are directly assigned according to Western orchestral conventions, the high-pitched flute can easily become overpowering, and the inner parts of the zhonghu (a type of Chinese bowed string instrument) can be drowned out. Second, the acoustic environment of rehearsal halls is far inferior to that of concert halls, with shorter reverberation times and a dominance of direct sound. The significant differences in the sound path from each part to the conductor's seat lead to an unbalanced listening experience at the conductor's position. Third, the balance parameters for the same piece need to be repeatedly readjusted in different rehearsal halls, with different numbers of participants, and with instruments of varying ages. Furthermore, the unique improvisational embellishments and free-rhythm rhythms of Chinese folk music make it difficult to adapt fixed balancing strategies. These difficulties are particularly pronounced in rehearsal settings without amplification equipment and a dedicated sound engineer.

[0003] Regarding the automatic arrangement of musical pieces, US Patent Application No. US11037538B2 discloses a method and system for automatic musical arrangement and instrument performance style transformation supported within an automatic music performance system. This method extracts music theory state descriptors, including roles, from the musical structure and automatically selects an instrument for each role to drive a deeply sampled virtual instrument to generate a digital performance. However, this solution focuses on the digital performance synthesis of virtual sampled instruments, and its role allocation only addresses the arrangement problem of "who plays," without addressing the measurement and adjustment of the perceived proportion of each voice part in the live sound field of a rehearsal hall. It cannot resolve the imbalance in the multi-part sound caused by the significant differences in the penetrating power of ethnic musical instruments.

[0004] Regarding energy balancing of multiple audio streams, Chinese patent application CN117316170A discloses a multi-channel audio mixing method, system, medium, and device based on energy assessment. This method uses speech activity detection to obtain the probability of human voices in each audio stream, multiplies this probability by the energy of each stream to obtain an energy assessment value, and then fades in or out audio streams with consecutive frames whose energy assessment values ​​are higher or lower than a preset threshold to participate in the mixing. However, this scheme uses the sound energy of each audio stream as the physical quantity for balancing and selection, and uses the probability of human voices rather than musical notation for weighting. It performs gain processing on the audio signal itself rather than providing adjustment instructions to the performer.

[0005] The common shortcoming of the aforementioned existing technologies is that they all use the sound energy or loudness of a particular voice as the physical quantity for voice balance. However, in rehearsal halls with short reverberation times and dominant direct sound, the auditory dominance of traditional Chinese instruments is determined by the penetrating power of the voice (i.e., the proportion of high-frequency energy and transient onset sharpness), rather than by the total energy. Therefore, using sound energy as the physical quantity for balance can lead to imbalances where one voice's energy is sufficient but its auditory impact is still overwhelming, or another voice's energy is sufficient but its auditory impact is still submerged. Furthermore, existing technologies cannot continuously correct this imbalance with each rehearsal iteration without amplification equipment or a dedicated sound engineer. How to achieve automatic arrangement and voice balance that continuously improves with each rehearsal iteration, addressing the multi-voice auditory imbalance caused by significant differences in the penetrating power of traditional Chinese instruments in short rehearsal halls without amplification equipment or a dedicated sound engineer, is a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0006] To address the technical bottleneck of existing technologies that use the sound energy of individual voices as a physical quantity for balancing in the rehearsal of folk music orchestras, which cannot cope with the auditory imbalance caused by the significant differences in the penetrating power of folk instruments, this invention provides a multi-part intelligent arrangement and voice balance optimization method for folk music orchestra rehearsals. By replacing sound energy with the weighted energy ratio of penetrating power as the target quantity and actual measurement for balance, and using the activity of the notes to be played in the score as a priori constraint, the mixed sound field is decoupled into the proportion of each voice part, and then the dynamic prompts are fed back to the performers via the micro-light of the music stand. This closed-loop arrangement, measurement, and tuning method achieves continuous improvement of the auditory imbalance of multi-part folk music orchestras through rehearsal iterations from the mechanism level of balancing physical quantities, without the need for sound reinforcement equipment and a dedicated sound engineer.

[0007] The technical solution of this invention is as follows: A method for intelligent multi-part arrangement and part balance optimization for folk music orchestra rehearsals includes the following steps: The full score uploaded from the conductor's platform is acquired; the functional roles of melodic backbone, countermelody, harmonic filler, and rhythmic support are identified segment by segment from the full score, and timbre profiles are constructed according to instrument register, brightness, phonation delay, and penetration. The score is then divided based on the matching degree between the functional roles and the timbre profiles to obtain the target penetration-weighted energy percentage for each voice part. The penetration-weighted energy percentage is the percentage obtained by weighting the frequency band energy of each voice part with a penetration weight; the mixed sound field signal is acquired through a multi-point pickup array distributed at the four corners of the orchestra and above the conductor's platform. Using the activity level of the notes to be played in the total score at the current moment as a priori constraint, the time-frequency fingerprint of the mixed sound field signal is decoupled to obtain the measured penetrating power weighted energy ratio of each voice part; the target penetrating power weighted energy ratio is compared with the measured penetrating power weighted energy ratio frame by frame, and the voice parts that deviate from the target range are given a dynamic prompt by the spectrometer's micro-light output; after the dynamic prompt, the multi-point pickup array re-acquires the mixed sound field signal and iterates the comparison until the measured penetrating power weighted energy ratio of each voice part falls into the target range.

[0008] The beneficial effects of this invention are as follows: First, by using the weighted proportion of penetration power instead of sound energy as the target and actual measurement for voice part balance, a listening balance consistent with the timbre characteristics of traditional Chinese musical instruments is achieved. The mechanism is as follows: in rehearsal halls with short reverberation times and dominant direct sound, the listener's perceived dominance of a voice part depends on the proportion of high-frequency harmonic energy reaching the listening point and the transient sharpness of that voice part, rather than its total energy. Therefore, the proportion obtained by weighting the frequency band energy of a voice part with a weight determined by penetration power is the balance physical quantity linearly related to the perceived dominance. Compared to using sound energy as the balance physical quantity, this scheme can still identify that the listening dominance of the qudi (a type of Chinese flute) is higher than that of the zhonghu (another type of Chinese erhu) even when their energies are equal, thus adjusting accordingly and avoiding the problem of unbalanced listening even when the energy meets the standard.

[0009] Secondly, by using the activity level of the notes to be played in the full score as a priori constraint to decouple the mixed sound field signal acquired by the multi-point pickup array through time-frequency fingerprinting, a reliable acquisition of the weighted energy proportion of the measured penetration force of each voice part is achieved under the condition that the direct sound of short reverberation is dominant. The mechanism is as follows: the early reflections in the short reverberation rehearsal hall are sparse, and the blind source separation based on reverberation statistics degenerates under this condition. Moreover, the high-frequency harmonic structures of various folk instruments such as wind instruments, bowed string instruments, and plucked string instruments highly overlap, and it is difficult to distinguish each voice part in the mixed sound field based solely on the time-frequency fingerprint of timbre. Introducing the activity level of the notes to be played in the full score at the current moment as a priori constraint is equivalent to telling whether each voice part should play at this moment and what note it should play during the decoupling process, thereby transforming the originally indistinguishable decoupling into a solvable problem under a priori constraint. Compared with the existing methods that separate based solely on timbre characteristics, this scheme can still stably output the proportion of each voice part in the overlapping sections of the four major types of instrument registers.

[0010] Third, by closing the three processes of editing, testing, and tuning—targeting the weighted proportion of penetration, measuring the a priori decoupling of the score, and adjusting the dynamic cues from the spectrometer—into an iterative loop, a synergistic effect exceeding the sum of the individual effects of each step is generated. The mechanism is as follows: the target quantity step defines what is being balanced; the decoupling measurement step provides the measured value of that value; the dynamic cues step drives the performer to adjust based on the difference between the target and the measured value; after adjustment, the sound is re-picked up and compared again. The loop is not established if any one of these three processes is missing. Individual target quantities cannot be verified, individual decoupling measurements cannot be utilized, and individual dynamic cues cannot be judged. Only when all three are closed can the measured weighted proportion of penetration energy for each voice part gradually converge to the target range with each rehearsal iteration, even in rehearsal scenarios without amplification equipment or a dedicated sound engineer. This achieves a system-level effect of self-balancing with iteration, which is not available in single-step techniques. Attached Figure Description

[0011] Figure 1 This is an overall flowchart of the method for intelligent multi-part arrangement and part balance optimization in folk music orchestra rehearsals according to the present invention; Figure 2 This is a flowchart of the process for constructing a timbre profile and determining the weighted energy percentage of target penetration in this invention; Figure 3 This is a flowchart of the decoupling of multi-point pickup array acquisition and spectral prior constraint time-frequency fingerprint in this invention; Figure 4 This is a flowchart of the pre-performance bias and frequency-varying penetration compensation based on the seat pitch in this invention; Figure 5 This is a flowchart of the frame-by-frame comparison and sub-sound spectrum stage low-light intensity prompt in this invention; Figure 6 This is a flowchart of the prior confidence dynamic modulation and impromptu paragraph processing in this invention; Figure 7 This is a flowchart of the rehearsal memory sedimentation, segmented reuse and penetrating power balanced fingerprint reverse correction in this invention. Detailed Implementation

[0012] The present invention will be further described below with reference to the accompanying drawings and embodiments. The following embodiments use the example of a folk music ensemble of approximately 60 people rehearsing a folk orchestral piece containing slow and fast movements in a rehearsal hall with a reverberation time of approximately 0.8 seconds as a background to specifically illustrate the steps of the method of the present invention. In this embodiment, the ensemble is divided into several sections, including gaohu, erhu, zhonghu, qudi, sheng, pipa, zhongruan, daruan, and percussion. The total number of sections is denoted as […]. Each voice part is numbered sequentially.

[0013] It should be noted that the overall process of the method of the present invention is as follows: Figure 1 As shown, Figure 1 The specific implementation process of each step is further as follows: Figures 2 to 7 As shown, steps S2 to S7 correspond sequentially. Figures 2 to 7 The diagrams are interconnected using the weighted energy ratio of the target penetration power for each sound part and the weighted energy ratio of the measured penetration power as the data interface.

[0014] like Figure 1As shown, in step S1, the full score uploaded from the conductor's console is obtained, and the functional roles of each segment of the full score are identified. The conductor uploads the full score of the rehearsed piece from the conductor's console via the music management terminal. The full score provides the pitch, duration, and start and end positions of each note in each voice part in digital music score format. After receiving the full score, the method divides it into several segments along the time axis according to musical phrases. The functional roles of each voice part in each segment are identified, and the functional role of each voice part in that segment is determined to be one of the following: melodic backbone, secondary melody, harmonic filler, or rhythmic support. Functional roles are identified based on a comprehensive assessment of the note density, interval leaps, pitch envelope correlation with other voices in the same section, and position within the vertical harmony of each voice part: notes carrying the main melody, with longer durations and pitch envelopes located above the vertical texture, are considered melodic backbones; notes related to the melodic backbone's pitch envelope and in a supporting role are considered secondary melodies; notes mostly falling on vertical chord components and filling the harmony with sustained or evenly divided rhythms are considered harmonic fillers; and notes supported by fixed rhythmic patterns with minimal pitch variation are considered rhythmic supporters. Functional roles are updated with each section; the same voice part may assume different functional roles in different sections. For example, the erhu voice part may act as the melodic backbone in a slow main section but become a harmonic filler in a fast accompaniment section. The identification results form a sequence of functional roles for each voice part, organized by section, for subsequent score division.

[0015] In the specific implementation of functional role recognition, a time window is set for each segment, with the window length being the duration of a musical phrase; in this embodiment, it is 2 to 4 measures. Within this time window, the note density (number of notes per unit time value), interval leap amplitude (average pitch difference between adjacent notes), pitch envelope (connection of the highest-pitched notes at each moment), and correlation coefficient between the pitch envelope of this voice part and the pitch envelopes of other voice parts in the same segment are statistically analyzed for each voice part. The judgment rules are as follows: those with pitch envelopes located above the vertical texture, longer note durations, and lower correlation coefficients with most voice parts are considered independent and are judged as the melody backbone; those with pitch envelopes having high correlation coefficients with the melody backbone and overall pitch lower than the melody backbone are judged as secondary melodies; those with notes mostly falling on the vertical chord constituents, appearing as sustained notes or evenly divided rhythms, and having weak melodic characteristics are judged as harmonic fillers; those with small pitch variations and fixed rhythmic patterns, appearing as strong beats or fixed rhythmic points, are judged as rhythmic support. When a voice part possesses multiple characteristics within the same segment, the one with the highest score among the four criteria mentioned above is taken as its functional role. This step is a purely score-level analysis and does not involve sound field acquisition. The output functional role sequence is used in subsequent steps to determine the score division and the weighted energy proportion of target penetration.

[0016] like Figure 2 As shown, in step S2, a timbre profile of each voice part is constructed, and the spectral analysis is performed by matching the functional role with the timbre profile, and the weighted energy ratio of the target penetration of each voice part is determined.

[0017] For each section of the orchestra, a timbre profile was constructed based on four dimensions: instrument register, brightness, sound delay, and penetration. Register was determined by the lower and upper limits of the instrument's commonly used range; sound delay was determined by the time from the start of the instrument's playing to the establishment of sound, with plucked and percussion instruments having shorter sound delays, while bowed and wind instruments had relatively longer delays. Brightness and penetration were quantitatively determined by the dry sound sample spectrum of the section.

[0018] Brightness Determine using the following formula: , in: For the first The brightness of a voice part is a scalar quantity with a range of values ​​of 1000. , dimensionless, is calculated by this formula, and characterizes the brightness of the timbre of this voice in the high-frequency region; is the voice part number, and is a scalar with a range of values. Dimensionless, determined by voice part number, used to identify each voice part; The total number of voices is denoted by , which is a scalar with a value range of positive integers not less than 2. It is dimensionless and determined by the orchestra's composition, representing the number of voices participating in the balance. Here, is the frequency, and is a scalar with a range of values. The unit is Hz, which is determined by the frequency point of the spectrum analysis and is used to integrate the band energy by frequency. It is a set of frequency points across the entire frequency band. It is a set of high-frequency bands and In this embodiment, Both are sets of frequency points not lower than 2000Hz, and are dimensionless. They are determined by the frequency band division of the spectrum analysis and are used for energy statistics of the full frequency band and high frequency band, respectively. For the first Voice part at frequency The frequency band energy at a given point is a scalar, taking values ​​that are real numbers not less than 0, and its unit is _____. The energy intensity of the sound part at that frequency is obtained by taking the square of the modulus of the short-time Fourier transform of the dry sound sample of the sound part and accumulating it according to the frequency points. The highest frequency for spectral analysis is denoted by scalar, which is a real number greater than 0, and its unit is Hz. In this embodiment, 8000Hz is used, determined by half the sampling rate, and is used to normalize the spectral centroid. ; The weighting coefficient for the centroid term in the spectral density is denoted as , where is a scalar and its value ranges from . , dimensionless, is taken as 0.5 in this embodiment, and is determined by the relative magnitude of the contribution of the spectral centroid and the high frequency ratio to the perceived brightness, and is used to balance the weights of the two items.

[0019] Penetration Determine using the following formula: , in: For the first The penetrating power of a sound part is a scalar quantity with a range of values ​​of 1000. , dimensionless, is calculated by this formula, and characterizes the auditory dominance of this voice part under the condition that the direct sound of short reverberation is dominant; is the weighting coefficient for the high-frequency component of penetration, and is a scalar with a value range of . , dimensionless, is taken as 0.6 in this embodiment, and is determined by the relative magnitude of the contribution of high frequency proportion and transient sharpness to the dominance of auditory perception, and is used to balance the two weights; For the first The transient onset sharpness of the sound part is a scalar quantity with a range of values ​​of 1000. , dimensionless, is calculated by this formula and characterizes the steepness of energy buildup when the voice begins to play; For the first The energy envelope of the dry acoustic sample is a scalar function, taking values ​​that are real numbers not less than 0, and its unit is 1. It is obtained by extracting the energy envelope of the single-note intro section of the voice part, and represents the rise of energy over time during the intro process; The time within the initial oscillation period is a scalar quantity, and its value range is the duration of the initial oscillation period, with units of _____. The time axis of the oscillation start-up segment is determined and used to calculate the time derivative with respect to the oscillation start-up envelope; The normalization factor is a scalar, taking values ​​of real numbers greater than 0, and its unit is . The maximum value of the time derivative of the vibration start-up envelope of the sound part on the sample set is used to normalize the transient sharpness. ; , , The definition is the same as the aforementioned brightness formula.

[0020] After obtaining the functional roles and timbre profiles of each voice part, the score is divided based on the matching degree between the functional roles and the timbre profiles. The principle is as follows: melodic backbone functional roles are preferentially assigned to voice parts with higher brightness and suitable registers; harmonic filling functional roles are assigned to voice parts with moderate penetration and strong vertical harmonic cohesion; and rhythmic support functional roles are assigned to voice parts with short vocal delays. Based on the score division results and the timbre profiles of each voice part, the penetration weight of each voice part is determined. The energy percentage weighted by the target penetration power. The penetration power weight is determined by the following formula: ,in: For the first The penetrating power weight of a vocal part is a scalar, and its value range is... , dimensionless, is calculated by this formula and represents the contribution of unit energy of this voice part to the degree of auditory dominance; Let be the voice part number to be summed, and be a scalar with a range of values ​​of . , dimensionless, used to traverse all voice parts; , The first , No. The penetrating power of a vocal part is defined in the same way as the aforementioned penetrating power formula; The definition is the same as before. This formula assigns a penetrating power weight that decreases in value from the highest to the lowest penetrating power of each voice part. Based on this, the weighted energy percentage of the target penetrating power for each voice part is determined by the following formula: ,in: For the first The voice part in the The target penetration weighted energy percentage of a frame is a scalar with a value range of [value range missing]. , dimensionless, is calculated by this formula, and represents the dominant auditory proportion that the voice part should occupy in the frame; is the frame number, a scalar, a positive integer not less than 1, dimensionless, determined by frame segmentation, and used for frame-by-frame comparison; For the first The voice part in the The frame is defined by the target energy given by the spectrum, which is a scalar, taking values ​​that are real numbers not less than 0, and is in units of . The energy of the voice part in this frame is calculated by combining the spectral results with the dynamic markings of the voice part. , The definition is the same as the aforementioned penetration weight formula; , The definition is the same as before. In the score, for at least two voices with the same functional role and whose penetration power is not lower than the penetration power threshold, these two voices are set to alternate in playing phrases, instead of playing simultaneously. For example, when the dizi and gaohu both bear the melody backbone in the same bright melody segment and their penetration power is higher than the penetration power threshold, they are set to alternate in playing phrases, instead of playing the same melody line simultaneously. The penetration power threshold is set to 0.7 in this embodiment. The basis for this is that when two voices with high penetration power play the same melody simultaneously, under the condition that the short reverberation direct sound is dominant, their penetration power is non-linearly superimposed and dominates the overall listening experience, resulting in a crowded listening experience; after changing to alternating in playing phrases, only one high-penetration voice is on the part at any given time, making it easier for the weighted energy ratio of the target penetration power of each voice to meet the target in the subsequent closed loop.

[0021] like Figure 3 As shown, in step S3, the mixed sound field signal is collected by a multi-point pickup array, and the time-frequency fingerprint is decoupled with the activity of the responding note as a priori constraint to obtain the weighted energy ratio of the measured penetration of each sound part.

[0022] A multi-point pickup array is set up at the four corners of the orchestra and above the conductor's seat. In this embodiment, one pickup is placed at each of the four corners of the orchestra, and one pickup is suspended above the conductor's seat, for a total of 5 points. When the orchestra plays together, the multi-point pickup array collects a mixed sound field signal. The mixed sound field signal is a time-domain waveform that is synchronously collected by each pickup point, and after short-time Fourier transform, it is represented as a complex vector at each time frequency point.

[0023] When decoupling the mixed sound field signal using time-frequency fingerprinting, the activity level of the responding notes in the total score at the current moment is used as a priori constraint. The activity level of the responding notes is given by the total score, characterizing whether each voice part should be played according to the score in the current frame and its pitch. In this embodiment, it is quantified as the priori proportion of the responding energy of each voice part in the current frame. Decoupling involves solving for the energy distribution of each voice part at each time-frequency point, minimizing the residual between the energy distribution of each voice part and the time-frequency fingerprint of each voice part's timbre and spatial steering vector after reconstruction of the mixed sound field signal, while simultaneously ensuring that the energy distribution aligns with the priori activity level of the responding notes. The energy distribution is solved using the following formula: ,in: For the first The voice part in the Frame rate The energy distribution obtained from the solution is a scalar, taking values ​​that are real numbers not less than 0, and its unit is . The value obtained by minimizing this formula represents the actual energy of the sound part at that time frequency point. For the first Frame rate The mixed sound field signal vector at the location is a complex vector with a dimension equal to the number of pickup points (5 dimensions). The values ​​of each component are complex numbers, and the unit is [unit missing]. It is obtained by the short-time Fourier transform of the signal collected by the multi-point pickup array, and represents the complex sound pressure of each pickup point at that time frequency. For the first Energy distribution in vocal parts, meaning the same as To optimize variables; For the first Voice part at frequency The timbre time-frequency fingerprint at each location is a vector with the same dimension as the number of pickup points, and the values ​​of each component range from [value range missing]. Dimensionless, determined by the normalized energy distribution of the dry sound sample of the voice part at each frequency point, characterizing the frequency domain features of the timbre of the voice part; For the first The spatial steering vector of the voice part is a complex vector with the same dimension as the number of pickup points. Each component takes the value of a complex number with a modulus of 1. It is dimensionless and is determined by the geometric relationship between the voice part's seat and each pickup point. It represents the phase and amplitude relationship of the sound wave of the voice part reaching each pickup point. This indicates that vectors are multiplied element by element; The L2 norm of a vector; For the first The voice part in the Frame rate The prior given by the liveness of the note to be played is a scalar, taking values ​​of real numbers not less than 0, and its unit is . It is obtained by converting the pitch of the voice part in the current frame of the score with the dynamic markings, and represents the energy that the voice part should have according to the score. For the first The voice part in the The constraint weights of the frame are scalars, and their values ​​range from 1 to 2. Dimensionless, determined by the prior confidence dynamic modulation described in step S6, characterizing the constraint strength of the prior decoupling of the active note of the responding note to the voice part; Here, is the regularization coefficient of the prior constraint term; is a scalar, taking values ​​of real numbers greater than 0, with units of . In this embodiment, the value is 0.2, which is determined by the dimensional balancing and relative importance of the reconstructed residual term and the prior constraint term, and is used to balance the two terms. , , , The definition is the same as before. The acquisition methods of the timbre time-frequency fingerprint and spatial steering vector of each voice part in the above formula are as follows. The timbre time-frequency fingerprint is obtained by collecting single-tone dry sound samples of each voice part instrument before rehearsal. After performing a short-time Fourier transform on the samples, they are normalized according to frequency points. The average energy distribution at each frequency point is taken as the timbre time-frequency fingerprint of that voice part at each frequency point, which is used to characterize the energy distribution differences of various instruments such as wind, string, plucked, and percussion instruments in the frequency domain. For example, the energy of the flute is concentrated in high-frequency harmonics, and the energy of the erhu is concentrated in mid-low frequencies. The spatial steering vector is determined by the geometric relationship between the voice part seat and each pickup point: the propagation distance of each pickup point relative to the voice part is obtained by the difference between the coordinates of the voice part seat and the coordinates of each pickup point. The phase difference of each pickup point relative to the reference pickup point is calculated according to the propagation distance and the speed of sound. This constitutes the spatial steering vector of the voice part to the multi-point pickup array, which is used to characterize the phase and amplitude relationship of the sound wave of the voice part reaching each pickup point. The frequency domain differences between voices are characterized by timbre time-frequency fingerprints, the spatial differences between voices are characterized by spatial steering vectors, and the response differences of voices on the time axis are characterized by the a priori activity of the responding notes. These three together constitute a triple constraint for decoupling the mixed sound field, so that even if the timbre time-frequency fingerprints of blowpipes and plucked instruments are similar, they can still be distinguished by spatial steering vectors and the a priori activity of the responding notes.

[0024] When solving the above equation, the energy distribution at each time and frequency point is solved point by point using alternating direction multiplier iteration. The energy distribution is driven by the reconstructed residual term to fit the measured mixed sound field, and the energy distribution is driven by the prior constraint term to fit the activity of the responding note. The two terms converge after being balanced by the regular coefficients, and the energy distribution of each voice part at each time and frequency point is obtained.

[0025] The energy distribution of each voice part obtained from the solution Accumulate by frequency band and weight by penetration power to obtain the measured penetration power weighted energy percentage for each sound part. Its form is consistent with the target penetration weighted energy ratio in step S2, the difference being that the energy used is the measured energy obtained from decoupling. Instead of the target energy, the mixed sound field is decoupled into the measured penetration power weighted energy percentage of each sound element for subsequent comparison.

[0026] like Figure 4 As shown, in step S4, the seat distribution is obtained, the pre-performance bias is set according to the sound path, and frequency-varying penetration compensation is performed on the measured penetration weighted energy ratio.

[0027] Obtain the seating distribution for each section, which is given by the seat coordinates of each section in the orchestra during rehearsals. Determine the acoustic distance from each section to the conductor's seat based on the seating distribution, that is, the straight-line distance from the center of each section's seat to the pickup point of the conductor's seat.

[0028] A pre-performance offset is set for each part based on the sound path to compensate for the propagation delay of the far-end parts reaching the conductor's seat, ensuring that the ensemble heard by the conductor is synchronized in time. The pre-performance offset is equal to the sound path of that part divided by the speed of sound in air. In this embodiment, the speed of sound in air at room temperature is taken as 343 m / s, meaning that for every 3.43 m increase in sound path, the pre-performance offset increases by approximately 10 ms. The far-end parts start playing earlier accordingly, ensuring that the sound waves of all parts reach the conductor's seat simultaneously.

[0029] A frequency-varying attenuation model is established based on the sound path to compensate for the high-frequency band attenuated by the sound path in the measured weighted energy ratio of penetration power. This compensation is necessary because: when sound waves propagate in air, the high-frequency components attenuate faster with propagation distance than the low-frequency components. Different sound paths in different sound parts lead to a systematic underestimation of the high-frequency penetration components of far-end high-penetration sound parts at the conductor's position. Without compensation, the weighted energy ratio of penetration power measured at the conductor's position will be distorted. The compensation under the frequency-varying attenuation model is performed using the following formula: , in: For the first Voice part at frequency The compensated equivalent frequency band energy of the performance position is a scalar, taking values ​​that are real numbers not less than 0, and its unit is . The value calculated by this formula represents the energy after the sound path attenuation is inverted back to the playing position; For the first Voice part at frequency The measured frequency band energy of the command post obtained by decoupling is a scalar, taking values ​​that are real numbers not less than 0, and the unit is . The energy distribution in step S3 is determined by frequency point, which characterizes the energy measured at the conductor's seat after sound path attenuation; For frequency The frequency-varying attenuation coefficient at a given point is a scalar, taking values ​​that are real numbers greater than 0, and its unit is 1 / 2 ppm. The energy attenuation rate per unit propagation distance of a sound wave at this frequency is obtained by calculation using this formula. For the first The path length of a voice part is a scalar quantity, taking values ​​that are real numbers greater than 0, and its unit is . Determined by the seating arrangement, it represents the straight-line distance from the seat of that section to the conductor's seat; Represents an exponential function with the natural constant as its base; Here are the constant coefficients of the frequency-varying attenuation model; are scalars, taking values ​​that are real numbers not less than 0, and their units are . In this embodiment, The low-frequency absorption under the temperature and humidity conditions of the rehearsal hall is determined, characterizing the fundamental attenuation that is independent of frequency. Here, represents the coefficient of the quadratic term in the frequency-varying attenuation model; is a scalar, taking values ​​of real numbers greater than 0, with units of . In this embodiment, This is determined by the law that air absorption increases with the square of frequency, characterizing the intensification of high-frequency attenuation with frequency. , The definition is the same as before. After compensation, it is... Alternative The weighted energy percentage of the measured penetration force equivalent to the performance position was recalculated and used for subsequent comparisons.

[0030] like Figure 5 As shown, in step S5, the target is compared with the measured weighted energy ratio of penetration power frame by frame, and the intensity of the light output of the spectrometer is indicated for the sound part that deviates from the target range.

[0031] The weighted energy percentage of the target penetration power for each voice part is compared frame by frame with the equivalent measured weighted energy percentage of the penetration power at the performance position. To avoid frequent prompts causing performers to make repeated adjustments, a target range is set for each voice part, centered on its target weighted energy percentage of penetration power, extending one level above and below it. The output intensity prompt for a voice part is only triggered when the measured weighted energy percentage of the penetration power for a certain voice part deviates from its target range for several consecutive frames. In this embodiment, the prompt is triggered when there is a continuous deviation of 20 frames, or approximately 200ms.

[0032] The intensity prompts are divided into three levels: intensity reduction prompt, intensity increase prompt, and number of performers supplement prompt. For voices whose measured weighted energy percentage of penetration is higher than the target range, an intensity reduction prompt is output; for voices whose weighted energy percentage is lower than the target range, if the deviation is within one level, an intensity increase prompt is output; if the deviation exceeds one level and the number of performers in that voice is insufficient to fill the intensity range, a number of performers supplement prompt is output. The intensity prompts are displayed on the conductor's console screen in sections according to voice, and are synchronized with the spectrometer lights of each voice: for voices requiring intensity reduction, the spectrometer light changes to one color and gradually dims; for voices requiring intensity increase, the spectrometer light changes to another color and gradually brightens; for voices requiring number of performers supplement, the spectrometer light flashes. The criteria for determining the three levels of intensity are as follows: For a given vocal part, if the measured weighted energy percentage of penetration is higher than the upper limit of the target range, the overall sound of that vocal part is relatively strong, and a reduction in intensity is indicated. If the measured weighted energy percentage of penetration is lower than the lower limit of the target range but within one level of deviation, the vocal part can still be compensated for by increasing intensity, and an increase in intensity is indicated. If the measured weighted energy percentage of penetration is lower than the lower limit of the target range but the deviation exceeds one level, it means that even if the performer of that vocal part tries their best to increase intensity, it will be difficult to raise the percentage to the target range. In this case, a supplementary performer suggestion is indicated, recommending that additional performers be added to that vocal part. In this embodiment, the magnitude of the first level is taken as 0.05 of the target weighted energy percentage of penetration.

[0033] The music stand's subtle lighting works as follows: each section's music stand is equipped with a micro-light module with adjustable color temperature and brightness. For sections requiring reduced intensity, the music stand's micro-light turns warm and gradually dims; for sections requiring increased intensity, the micro-light turns cool and gradually brightens; and for sections needing additional members, the micro-light flashes at approximately 1Hz. The brightness variation of the micro-light is directly proportional to the deviation of the section's measured proportion from the target range; the greater the deviation, the more pronounced the dimming or brightening, allowing performers to perceive the amount of adjustment based on the micro-light's brightness. Simultaneously, the conductor's screen displays the type of prompt and the deviation range for each section, allowing the conductor to oversee the overall balance of the orchestra. Using the music stand's micro-light to provide dynamic cues allows performers to adjust the dynamics without interrupting the performance or responding to verbal commands, avoiding the need for verbal calls to stop, which disrupts rehearsal continuity. After the intensity prompt, the multi-point pickup array re-acquires the mixed sound field signal and iterates the aforementioned comparison until the weighted energy ratio of the measured penetration of each sound part falls into the target range, thus forming a closed loop of editing, testing and tuning.

[0034] like Figure 6 As shown, in step S6, the prior mismatch is calculated and the constraint weights are dynamically adjusted. For improvisational passages, only recording is performed without intervention. For the improvisational embellishments and free-rhythm rhythms unique to folk music, the prior activity of the notes given in the score may deviate from the actual performance in these passages. If decoupling is still performed with fixed-intensity constraints, the energy distribution obtained from decoupling will be biased by unreliable priors. Therefore, the prior mismatch between the prior activity of the notes and the energy distribution obtained from decoupling is calculated, and the constraint weights are dynamically adjusted accordingly. The prior mismatch is calculated using the following formula: ,in: For the first The voice part in the The a priori mismatch degree of the frame is a scalar, a real number not less than 0, dimensionless, calculated by this formula, and characterizes the degree of deviation between the actual performance of the voice part and the a priori performance according to the score. For the first The voice part in the The cumulative energy distribution obtained from frame decoupling across the entire frequency band is a scalar, taking values ​​that are real numbers not less than 0, and its unit is . The energy distribution in step S3 is accumulated according to the frequency band, which characterizes the measured energy of the sound part. For the first The voice part in the The prior energy of a frame is a scalar, taking values ​​that are real numbers not less than 0, and its unit is . It is calculated from the activity level of the note to be played, and represents the energy that the voice part should have according to the score. To prevent small positive numbers with a denominator of zero from being counted, it is a scalar, taking values ​​of the smallest real numbers greater than 0, with units of 1. In this embodiment, The value is determined by numerical stability requirements and is used to avoid division by zero. , The definition is the same as before. The constraint weights are dynamically modulated according to the following formula: ,in: For the first The voice part in the The constraint weights of the frames, as described in step S3, are calculated using this formula. The upper limit of the constraint weight is a scalar, which takes the value of a real number greater than 0. It is dimensionless. In this embodiment, it is taken as 1.0. It is determined by the maximum constraint strength when the prior is completely reliable, and it represents the maximum weight of the prior constraint. is the mismatch attenuation coefficient, a scalar, a real number greater than 0, dimensionless, and in this embodiment is 2.0. It is determined by the attenuation sensitivity of the constraint weight to the prior mismatch degree. The larger the value, the faster the constraint weight decreases with mismatch. The meaning is the same as before; The definition is the same as the aforementioned prior mismatch formula; , The definition is the same as before. This formula makes the constraint weight smaller when the prior mismatch is greater, thereby automatically reducing the prior constraint on decoupling of the active notes in the improvisation section and avoiding unreliable prior biased decoupling.

[0035] Furthermore, for the improvisational and free-rhythm sections marked by the conductor on the score, the method only records the weighted energy percentage of the measured penetration power of each voice part, without performing comparisons or dynamic cues. This respects the traditional improvisational style of folk music in these sections and avoids automatic balancing interference. The conductor selects these sections using section markers on the music management terminal when uploading the score. The method determines the start and end frames of these sections on the timeline based on these markers. When rehearsals reach a marked section, the method pauses the comparisons and dynamic cues for each voice part in that section, and the music stand's dim light remains off, allowing performers to freely improvise embellishments or adjust their rhythm according to the free-rhythm without being constrained by automatic balancing. Simultaneously, the method continuously decouples and records the weighted energy percentage of the measured penetration power of each voice part in the background, forming a measured percentage curve for that section. The recorded measured proportions are used for two purposes: firstly, for memory retention in step S7, to statistically analyze the group's actual balance habits during improvisational sections after multiple rehearsals; and secondly, when a marked section ends and the group returns to a regular section, the method uses the balance state of the most recent regular section before the section switch as the starting point for comparison, avoiding jumps during the return due to the pause in the improvisational section. Thus, the method preserves the freedom of folk music improvisation while maintaining the continuity of balance before and after the improvisational section.

[0036] like Figure 7 As shown, in step S7, the rehearsal memory and penetrating power balance fingerprint are settled, and segmented reuse and reverse correction are performed.

[0037] The orchestra's section balance memory is created by storing the score, the target penetration weighted energy percentage for each voice part, the measured penetration weighted energy percentage for each voice part, and the conductor's on-site adjustments from each rehearsal. This section balance memory is stored separately by movement and section to address the section-specific differences in tempos, where bowed strings need to be emphasized in slow movements, and plucked and percussion strings in fast movements. In slow sections, the target penetration weighted energy percentage for bowed strings is higher than that for plucked and percussion strings; in fast sections, the target penetration weighted energy percentage for plucked and percussion strings is higher than that for slow sections. During the next rehearsal of the same piece, the target penetration weighted energy percentage is loaded using this section balance memory as the initial value. Only voice parts that deviate from the measured value in the section balance memory are compared and given dynamic prompts, thus avoiding the need to start from zero each time.

[0038] Furthermore, the target penetration weighted energy ratio from multiple rehearsals, the measured penetration weighted energy ratio equivalent to the performance position, and the pre-performance bias and frequency-varying attenuation model are precipitated as the penetration balance fingerprint for this orchestra and its main hall. When changing rehearsal halls, the initial value of the target penetration weighted energy ratio is obtained by migrating the penetration balance fingerprint to the new hall's acoustic characteristics. The migration is performed using the following formula: ,in: For the first The voice part in the The initial value of the target penetration weighted energy ratio in the new hall is a scalar, and its value range is... , dimensionless, is calculated from this formula, and represents the proportion of the target used as the initial value after migration; To balance the penetration power stored in the fingerprint The voice part in the The target penetration weighted energy percentage of a frame is a scalar with a value range of [value range missing]. Dimensionless, read from the balance memory of the voice part, representing the target proportion of the original hall; , Let be the acoustic characteristic vectors of the rehearsal halls in the new hall and the original hall, respectively. is a vector, and in this embodiment, it is taken as a two-dimensional vector, with components representing reverberation time and direct-to-mixing ratio, respectively. The unit of reverberation time is . The direct mixing ratio is dimensionless and is obtained from acoustic measurements in each hall, characterizing the acoustic environment of the rehearsal hall; Here, is the migration coefficient vector, and is a vector with the same dimensions as the acoustic feature vector. The components take values ​​in the real number range, and the units of the components are respectively... Dimensionless, determined by historical multi-hall memory regression, characterizes the sensitivity of target proportion to differences in acoustic features; Indicates vector transpose; , The definition is the same as before. When the measured penetration weighted energy ratio of a certain voice part still does not fall into the target range after multiple iterations, it is determined that the voice part cannot achieve balance by force adjustment alone. The functional role of the voice part in the spectrum is corrected based on the penetration balance fingerprint, or the seat of the voice part in the seat distribution is corrected. This reverse correction indication is determined by the following formula: ,in: For the first The reverse correction indicator for a voice part is a scalar with a value range of [value range missing]. , dimensionless, is calculated by this formula, and when it is 1, it indicates that the spectral function role or seat of the voice part needs to be reversed; This is an indicator function. It takes the value 1 if the condition inside the parentheses is true and 0 otherwise. It is dimensionless and is used to convert consecutive non-compliance events into binary indicators. Let be the iteration index, and be a scalar with a range of values. , dimensionless, used to iterate through the most recent few iterations; The threshold for the number of consecutive iterations is a scalar, a positive integer not less than 2, and dimensionless. In this embodiment, it is set to 5, which is determined by the allowable number of adjustment attempts and represents the length of the observation window before triggering the reverse correction. For the first The voice part in the The weighted energy percentage of the equivalent measured penetration force at the performance position in the next iteration is a scalar with a value range of [value missing]. , dimensionless, obtained from step S4, characterizing the measured proportion of this iteration; For the first The lower limit of the target range for the voice part is a scalar, with a value range of [value missing]. Dimensionless, determined by the weighted energy ratio of the target penetration power, representing the lower bound of the target achievement; The definition is the same as before. This reverse correction allows for the adjustment of the functional role or position of voices that have been unable to achieve a balance through dynamics over a long period of time, thereby realizing a closed-loop correction of the arrangement based on rehearsal feedback.

[0039] Step S8: Folk music orchestration mapping and adaptive sparsification of percussion rhythm density. In the score, functional roles are further mapped to voice parts based on the characteristics of folk music orchestration. The harmonic filling role is assigned to the sheng, utilizing its natural harmonic characteristics to handle vertical harmonic bonding; the melodic backbone and secondary melody roles are assigned among the brighter voices such as the gaohu, qudi, and erhu according to the aforementioned alternating pronunciation principle; the mid-to-low-range voices such as the zhongruan, daruan, and zhonghu connect to the harmonic voices, avoiding hollowness in the mid-to-low range. The rhythmic support role is assigned to the percussion voices, and the pronunciation of the percussion voices is sparsified according to the rhythmic skeleton density of the score. The lower the rhythmic skeleton density, the higher the degree of sparsification. That is, in slow passages with sparse texture, the percussion voices are made to sound sparsely, while in fast passages with dense texture, the percussion voices are made to sound densely, thus avoiding the high-pitched crowding and percussion noise commonly seen in folk music orchestras.

[0040] The register dimension in timbre profiling is also used in orchestration mapping: when the same functional role needs to be assigned among multiple candidate voices, the voice with a register that matches the pitch range of the functional role in the score is prioritized. For example, if the melody backbone is in the high-middle register, it is prioritized for assignment to the gaohu or qudi; if it is in the low-middle register, it is prioritized for assignment to the erhu or zhonghu. This avoids forcing a voice with an incompatible register to undertake a functional role, resulting in an abrupt timbre. The vocal delay dimension is used to align the start of multiple voices for the same functional role: if a wind instrument with a long vocal delay and a plucked instrument with a short vocal delay share the same rhythmic point, the method is to reserve an advance for the wind instrument based on the difference in their vocal delays, so that the two sounds are established simultaneously, avoiding rhythmic misalignment caused by the difference in vocal delays. Through the above-mentioned functional roles to the folk music instrumentation mapping, alternating pronunciation according to musical phrases, adaptive sparsification of rhythm density of percussion parts, and adaptation of register and vocal delay, the target penetration weighted energy ratio obtained from the score not only matches the timbre characteristics of folk instruments, but also avoids high-pitched crowding and mid-low-pitched hollowness, providing a reasonable target for subsequent part balance.

[0041] Taking a slow movement rehearsal in the rehearsal hall in this embodiment as an example, the coordination of each step is illustrated. After uploading the score, the method identifies that in this slow movement segment, the erhu and gaohu bear the melodic backbone, the sheng and zhongruan bear the harmonic filling, and the daruan bears the rhythmic support. After constructing the timbre profile, the penetration power of the dizi is approximately 0.82, the gaohu is approximately 0.76, and the zhonghu is approximately 0.41. According to the penetration power weight and the score, the weighted energy proportion of the target penetration power for this segment is approximately 0.55 for the melodic backbone, approximately 0.30 for the harmonic filling, and approximately 0.15 for the rhythmic support. During the ensemble, a multi-point pickup array collects mixed sound field signals. Using the activity of the responding notes as a priori constraint for decoupling, the measured weighted energy proportion of the zhonghu's penetration power is approximately 0.08, which is lower than its target range lower limit of 0.12. Furthermore, the zhonghu's sound path in this segment is approximately 7.0m, and its high-frequency components are underestimated at the conductor's position. After frequency-varying penetration power compensation, the equivalent measured proportion at the performance position is approximately 0.10, which is still lower than the lower limit. Based on this, the method outputs a volume increase prompt on the zhonghu (a two-stringed bowed instrument) music stand. After the zhonghu's volume is increased and it is re-picked, the measured proportion rises to 0.13, falling into the target range. The qudi (a two-stringed flute), because it shares the same melodic backbone as the gaohu (a two-stringed bowed instrument) and both have a penetration power exceeding the penetration power threshold of 0.7, has already been set to alternate sounding according to musical phrases in the score, thus avoiding overlapping and crowding. Entering the allegro movement of the same piece, the method identifies from the score that the pipa (a four-stringed lute) and liuqin (a five-stringed plucked instrument) bear the melodic backbone, the percussion parts provide rhythmic support, and the sheng (a three-stringed sheng) provides harmonic filling in this allegro segment. Based on the segmented target penetration power weighted energy proportion, the proportion of plucked and percussion parts in the allegro section is increased, with plucked parts at approximately 0.40, percussion parts at approximately 0.25, harmonic filling at approximately 0.20, and bowed string secondary melodies outside the melodic backbone at approximately 0.15. This creates a segmental difference compared to the distribution of bowed string emphasis in the slow movement. During ensemble playing, the method revealed that the measured weighted energy percentage of the percussion part reached 0.34, exceeding the upper limit of its target range. Furthermore, the allegro had a dense texture and high rhythmic skeleton density. The method involved, on the one hand, reducing the volume of the percussion part on the music stand; on the other hand, maintaining a low degree of sparsity in the percussion part's sound based on the rhythmic skeleton density to avoid excessive sparsity disrupting the allegro's dynamics. After reducing the volume of the percussion part and re-picking it, its measured percentage dropped to 0.26, falling within the target range. Once the allegro section reached equilibrium, its score, target percentage, measured percentage curves, and conductor adjustment records were stored separately in the orchestra's section balance memory, according to movement and section, coexisting with the scheme for the slow section.

[0042] Once the slow section reaches equilibrium, its parting, target proportion, measured proportion curve, and conductor's adjustment record are stored in the orchestra's part balance memory. The next time the same piece is rehearsed in the same hall, this memory is used as the initial value, with only minor adjustments made to the deviating parts, thus avoiding starting from zero. If the zhonghu still fails to reach the target after five consecutive iterations, its position is adjusted from an inner harmonic part to closer to the conductor's seat based on the penetration balance fingerprint, or part of its harmonic filling is transferred to the sheng, achieving a reverse correction of the arrangement based on rehearsal feedback. When the orchestra moves to another rehearsal hall with a reverberation time of approximately 1.2 seconds to rehearse the same piece, the method uses the orchestra's own hall's penetration balance fingerprint as a benchmark. Based on the difference between the new hall's reverberation time and direct mixing ratio relative to the original hall, the initial value of the target penetration weighted energy proportion for each part is obtained, allowing the first rehearsal in the new hall to start from an initial value close to equilibrium, rather than starting from zero.

[0043] The above-described embodiments are merely preferred embodiments of the present invention, and the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or changes made by those skilled in the art within the scope of the technology disclosed in the present invention without departing from the technical solution of the present invention shall be covered within the scope of protection of the present invention.

Claims

1. A method for intelligent multi-part arrangement and part balance optimization for folk music orchestra rehearsals, characterized in that, include: Obtain the full score uploaded from the command post; The functional roles of melody backbone, sub-melody, harmonic filling and rhythmic support are identified segment by segment of the full score. A timbre profile is constructed according to the instrument register, brightness, sound delay and penetration. The score is divided according to the matching degree between the functional roles and the timbre profile to obtain the target penetration weighted energy ratio of each voice part. The penetration weighted energy ratio is the ratio obtained by weighting the frequency band energy of the voice part with the penetration weight. The mixed sound field signal is collected by a multi-point pickup array distributed at the four corners of the orchestra and above the conductor's seat. The time-frequency fingerprint decoupling of the mixed sound field signal is performed using the activity of the notes to be played in the score at the current moment as a priori constraint to obtain the measured penetrating power weighted energy ratio of each part. The weighted energy ratio of the target penetration is compared with the measured weighted energy ratio of the penetration on a frame-by-frame basis, and the intensity of the sound part that deviates from the target range is indicated by the micro-light output of its spectrometer. After the force indication, the multi-point pickup array re-acquires the mixed sound field signal and iterates the comparison until the weighted energy ratio of the measured penetration of each sound part falls into the target range.

2. The method for intelligent multi-part arrangement and part balance optimization for folk music orchestra rehearsals according to claim 1, characterized in that, In the timbre profile, brightness is the ratio of the centroid of the sound spectrum to the energy of the high-frequency band, and penetration is the weighted value of the energy ratio of the high-frequency band to the transient onset sharpness of the sound. The penetration weight is determined by decreasing the penetration of each sound from the highest to the lowest. For at least two sound parts with the same functional role and penetration that is not lower than the penetration threshold, the at least two sound parts are set to alternate in musical phrases in the score to replace simultaneous overlapping.

3. The method for intelligent multi-part arrangement and part balance optimization for folk music orchestra rehearsals according to claim 2, characterized in that, It also includes obtaining the seating distribution of each section and determining the acoustic path from each section to the conductor's seat based on the seating distribution; setting a pre-performance offset for each section based on the acoustic path to align the arrival time of each section at the conductor's seat; and establishing a frequency-varying attenuation model based on the acoustic path to compensate for the portion of the high-frequency band attenuated by the acoustic path in the measured penetrating power weighted energy ratio, thereby obtaining the equivalent measured penetrating power weighted energy ratio of the performance position.

4. The method for intelligent multi-part arrangement and part balance optimization for folk music orchestra rehearsals according to claim 3, characterized in that, The time-frequency fingerprint decoupling includes: using the activity of the responding note, the time-frequency fingerprint of the timbre of each voice part, and the spatial steering vector determined by the seat distribution as triple constraints, solving the energy distribution of each voice part at each time-frequency point of the mixed sound field signal, so that the reconstruction residual of the energy distribution of each voice part is minimized under the triple constraints, and obtaining the measured penetration weighted energy ratio from the energy distribution.

5. The method for intelligent multi-part arrangement and part balance optimization for folk music orchestra rehearsals according to claim 4, characterized in that, Also includes: Calculate the a priori mismatch between the activity level of the responding note and the energy distribution; dynamically modulate the constraint weights corresponding to the activity level of the responding note in the triple constraints according to the a priori mismatch, where the larger the a priori mismatch, the smaller the constraint weights; and for the improvisational embellishments and free-rhythm sections marked by the conductor on the score, only record the weighted energy percentage of the measured penetration of each voice part, without performing the comparison and the output of the dynamic prompts.

6. The method for intelligent multi-part arrangement and part balance optimization for folk music orchestra rehearsals according to claim 5, characterized in that, Also includes: The weighted energy ratio of the target penetration force from multiple rehearsals, the measured weighted energy ratio of the penetration force equivalent to the performance position, the pre-performance bias, and the frequency-varying attenuation model are precipitated as the penetration force balance fingerprint of this group and this hall. If the measured penetration weighted energy ratio of a certain voice part still does not fall into the target range after multiple iterations, the functional role of that voice part in the spectrum is corrected based on the penetration balance fingerprint, or the seat of that voice part in the seat distribution is corrected.

7. The method for intelligent multi-part arrangement and part balance optimization for folk music orchestra rehearsals according to claim 1, characterized in that, Also includes: The score of each rehearsal, the weighted energy ratio of the target penetration power of each voice part, the weighted energy ratio of the measured penetration power of each voice part, and the on-site adjustment record of the conductor are stored as the voice part balance memory of the group. When rehearsing the same piece for the next time, the weighted energy ratio of the target penetration power is loaded with the voice part balance memory as the initial value, and the comparison and dynamic prompt output are only performed on the voice parts that deviate from the voice part balance memory.

8. The method for intelligent multi-part arrangement and part balance optimization for folk music orchestra rehearsals according to claim 7, characterized in that, The vocal balance memory is stored separately according to movement and section. In the weighted energy ratio of target penetration power in the slow movement section, the proportion of bowed string parts is higher than that of plucked and percussion parts. In the weighted energy ratio of target penetration power in the fast movement section, the proportion of plucked and percussion parts is higher than that in the slow movement section.

9. The method for intelligent multi-part arrangement and part balance optimization for folk music orchestra rehearsals according to claim 1, characterized in that, The intensity prompts include intensity reduction prompts, intensity increase prompts, and number of personnel prompts. The intensity prompts are displayed on the control console screen according to the sound section and are synchronously presented with the dim light of each sound section's spectrum. Specifically, for sound sections whose measured penetrating power weighted energy ratio is higher than the target range, the intensity reduction prompt is output; for sound sections whose measured penetrating power weighted energy ratio is lower than the target range, the intensity increase prompt or number of personnel prompt is output.

10. The method for intelligent multi-part arrangement and part balance optimization for folk music orchestra rehearsals according to claim 1, characterized in that, In the spectrum, the harmonic filling function is assigned to the sheng to undertake the longitudinal harmonic bonding, the rhythmic support function is assigned to the percussion part, and the pronunciation of the percussion part is sparsified according to the rhythmic skeleton density of the spectrum, and the lower the rhythmic skeleton density, the higher the degree of sparsification.

Images