TWI718716B - Method for detecting scales triggered in musical instrument - Google Patents

Abstract

The disclosure provides a method for detecting method for detecting scales triggered in musical instrument, comprising: inputting an audio signal of each scale of a target musical instrument; performing an aliasing effect on the audio signal to obtain a non-aliased audio signal; performing feature extraction on the non-aliased audio signal, optimizing the extracted features to obtain a final feature table; continuously inputting an audio signal to-be-tested; performing the aliasing effect and frequency transformation on the audio signal to-be-tested; and comparing with the features in the final feature table to determine if a certain scale being triggered.

Description

Detection method of musical instrument scale trigger

The invention relates to a method for detecting musical scale triggering.

Percussion instruments (or percussion instruments) not only account for a considerable proportion of the music and music education market, the mini or simplified version of percussion instruments is a classic toy that lasts forever. For example, the simplest combination is a set that can be sent separately. Do, Re, Mi, Fa, So, La, Si, Do scales, a toy musical instrument with a total of eight bells; simple in structure, it has many functions for children's entertainment, sound recognition, or music inspiration. On the other hand, due to the increasing popularity of smart electronic products, more and more electronic products have added sound-based related technologies to increase ease of use and market competitiveness. Therefore, for the aforementioned toys of percussion instrument type, a set of algorithms is designed to detect whether those scales are triggered while only relying on the microphone to pick up the sound, and with the function of detecting whether the instrument is triggered, increase the The new play method of percussion type toys will have a wider range of applications.

An embodiment of the present invention discloses a detection method of musical instrument scale trigger, including: a musical instrument scale feature value table generation stage, and a musical instrument scale trigger detection stage; wherein the musical instrument scale feature value table generation stage further includes: input one The sound signal of each scale of the target musical instrument; perform anti-frequency effect processing on the sound signal of each scale to obtain an anti-aliased sound signal; perform feature extraction on the anti-aliased sound signal to obtain a primary feature value table; The initial eigenvalue table optimizes the eigenvalues to obtain a final eigenvalue table; the instrument scale trigger detection stage further includes: continuously inputting a sound signal to be tested; processing the sound signal to be tested for frequency effect processing to obtain a sound signal. Falsify the sound signal to be tested; perform audio frame and time-frequency conversion processing on the sound signal to be tested without falsification to obtain a frequency domain signal to be tested; compare the frequency domain signal to be tested with the final characteristic value table, To determine whether a certain scale of the target instrument is triggered by the detection result.

In a preferred embodiment, the aliasing effect processing step further includes: performing finite impulse response (FIR) low-pass filtering on the sound signal of each scale of the target instrument to obtain a filtered signal; and, The filtered signal is resampled to obtain an audio signal without aliasing.

In a preferred embodiment, the feature extraction step further includes: sound-framing the unaltered sound signal to generate a sound-framed sound signal; and time-frequency conversion of the sound-framed sound signal to obtain A frequency domain signal; the frequency domain signal is statistically averaged by percussion sound energy to obtain a percussion energy characteristic; the strongest and the second-strongest frequency energy characteristics of the percussive scale are counted, and the audio energy of all scales is counted and knocked The statistical features of the percussion energy are combined to obtain the primary feature value table.

In a preferred embodiment, the feature value optimization step further includes: performing feature value cross-comparison on the primary feature value table, which is to compare all feature values in the primary feature value table with other feature values in pairs. Check whether the frequency characteristics of scales are too close to each other; perform trigger and false trigger tests, input the non-articulated sound signal to check whether each scale has a normal trigger and whether it will cause false triggers; change the initial-level characteristics In the value table, the eigenvalue cross-comparison result compares the scales with too similar frequency characteristics to optimize the tune threshold value of the similar characteristics; and, in the primary eigenvalue table, the trigger and false trigger test steps are tested If the frequency characteristics of the scale with a false trigger are replaced by the characteristics of the false scale, the scale characteristics are extracted again and the false trigger is eliminated, and a final characteristic value table is obtained after all are completed.

In a preferred embodiment, the step of comparing the frequency domain signal to be measured with the final feature value table is to compare the frequency domain signal with the tap sound energy feature counted in the final feature value table , It can be judged whether a certain scale of the instrument is struck, and the relative distance of the instrument struck.

The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand the other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied by other different specific examples, and various details in the specification of the present invention can also be modified and changed based on different viewpoints and applications without departing from the spirit of the present invention.

Among them, the structure, ratio, size, etc. shown in the drawings in this specification are only used to match the content disclosed in the specification for the understanding and reading of those who are familiar with the technology, and are not intended to limit the implementation of the present invention. Conditions, so it does not have technical significance. Any structural modification, proportion change or size adjustment, without affecting the effect of the present invention and the purpose that can be achieved, shall fall within the disclosure of the present invention. The technical content must be able to cover the scope.

As shown in FIG. 1, the embodiment of the present invention discloses a detection method of musical instrument scale trigger, which is suitable for a target musical instrument with a plurality of scales. The detection method includes: a musical instrument scale feature value table generation stage 110, and a The musical instrument scale trigger detection stage 120; wherein the musical instrument scale feature value table generation stage 110 further includes: step 111, inputting a sound signal of each scale of a target musical instrument; step 112, performing a frequency effect on the sound signal of each scale Process to obtain a non-forged sound signal; step 113, perform feature extraction on the forged sound signal to obtain an initial feature value table; step 114, perform feature value optimization on the primary feature value table to obtain a final feature Value table; the musical instrument scale trigger detection stage 120 further includes: step 121, continuously inputting a sound signal to be tested; step 122, processing the sound signal to be tested with forged effect processing to obtain a sound signal without forged frequency to be tested; Step 123: Perform audio frame and time-frequency conversion processing on the sound signal to be measured without false frequency to obtain a frequency domain signal to be measured; Step 124: Compare the frequency domain signal to be measured with the final characteristic value table to The detection result that determines whether a certain scale of the target instrument is triggered.

In other words, an instrument scale trigger detection method of the present invention first processes the sound signals of each scale of a target instrument with multiple scales in the previous stage to obtain a sound including each scale of the target instrument The final characteristic value table of the characteristic value; then in the later stage, the input sound signal to be measured is continuously received, and the sound signal to be measured is determined by the comparison method whether the sound signal to be measured matches the sound characteristic value of each scale of the target instrument, and Determine whether each scale of the target instrument is triggered.

FIG. 2 is a schematic diagram of the process of processing the fortification effect of the sound signal in the detection method of the musical instrument scale trigger of the present invention; as shown in FIG. 2, the fortification effect processing step 112 further includes: step 112a, the target instrument The sound signal of each scale is subjected to finite impulse response (FIR) low-pass filtering to obtain a filtered signal; and, 112b, the filtered signal is resampled to obtain a frequency-free sound signal.

Specifically, the main purpose of the aliasing effect processing step is to filter out the noise in the original sound signal, and then re-sampling, so as to obtain a so-called clean and noise-free sound signal.

FIG. 3 is a schematic diagram of the process of feature extraction of the forged sound signal in the detection method of musical instrument scale trigger of the present invention; as shown in FIG. 3, the feature extraction step 113 further includes: step 113a, performing the non-forged sound signal Sound framed to generate a sound framed sound signal; step 113b, time-frequency conversion is performed on the sound framed sound signal to obtain a frequency domain signal; step 113c, percussion sound energy statistical average of the frequency domain signal Then obtain a percussive energy feature; and, in step 113d, count the strongest and second-strongest frequency energy features in the percussive scale, and merge the audio energy statistics of all the scales with the percussive energy statistical features to obtain the primary feature value table .

It is worth noting that the so-called frame is to gather N sampling points into one observation unit, called the frame. Usually, the value of N is 256 or 512, and the time covered is about 20-30 ms. In order to avoid excessive changes between two adjacent sound frames, there is usually an overlap area between two adjacent sound frames. It is worth noting that the aforementioned N value, the length of time covered, and whether the sound frames overlap or not are all conventionally used to illustrate the embodiments of the present invention, but are not limited to this in practical applications. Furthermore, in a preferred embodiment, the conversion method used in the time-frequency conversion is fast Fourier conversion, but it is not limited to this. Similarly, in a preferred embodiment, the calculation function used in the calculation of the spectral energy in the statistical average of the percussion sound energy is an absolute value function, but it is not limited to this.

FIG. 4 is a schematic diagram of a flow chart of eigenvalue optimization in a detection method of musical instrument scale triggering of the present invention; as shown in FIG. 4, the eigenvalue optimization step 114 further includes: step 114a, performing eigenvalues on the primary eigenvalue table Cross-comparison is to compare all the feature values in the primary feature value table with other feature values in pairs to check whether the frequency features of the scales are too close to each other; step 114b, perform trigger and false trigger tests, and then Input the false frequency sound signal to check whether each scale has a normal trigger and whether it will cause a false trigger; step 114c, optimize the threshold value of the scale with similar characteristics, and cross-compare the result from the characteristic value in the initial-order characteristic value table Compare the scales with too similar frequency characteristics to optimize the threshold value of the scales with similar characteristics; and, in step 114d, replace the characteristics of the mistouched scales, and replace the features of the initial-order feature value table with the trigger and false trigger test steps. If the frequency characteristics of the scale with a false trigger are replaced by the characteristics of the false scale, the scale characteristics are extracted again and the false trigger is eliminated, and a final characteristic value table is obtained after all are completed.

Wherein, in step 114a, whether there is a condition that the frequency characteristics of the scales are too close to each other can be determined according to whether the difference in frequency characteristics is lower than a predetermined threshold. Furthermore, in step 114b, because the non-aliased sound signal represents the signal that is resampled after filtering, it does not contain other noise; on the other hand, the so-called false triggering refers to the frequency characteristics of the scale If there is noise that exceeds a predetermined threshold, it can be judged as a false trigger.

It is worth noting that, in step 122 of the musical instrument scale trigger detection stage 120, the sound signal to be tested is processed for the forged frequency effect to obtain a sound signal to be measured without forged frequency, which is related to the scale feature value table of the musical instrument. Step 112 in the generation stage 110 is the same, therefore, the details will not be repeated here; similarly, in the step 123 of the instrument scale trigger detection stage 120, the sound signal to be tested is framed and time-frequency. The conversion process to obtain a frequency domain signal to be measured is the same as the steps 113a and 113b in the feature extraction step 113, so the details are not repeated here.

Specifically, the step of comparing the frequency domain signal to be measured with the final feature value table is to compare the frequency domain signal with the tap sound energy features counted in the final feature value table, and it can be judged Whether a certain scale of the instrument is struck, and the relative distance of the instrument struck.

Although the implementation has been described with reference to many illustrative embodiments of the present application, it should be understood that those skilled in the art can think of many other changes and embodiments, and these changes and embodiments will fall within the spirit and scope of the principles of the present disclosure. Inside. In particular, within the scope of the present disclosure, the drawings and the attached patent applications, various changes and modifications can be made to the components and/or arrangements of the subject combination arrangement. In addition to changes and modifications to the components and/or settings, alternative uses will be obvious to those skilled in the art.

110: Musical instrument scale feature value table generation stage 111: Input the sound signal of each scale of the target instrument 112: Perform anti-frequency effect processing on the sound signal of each scale to obtain a non-aliased sound signal 112a: Perform finite impulse response low-pass filtering on the sound signal of each scale of the target instrument to obtain a filtered signal 112b: Resample the filtered signal to obtain an audio signal without aliasing 113: Perform feature extraction on the artificial sound signal to obtain an initial feature value table 113a: Frame the audio signal without aliasing to generate a framed audio signal 113b: Perform time-frequency conversion on the framed sound signal to obtain a frequency domain signal 113c: The frequency domain signal is statistically averaged by percussion sound energy to obtain a percussion energy characteristic 113d: Count the strongest and second-strongest frequency energy characteristics of the percussive scale, combine the audio energy statistics of all scales with the percussive energy statistical characteristics to obtain the primary feature value table 114: Perform feature value optimization on the primary feature value table to obtain a final feature value table 114a: Cross-compare the eigenvalues of the primary eigenvalue table 114b: Perform trigger and false trigger test 114c: Optimization of scale threshold with similar features 114d: Feature replacement of mistouched scale 120: Instrument scale trigger detection stage 121: Continuously input a sound signal to be tested 122: Perform anti-frequency effect processing on the sound signal to be tested to obtain a sound signal to be tested without anti-frequency 123: Perform audio frame and time-to-frequency conversion processing on the unaltered sound signal to be measured to obtain a frequency domain signal to be measured 124: The detection result of comparing the frequency domain signal to be measured with the final characteristic value table to determine whether a certain scale of the target instrument is triggered

FIG. 1 is a schematic flow chart of a method for detecting musical scale triggering according to the present invention; 2 is a schematic diagram of the process of processing the fore-frequency effect of a sound signal in a method for detecting musical scale triggers of the present invention; 3 is a schematic diagram of the process of feature extraction of the forged sound signal in a method for detecting musical scale triggering of the present invention; 4 is a schematic diagram of a flow chart of eigenvalue optimization in a detection method of musical instrument scale triggering of the present invention.

100: musical instrument scale feature value table generation stage

111: Input the sound signal of each scale of the target instrument

112: Perform anti-frequency effect processing on the sound signal of each scale to obtain a non-aliased sound signal

113: Perform feature extraction on the artificial sound signal to obtain an initial feature value table

114: Perform feature value optimization on the primary feature value table to obtain a final feature value table

120: Instrument scale trigger detection stage

121: Continuously input a sound signal to be tested

122: Perform anti-frequency effect processing on the sound signal to be tested to obtain a sound signal to be tested without artificial frequency

123: Perform audio frame and time-to-frequency conversion processing on the unaltered sound signal to be measured to obtain a frequency domain signal to be measured

124: The detection result of comparing the frequency domain signal to be measured with the final characteristic value table to determine whether a certain scale of the target instrument is triggered

Claims (3)

A method for detecting musical instrument scale triggering is suitable for a target musical instrument with a plurality of scales, including: a musical instrument scale feature value table generation stage; and a musical instrument scale trigger detection stage; wherein the musical instrument scale feature value table generation stage It further includes: inputting a sound signal of each scale of a target instrument; performing anti-frequency effect processing on the sound signal of each scale to obtain a non-artificial sound signal; performing feature extraction on the anti-frequency sound signal to obtain an initial feature Value table; eigenvalue optimization of the primary eigenvalue table to obtain a final eigenvalue table; the musical instrument scale trigger detection stage further includes: continuously inputting a sound signal to be tested; performing a false frequency effect on the sound signal to be tested Process to obtain a sound signal to be measured without forged frequency; perform audio frame and time-frequency conversion processing on the sound signal to be measured without forged frequency to obtain a frequency domain signal to be measured; the frequency domain signal to be measured and the final characteristic value Tables are compared to determine the detection result of whether a certain scale of the target instrument is triggered; wherein, the feature extraction step further includes: frame the sound signal without fortification to generate a framed sound signal ; Perform time-frequency conversion on the framed sound signal to obtain a frequency domain signal; perform percussion energy statistical averaging on the frequency domain signal to obtain a percussion energy feature; count the strongest and secondary percussive scales Strong frequency energy features, combining the audio energy statistics of all scales with the percussion energy statistics to obtain the primary feature value table; Wherein, the eigenvalue optimization step further includes: comparing all the eigenvalues in the primary eigenvalue table with other eigenvalues in pairs, checking whether the frequency characteristics of the scales are too similar to each other; inputting the non-forged sound signal , Check whether each scale has a normal trigger and whether it will cause a false trigger; in the initial-level feature value table, the cross-comparison result of the feature value shows that the frequency feature is too similar to optimize the scale threshold value of the similar feature ; And replace the false-touch scale feature with the frequency characteristics of the scale whose false trigger is detected in the trigger and false-trigger test steps in the primary feature value table, extract the scale feature again and eliminate the false trigger After all the features are completed, a final feature value table will be obtained. For example, the detection method of musical instrument scale trigger in claim 1, wherein the frequency artifact processing step further comprises: performing finite impulse response (FIR) low-pass filtering on the sound signal of each scale of the target instrument to obtain a filtered signal ; And, re-sampling the filtered signal to obtain an audio signal without aliasing. For example, the detection method of musical instrument scale trigger of claim 1, wherein the step of comparing the frequency domain signal to be measured with the final characteristic value table is to compare the frequency domain signal with the statistics in the final characteristic value table By comparing the energy characteristics of the percussion sound, it can be judged whether a certain scale of the instrument has been percussive, and the relative distance of the percussion.

Images