Automatic frequency modulation method of sound reed

Abstract

The invention discloses an automatic frequency modulation method for a sound reed, which comprises the following steps of: when the vibration frequency of the sound reed is unqualified, forming a two-dimensional vector by adopting the number of the sound reed and the frequency deviation value of the sound reed by acquiring the difference between the current vibration frequency and the reference frequency of the sound reed, namely the frequency deviation value; the trained deep learning model is input for prediction, five grinding parameters are obtained, grinding equipment is adopted to grind the sound reed based on the five obtained grinding parameters, and therefore automatic frequency modulation of the sound reed is achieved; the frequency modulation device has the advantages that efficient, stable and high-precision frequency modulation can be realized, the requirement of the sound reed frequency modulation on the professional degree of personnel is greatly reduced, and the frequency modulation labor cost is reduced.

Description

Technical Field

[0001] The invention relates to a frequency modulation method, in particular to an automatic frequency modulation method for a reed piece. Background Art

[0002] Reed vibrating instruments are a type of musical instrument that uses the vibration of reeds to produce sound. Reeds can be of various shapes and materials, but their common feature is that when subjected to external forces, they can vibrate and excite the surrounding air to produce sound. Such instruments as organs, accordions, harmonicas, and octaves, etc., produce sound by vibrating a free reed fixed at one end. For a good musical instrument, each note needs to match the standard pitch. The more accurate, the more pleasant the sound. Therefore, in the production process of musical instruments, the precise frequency tuning of reeds is a very core technical process. For example, in the harmonica industry, at present, both domestic and foreign manufacturers still mainly perform frequency tuning of reeds manually. When tuning a reed, the reed is first deformed under the action of air to start vibrating and producing sound. The microphone collects the sound information generated by the reed and converts it into an electrical signal and outputs it to the oscilloscope. The oscilloscope converts the electrical signal into a visual graph. The tuner compares the graph displayed by the oscilloscope with the standard graph of the reed to determine whether they are consistent. If they are inconsistent, the reed is ground by grinding tools such as milling cutters or files to achieve frequency tuning. Since the recognition ability of tuners is different and unstable, they will also get tired, so the efficiency is low. At the same time, it is affected by human subjectivity, and the frequency tuning accuracy is unstable, which affects the quality and performance of the final product. In addition, since the accuracy of the reed frequency tuning mainly depends on the professionalism of the tuner, the professionalism of the tuner is required to be high. It takes more than a year for an experienced tuner to be trained. Therefore, the labor cost of frequency tuning is high. Once the staff is lost, it will have a great impact on the company. Summary of the invention

[0003] The technical problem to be solved by the present invention is to provide an automatic frequency tuning method for a reed piece which can achieve efficient, stable and high-precision frequency tuning and can reduce the labor cost of frequency tuning.

[0004] The technical solution adopted by the present invention to solve the above technical problems is: an automatic frequency tuning method for a reed piece, firstly setting a number for each reed piece of a reed vibrating musical instrument, each reed piece having a different number, obtaining the current vibration frequency of each reed piece respectively, comparing the current vibration frequency of each reed piece with its reference frequency, if the two are consistent, it indicates that the frequency of the reed piece is qualified, if they are inconsistent, it indicates that the frequency of the reed piece is unqualified, at this time, the difference between the current vibration frequency of the reed piece and its reference frequency is calculated, that is, the frequency deviation value of the reed piece, and a two-dimensional vector is formed by using the number of the reed piece and the frequency deviation value of the reed piece, which is input into a trained deep learning model for prediction, and a 5-dimensional feature data output is obtained, wherein the 5-dimensional feature data respectively correspond to five grinding parameters of a grinding device: milling cutter rotation angular velocity, milling cutter movement speed, milling cutter grinding length, milling cutter grinding depth and milling cutter grinding path, and then after setting the five grinding parameters of the grinding device according to the 5-dimensional feature data, the grinding device is used to grind the reed piece so that its vibration frequency is equal to its reference frequency.

[0005] Compared with the prior art, the advantage of the present invention is that when the vibration frequency of the sound reed piece is unqualified, by obtaining the difference between the current vibration frequency of the sound reed piece and its reference frequency, that is, the frequency deviation value, the number of the sound reed piece and the frequency deviation value of the sound reed piece are used to form a two-dimensional vector, and the trained deep learning model is input for prediction to obtain five grinding parameters. Based on the obtained five grinding parameters, the sound reed piece is ground by a grinding device, thereby realizing automatic frequency modulation of the sound reed piece, wherein the vibration frequency of the sound reed piece can be realized by the current mature signal acquisition technology, so that the obtained frequency deviation value eliminates human subjective intervention and has high precision, and the deep learning model has a feature learning function, which can accurately predict the five grinding parameters after training, so that after the grinding device grinds the sound reed piece based on the predicted five grinding parameters, the vibration frequency of the sound reed piece can be equal to its reference frequency. Therefore, the present invention can realize efficient, stable and high-precision frequency modulation, and greatly reduce the requirements for the professionalism of personnel for the frequency modulation of the sound reed piece, and reduce the labor cost of frequency modulation.

[0006] Furthermore, the specific process of obtaining the vibration frequency of each reed piece is: first drive the reed piece to vibrate and emit sound, then collect the sound signal, and then use Fourier transform to convert the sound signal into a frequency signal, and the frequency signal is recorded as the current vibration frequency of the reed piece.

[0007] Furthermore, the trained deep learning model is obtained by training the deep learning model using a training set, wherein the training set is constructed by the following steps:

[0008] S1. Each reed of the reed vibrating instrument is processed as follows:

[0009] S1.1. Set five grinding parameters of the grinding equipment: milling cutter rotation angular velocity, milling cutter moving speed, milling cutter grinding length, milling cutter grinding depth and milling cutter grinding path;

[0010] S1.2, obtaining the current vibration frequency of the reed piece, that is, the vibration frequency before this grinding, and after grinding the reed piece with a grinding device, obtaining the current vibration frequency of the reed piece, that is, the vibration frequency after this grinding, and calculating the difference between the vibration frequencies of the reed piece before and after this grinding, that is, the frequency deviation value corresponding to a set of grinding parameters consisting of the current five grinding parameters;

[0011] S1.3, four of the five grinding parameters of the grinding equipment are used as holding parameters, and the other one is used as an adjustment parameter, the four holding parameters remain unchanged, and the data of the adjustment parameter is changed to obtain a set of grinding parameters consisting of five grinding parameters again;

[0012] S1.4, using the same method as step S1.2, obtain the frequency deviation value corresponding to a set of grinding parameters consisting of the current five grinding parameters;

[0013] S1.5, repeating steps S1.3 and S1.4 until the frequency deviation values ​​corresponding to m groups of grinding parameters are obtained, where m is an integer greater than or equal to 50, wherein when repeating step S1.3, the five grinding parameters of the grinding equipment are changed as adjustment parameters respectively;

[0014] S2. Each training sample is represented as [X, Y], where X is a two-dimensional vector as input data, X = [x1, x2], x1 is the number of the reed piece, x2 is the frequency deviation value, and Y is a five-dimensional vector as output data, i.e., label, Y = [a1, a2, a3, a4, a5], a1 represents the angular velocity of the milling cutter, a2 represents the moving speed of the milling cutter, a3 represents the grinding length of the milling cutter, a4 represents the grinding depth of the milling cutter, and a5 represents the grinding path of the milling cutter.

[0015] S3, respectively constructing training samples for each reed piece of the reed vibrating instrument, specifically: using the number of the reed piece, the m groups of grinding parameters of the reed piece in step S1, and the frequency deviation values ​​corresponding to the m groups of grinding parameters to form a sample, to obtain m training samples of the reed piece;

[0016] S4. The training samples of all the reeds of the reed-vibrating musical instrument are used to form a training set. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 The present invention is a flow chart of the automatic frequency tuning method of the reed piece. DETAILED DESCRIPTION

[0018] The present invention is further described in detail below with reference to the accompanying drawings.

[0019] Embodiment 1: Figure 1 As shown, a method for automatic frequency tuning of a reed piece is provided. First, a number is set for each reed piece of a reed vibrating musical instrument, and each reed piece has a different number. The current vibration frequency of each reed piece is obtained respectively, and the current vibration frequency of each reed piece is compared with its reference frequency. If the two are consistent, it indicates that the frequency of the reed piece is qualified. If they are inconsistent, it indicates that the frequency of the reed piece is unqualified. At this time, the difference between the current vibration frequency of the reed piece and its reference frequency is calculated, that is, the frequency deviation value of the reed piece. The number of the reed piece and the frequency deviation value of the reed piece are used to form a two-dimensional vector, which is input into a trained deep learning model for prediction to obtain a 5-dimensional feature data output, wherein the 5-dimensional feature data respectively correspond to five grinding parameters of a grinding device: milling cutter rotation angular velocity, milling cutter movement speed, milling cutter grinding length, milling cutter grinding depth and milling cutter grinding path. Then, after setting the five grinding parameters of the grinding device according to the 5-dimensional feature data, the grinding device is used to grind the reed piece so that its vibration frequency is equal to its reference frequency.

[0020] In this embodiment, the existing mature automatic data acquisition technology can be used to realize the frequency acquisition of the reed piece, and compared with the reference frequency to obtain the deviation of the digital quantity (i.e., the frequency deviation value), eliminate human subjective intervention, ensure the accuracy of the frequency deviation value, predict five grinding parameters through the deep learning model, and use the existing mature grinding equipment to automatically perform grinding for frequency modulation, which ensures the reliability and stability of the entire frequency modulation process. At the same time, because a large-scale training set can be established for the deep learning model, the deep learning model can have a high prediction accuracy, and the five predicted grinding parameters are more accurate, ensuring a high frequency modulation accuracy. In addition, because it is possible to combine automatic data acquisition technology, deep learning models and grinding equipment for automatic frequency modulation, the consistency of frequency modulation is guaranteed, which greatly reduces the requirements for the professionalism of personnel in the frequency modulation of the reed.

[0021] Embodiment 2: This embodiment is basically the same as Embodiment 1, except that: in this embodiment, the specific process of obtaining the vibration frequency of each reed piece is: first drive the reed piece to vibrate and emit sound, then collect the sound signal, and then use Fourier transform to convert the sound signal into a frequency signal, and the frequency signal is recorded as the current vibration frequency of the reed piece.

[0022] Embodiment 3: This embodiment is basically the same as Embodiment 1, except that: in this embodiment, the trained deep learning model is obtained by training the deep learning model using a training set, wherein the training set is constructed by the following steps:

[0023] S1. Each reed of the reed vibrating instrument is processed as follows:

[0024] S1.1. Set five grinding parameters of the grinding equipment: milling cutter rotation angular velocity, milling cutter moving speed, milling cutter grinding length, milling cutter grinding depth and milling cutter grinding path;

[0025] S1.2, obtaining the current vibration frequency of the reed piece, that is, the vibration frequency before this grinding, and after grinding the reed piece with a grinding device, obtaining the current vibration frequency of the reed piece, that is, the vibration frequency after this grinding, and calculating the difference between the vibration frequencies of the reed piece before and after this grinding, that is, the frequency deviation value corresponding to a set of grinding parameters consisting of the current five grinding parameters;

[0026] S1.3, four of the five grinding parameters of the grinding equipment are used as holding parameters, and the other one is used as an adjustment parameter, the four holding parameters remain unchanged, and the data of the adjustment parameter is changed to obtain a set of grinding parameters consisting of five grinding parameters again;

[0027] S1.4, using the same method as step S1.2, obtain the frequency deviation value corresponding to a set of grinding parameters consisting of the current five grinding parameters;

[0028] S1.5, repeating steps S1.3 and S1.4 until the frequency deviation values ​​corresponding to m groups of grinding parameters are obtained, where m is an integer greater than or equal to 50, wherein when repeating step S1.3, the five grinding parameters of the grinding equipment are changed as adjustment parameters respectively;

[0029] S2. Each training sample is represented as [X, Y], where X is a two-dimensional vector as input data, X = [x1, x2], x1 is the number of the reed piece, x2 is the frequency deviation value, and Y is a five-dimensional vector as output data, i.e., label, Y = [a1, a2, a3, a4, a5], a1 represents the angular velocity of the milling cutter, a2 represents the moving speed of the milling cutter, a3 represents the grinding length of the milling cutter, a4 represents the grinding depth of the milling cutter, and a5 represents the grinding path of the milling cutter.

[0030] S3, respectively constructing training samples for each reed piece of the reed vibrating instrument, specifically: using the number of the reed piece, the m groups of grinding parameters of the reed piece in step S1, and the frequency deviation values ​​corresponding to the m groups of grinding parameters to form a sample, to obtain m training samples of the reed piece;

[0031] S4. The training samples of all the reeds of the reed-vibrating musical instrument are used to form a training set.

[0032] In this embodiment, the settings and changes of the five grinding parameters at the grinding equipment are set according to actual needs, and the value of m is set according to actual needs, so that the prediction accuracy of the trained deep learning model can meet the actual application needs.

[0033] To sum up, the automatic frequency tuning method of the reed piece of the present invention realizes the automatic frequency tuning of the reed piece based on the current mature signal acquisition technology, deep learning model and grinding equipment, eliminates human subjective intervention, can achieve efficient, stable and high-precision frequency tuning, and greatly reduces the requirements for the professionalism of personnel in the frequency tuning of the reed piece, reduces the labor cost of frequency tuning, and provides great convenience for the frequency tuning of the reed piece. The application and promotion in the industry will greatly enhance the productivity of the musical instrument industry and has broad application prospects.

Claims

1. A method for automatically tuning the frequency of a reed, characterized in that First, a number is set for each reed piece of the reed vibrating instrument, and each reed piece has a different number. The current vibration frequency of each reed piece is obtained respectively, and the current vibration frequency of each reed piece is compared with its reference frequency. If the two are consistent, it indicates that the frequency of the reed piece is qualified. If they are inconsistent, it indicates that the frequency of the reed piece is unqualified. At this time, the difference between the current vibration frequency of the reed piece and its reference frequency is calculated, that is, the frequency deviation value of the reed piece. The number of the reed piece and the frequency deviation value of the reed piece are used to form a two-dimensional vector, which is input into the trained deep learning model for prediction to obtain a 5-dimensional feature data output, wherein the 5-dimensional feature data respectively correspond to the five grinding parameters of the grinding equipment: milling cutter rotation angular velocity, milling cutter movement speed, milling cutter grinding length, milling cutter grinding depth and milling cutter grinding path. Then, after setting the five grinding parameters of the grinding equipment according to the 5-dimensional feature data, the grinding equipment is used to grind the reed piece so that its vibration frequency is equal to its reference frequency.

2. The automatic frequency tuning method of a reed according to claim 1, characterized in that The specific process of obtaining the vibration frequency of each reed piece is: first drive the reed piece to vibrate and make a sound, then collect the sound signal, and then use Fourier transform to convert the sound signal into a frequency signal, and the frequency signal is recorded as the current vibration frequency of the reed piece.

3. The automatic frequency tuning method of a reed according to claim 2, characterized in that The trained deep learning model is obtained by training the deep learning model using a training set, wherein the training set is constructed by the following steps: S1. Each reed of the reed vibrating instrument is processed as follows: S1.

1. Set five grinding parameters of the grinding equipment: milling cutter rotation angular velocity, milling cutter moving speed, milling cutter grinding length, milling cutter grinding depth and milling cutter grinding path; S1.2, obtaining the current vibration frequency of the reed piece, that is, the vibration frequency before this grinding, and after grinding the reed piece with a grinding device, obtaining the current vibration frequency of the reed piece, that is, the vibration frequency after this grinding, and calculating the difference between the vibration frequencies of the reed piece before and after this grinding, that is, the frequency deviation value corresponding to a set of grinding parameters consisting of the current five grinding parameters; S1.3, four of the five grinding parameters of the grinding equipment are used as holding parameters, and the other one is used as an adjustment parameter, the four holding parameters remain unchanged, and the data of the adjustment parameter is changed to obtain a set of grinding parameters consisting of five grinding parameters again; S1.4, using the same method as step S1.2, obtain the frequency deviation value corresponding to a set of grinding parameters consisting of the current five grinding parameters; S1.5, repeating steps S1.3 and S1.4 until the frequency deviation values ​​corresponding to m groups of grinding parameters are obtained, where m is an integer greater than or equal to 50, wherein when repeating step S1.3, the five grinding parameters of the grinding equipment are changed as adjustment parameters respectively; S2. Each training sample is represented as [X, Y], where X is a two-dimensional vector as input data, X = [x1, x2], x1 is the number of the reed piece, x2 is the frequency deviation value, and Y is a five-dimensional vector as output data, i.e., label, Y = [a1, a2, a3, a4, a5], a1 represents the angular velocity of the milling cutter, a2 represents the moving speed of the milling cutter, a3 represents the grinding length of the milling cutter, a4 represents the grinding depth of the milling cutter, and a5 represents the grinding path of the milling cutter. S3, respectively constructing training samples for each reed piece of the reed vibrating instrument, specifically: using the number of the reed piece, the m groups of grinding parameters of the reed piece in step S1, and the frequency deviation values ​​corresponding to the m groups of grinding parameters to form a sample, to obtain m training samples of the reed piece; S4. The training samples of all the reeds of the reed-vibrating musical instrument are used to form a training set.

Images