TW201034000A - Automatic grading method for karaoke song singing - Google Patents

Abstract

The present invention provides an automatic grading method for karaoke song singing. The method obtains the score for sensation of tone, score for tempo and score for feeling respectively by comparing the pitch, tempo point location and volume of the signer to the pitch, tempo point location and volume of main melody; then sums up the weighted total score in a weighted scoring manner. By the design of this method, the signer's errors in pitch, tempo point location and volume at each song section can be accurately calculated; and uses the display effect of pitch curve and volume curve to let the signer easily know which part he does not sign accurately and which part needs improvement, thus this invention providing dual effects of teaching and entertainment thereby having the features of practical advancement.

Description

201034000. VI. Description of the invention: [Technical field of invention] The present invention relates to an automatic evaluation of karaoke song accompaniment, which refers to an innovative designer who calculates scores based on multiple scores, such as pitch, rhythm and emotion. . [Prior Art] Press, during the karaoke (KARA0K) song accompaniment process, the ® player is usually accompanied by an automatic scoring function. However, such a function meter is often only a rough estimate of the overall score, or it may be a decibel value of only the song sound. High and low as the only reference for the evaluation, the score of the record player, even related to the quality of the song sings, so can only achieve a little entertainment effect, just judge the song sing good or bad, so for singing The practice of singing is not helpful. ❿ 是以 是以 是以 是以 是以 是以 针对 针对 针对 针对 针对 针对 针对 针对 针对 针对 针对 针对 针对 针对 针对 针对 针对 针对 针对 针对 针对 针对 针对 针对 针对 针对 针对 针对 针对 针对 针对 针对 针对 针对 针对 针对 针对 针对 针对 针对In view of this, the inventor has been engaged in the related product design and design experience for many years. In view of the above objectives, the design and the prudence have finally become practical and practical. [Description of the Invention] The method, special, and the current knowledge of the current companion is based on the sing and some of the companionships are not really true, after the use of the practicality of the standard and the direction of manufacturing development and evaluation, 3 201034000 The main purpose of the present invention is to provide an automatic scoring method for Karaokea song accompaniment, and the problem to be solved is that the automatic scoring function for the conventional Karaoke karaoke machine can not really judge the singing quality. For the singer to practice singing, there is no problem to help to solve the problem; the technical feature of the problem solving of the present invention lies in the automatic scoring method of the karaoke song accompaniment, mainly by comparing the pitch and beat of the singer The position and volume and the pitch, beat position and volume of the main melody of the song are respectively obtained into the pitch score, the rhythm score and the emotion score, and finally the weighted total score is calculated by the weighted square formula 0; thereby, the innovative unique design makes the invention Compared with the prior art, the pitch, the beat position and the volume error of the singer in each song passage can be accurately calculated, and The use of pitch curve, showing volume curve, so that the singer can easily know where to sing inaccurate and which ground to make up, but also has the dual effect of education and entertainment with practical and progressive _ indeed. [Embodiment] Please refer to the patent application of the karaoke song accompaniment automatic scoring method of the present invention as shown in the flute 1 to 16, but these examples are for illustrative purposes only. Restricted by this structure; the Karaoke song sings the position of the selfie point and the chic point, mainly by comparing the pitch, the way of the singer, the volume of the gossip and the pitch of the main theme of the song, the position of the beat and the volume For the sub-items, the most (1) scores of the pitch score, rhythm score and sentiment score are obtained, and the scores of the weighted total scores of the scores are calculated by weighted scoring. Animal-personal t(10) % song, in addition to the characteristics of personal voice 'To comment 4 201034000 • The four songs of its songs and songs should mainly include three senses, / is the sense of sound, the second is the sense of rhythm, the third is the emotion 'sense is the judgment The pitch is relative to the pitch accuracy of each note; the sense of rhythm is the error that determines the position of the beat', including the vocal beat and the end of the beat; the emotion is to determine the change in its volume, including the volume change of each sentence. And the overall volume change. The methods for specifically obtaining the pitch score, rhythm score and emotion score are respectively described as follows: (1) Sound score: Please refer to Fig. 1 for every short period of time (for example, 0·1 second) 0, by singing The microphone audio sung by the singer calculates the pitch of the singer. This pitch estimation is the fundamental frequency of the human voice, and the acquisition method can usually use the autocorrelatio n function. The method obtains, and then converts the fundamental frequency into a relative scale by the pitch estimator, and then compares the scale of the scale with the scale captured in the main melody of the music, and gives the scale a pitch score, thus calculating all The pitch score of the scale is until the end of the concert, and then the average pitch score is output. As shown in Figure 2, the details are as follows: ® First is the "initial parameter setting", in which the number of scales initialized is n = 0, and the vocal match with the scale of the scale is NoteHit = 0, and the bass sense is matched. The value NoteHitAround = 0, NoteHit indicates the number of time periods during which the human voice is highly matched during the performance of the scale, and NoteHitAround indicates the number of time periods in which the human voice is higher than the scale within one semitone, and then the next period of time is obtained. The main melody scale and the person who calculated the period of time have a high voice. The main melody scale is directly obtained from the file such as midi. The scale of the performance relative to the time is obtained according to the increase of time. The human voice is high (the fundamental frequency), and can be transcoded. The table is converted to a scale relative to the pitch. For example, the scale "A4 5 201034000 • The frequency is 440 Hz. For every octave, the frequency is increased by two times. For example, the frequency of the scale "A5" is 880 Hz, an octave has 12 semitones, the frequency difference between each semitone is 2 (9) 2) times, because if the vocal is different from the frequency of the scale by 2 仏 or 1/2 times, etc. It is the same, so we adjust the calculated human sound level N〇te-P and the main melody scale Note-» through the scale ± 12 semitones, so that the difference is between +6 semitones v and 5 semitones. 'Note-p = Note_p +12*i, i is a non-〇 integer, so -5 <= Note_p - Notejn <=6. Then, determine whether it is a new sound © 卩自卩 什 算 算 上 上The pitch score ' then reset the starting parameters, NoteHit = 0 and NoteHitAround = 0 and the number of scales n = η ten 1 'If otherwise compare whether the main melody scale matches the human sound level, this match means that the error is, one Within a relatively small allowable range, such as 〇. 5 semitones. If matched, increase the treble match value of the scale N〇teHit = NoteHi t + 1, otherwise judge whether the main melody scale and the human sound level match the bass sense. The low-day-sensing match indicates that the error is within a relatively large allowable range, such as within one semitone of the difference, and if so, increase the scale-to-sound match value NO oteHltAr〇und = NoteHitAround + 1, and then return to the master who obtained the next period of time. The rhythm scale and the calculation of the human voice are high. The above-mentioned "calculation of the pitch score of the previous scale", the algorithm is as shown in Fig. 3: First, the length of the main string of the music is N〇teLength(m), where: m = 〇 1, 2, ..., Μ The Μ is the total number of scales ' then judge the high-pitched sound matching value N〇teHi1; whether it is greater than zero', then calculate the high-pitched scale matching score:

PitchScore(m)- PSH + Kl * NoteHit(m)/ NoteLength(m); where .PSH ’ K1 is an adjustable empirical value parameter, otherwise the calculation is low 6 201034000 . Phonological scale matching score: .

PitchScore(m)= PSL - K2 * NoteHitAround(m)/ NoteLength(m); where: PSL, K2 are adjustable empirical value parameters, and limit: 0<=PitchScore(m)<= 100 The last scale, if otherwise repeat the above process, if it is, then "calculate the average pitch score", the algorithm is the weighted average of all PitchScor e (in) with the length of NoteLength (m) as the weighted proportion, as follows: Let the total length of the scale Mv = I = 汍fmj, Score of Pitch: 1 1VX ~1

NL SOP =-^PitchScore(m) · NoteLength(m) φ (2) Rhythm score: Rhythm sense is the calculation of the vocal beat point and the start time of the music main melody scale and the vocal end beat point and the music main melody scale The degree of matching of the end time is determined. To accurately estimate the position of the singer's beat at each beat, here we estimate the singer's pitch as a change in the time of the different scales, and then determine the accuracy of the beat, as shown in Figure 4. It is shown that the method described in the first figure first estimates the pitch of the human voice and the scale of the music main melody, and then generates an average rhythm sensation score through the rhythm estimator. 7 201034000 . Through the rhythm estimator, the human voice is first converted to the relative scale and then compared to the scale, and the scale obtained in the main melody is in time error. The error of this time includes the early or delayed attack point. The end time is taken and the time error of each scale is recorded, and then the rhythm sense of the scale is given, so that the rhythm scores of all the scales are calculated until the end of the singing and then the average rhythm score is output. As shown in Fig. 5, the rhythm delay matcher and the rhythm sense lead matcher can be used to calculate the person 0 and the scale at the time of the current, previous and next music main chord scales of the converted human voice level, respectively. The degree of matching between the delay or the lead, the delay time and the lead time of the vocal beat point or the vocal beat point, and then the score of the rhythm of the scale is obtained by calculating the score of the scale, and accordingly, starting from a scale, We calculate the rhythm error for each scale until the end of the next scale, and then calculate the average rhythm score. Please refer to Figure 6, the rhythm delay matcher is the beginning of whether the new music scale is broken, otherwise it is judged whether the start point delay time has been set, and if so, the end, otherwise judge the sound level and sound Whether the scale matches, if otherwise increase the slap beat delay time, if yes, set the slap beat delay time, and then end, this delay time indicates the time error after the start of the music stage, the vocal is later than it; if it is a new music sound At the beginning, reset the singer beat delay time and record the end of the previous scale, then judge whether the human sound level is equal to the previous music main chord scale, and if so, determine whether the next personal sound level is the same as the previous music main chord scale. Match until no, and then set the end beat delay time after the bundle, this delay time indicates the time error of the vocal than the end of the last musical scale. Difference > Sense, Sound Beam ΛΛ - that is, the most vocal music 6 and scales with the law of the night 8 201034000 - Please refer to Figure 7, the rhythm sense lead matcher, first determine whether it is a new musical scale Start, if not, determine whether the vocal matches the current musical scale, and if so, record the human sound between the knots, otherwise set the ending beat time, and then end, the advance indicates that the vocal is earlier than it before the end of the musical scale If the end time is wrong, the ending time of the new music scale is reset, and the ending time of the ending beat is started and the time is started. Then, it is determined whether the human sound level matches the main musical scale, and if so, whether the previous human sound level is related to the sound. ❹ Match, until no, then set the start time of the singer beat time and end the lead time to indicate that the vocal is earlier than the beginning of the sound before the start of the musical scale. Then, the vocal beat delay time, the vocal beat point lead time delay point time and the end beat time advance time, calculate the score rhythm number SOB (Score of Beat), the algorithm is as follows: Let the vocal beat time error is 7Λ9, then, the starting point score S): ❿ SOBS = ^^ + 100 -(l- TDS / Ls) where TDS = singer beat delay time (NoteOnLag) + singer time (NoteOnLead), As and Ls are preset empirical value parameters. Let the beat time error be TDE, then: end beat score (S0BE): SOBE = Ae+ 100-(1- TDE / Le)

It is the time difference of the scale beam; record the chord order, the time and the knot sense (S0B point is over 9 201034000. Where 'TDE = end beat delay time (Note0ffL) + end beat lead time (NoteOffLead) , Ae and Le are preset empirical value parameters, the scale rhythm score (S0B):

SOB = SOBS · R + SOBE where R is a preset weighting parameter and 0 <= R <= 1 . (3) Emotional scores: _ Emotion is a parameter that is difficult to measure objectively. Here we use the degree of matching between the average amplitude of the vocal and the average amplitude of the main melody of the music. The average amplitude of the vocal is obtained by calculating the RMS (Root of Mean Square) value of each/personal sound segment. The average amplitude of the main melody of the music can also be calculated by calculating the RMS value of each main melody sound segment or directly by synthesis. The amplitude parameter in the music information is obtained. The algorithm of the RMS is as follows: e RMS II

κ~\ Σ>2(7) where x(i),i = 〇,1,...,Κ-1,κ represents the sound sample points (Samples) of the sound segment, the RMS value, in actual operation, It can also be replaced by other methods such as average amplitude or maximum amplitude. As shown in Fig. 8, the sentiment score estimator calculates the RMS value of the vocal signal and the main melody of the music once every time (about 〇·丨sec), and obtains the RMS sequence of the vocal and music 'hypothesis respectively. MicV 〇1<^ and 丨(4), 卩=〇10 201034000, 1, Nl, ,,, represent the RMS value of the ηth time period, where N is the total length of the song' and MicVol(η) The energy level is adjusted to be the same as MelVol (η), and then averaged according to the length of each scale, and the average RMS sequence of the mth scale of vocals and music is AvgMe lVol (m), AvgMicVol (m) ); AvgMelVol (η), AvgMicVol (η) can be used to calculate the score of S0E (Score of Emotion), first obtain and calculate the overall match degree S0ET of the vocal amplitude curve and the music amplitude curve, which can represent the overall emotional change score, as follows:

Where Μ is the total number of scales, and

Therefore S0ET <= 1〇〇. Then, the calculation of each sentiment score S0MS can be performed. First, AvgMicVol(m) and AvgMelVol(m) are cut into one sentence, assuming that the starting scale of each lyric is S(j), j = 〇, 1, 2,...,L-1, start 'where L is the total number of lyrics, and let S(L) =M, then the emotional change score of each sentence is: 201034000 SOES(j) sa+oi [AvgMic Vol (m )AvgMel Vol (m) m^SU)

XlOO (5(;+l)-l V 5(7+1)-1 , J] AvgMic Vol2 (m) J] AvgMelVol2 {m) m=S(j) j - 0, 1, 2, ··· , Ll , and then calculate the relative emotional change score for each sentence, which is the change in volume of each sentence relative to the overall volume:

First, let Z AvgMic Vol(m)AvgMeI Vol(m) (5(7+1)-1 λ y] AvgMic Vol2 {m) ^ m=S(j) j f M-l ,

AvgMic Vol(m)AvgMel Vol(m) A =

\/w=0_/

' N-l ^ AvgMic Vol2 {m) \m-0 Then SOEA(j) = j i j = 0, L 2, ..., L-l --100,

Uu) 12 201034000 From the above, the average emotional score SOE = a- SOET + ~ Σ (^ * S〇ES(j) + r · SOEA(j)) where α, /3, τ are weighting coefficients, and α + cold + / two. (4) Weighted total score. (Please refer to Figure 9) The Average Evaluated Score obtained by the above SOP, SOB, and S0E is as follows:

AES = p · SOP + q · SOB + r · SOE where P, Q, r are weighting coefficients, and p + r = }. Example of implementation: Taking a song as an example, we calculate the vocal pitch MicPitch(n) and the RMS mean MicVol(n) every 0.1 seconds, and listen to the pitch of the music main melody note MelNoteOi) and calculate it. The RMS average 俭 / / (8), η = 〇, 1' 2, ..., Ν ' Ν indicates the total length of the song, without losing the generality, for the square 说明 explain 'here Ν = 280, indicating the total length of the song The degree is 28 seconds. As shown in Fig. 10, it is a graph of MicPitch(n) and MelNote(n). The solid line in the figure represents the main melody _ the pitch of the note. The vertical axis is the pitch code, and each integer interval is One semitone, 60 means midrange])〇, 61 means the midrange rises Do, 69 means the midrange La' and so on. The dot represents the pitch calculated by the human voice and turns it into a scale code. The height has been adjusted by plus or minus 12, so that the human voice is still closest to the pitch of the main note. The solid line in the figure is a segment, each segment represents a continuous scale, and the height of each segment is undulating, indicating the change of the scale. When the main melody scale is -1, it indicates that the note is a pause 13 201034000 - Or an empty scale, skipping the ignore, when the dot in the figure is zero, it means that the vocal is not calculated, and the vocal may be silent, mute or murmur, etc., and will be regarded as unvoiced. First, by the algorithm of the above-mentioned pitch score, the high-pitched sound matching value NoteHit (m) of the mth scale (as shown by the circle in Fig. 11) and the bass sense matching value NoteHitAround (m) can be obtained (as shown in Fig. 11). In the middle triangle), m = 0,1,2,...Μ, Μ = 3, as shown in Figure 11, let PSH = 50, K1 = 100, and PS L = 35, K2 = 50, Obtain the pitch score for each scale/» (as shown by the rectangle in Figure 11H), and calculate the average pitch score ScoreOfPitch (S0P) after a weighted average of the scale length (as shown by the star in Figure 11). 98 ο Then, by the above algorithm of the rhythm score, the NoteOnLag (m) (circle) and NoteOnLead (m) (star) of the mth scale can be obtained, so that As = 10 and Ls = 10, BeatOnScore can be calculated. m) (rectangle), as shown in Figure 12 and get NoteOffLag (m) (circle) and NoteOffLead (m) (star), let Ae = 50, Le = No teLength (scale length), can calculate BeatOffScore 〇 (m) (circle), as shown in Figure 13, after the weighted average calculation of the scale length, you can get ScoreOfBeatStart (SOBS) = 93. 19, ScoreOfBeatEnd (S0BE) = 99 82 '令R二0. 5, SOB = 96. 5. Then, by the algorithm of the above emotion score, the RMS sequence MelVol (η) of the vocal and music main melody (as shown in FIG. 14 L1) and ic icVol (η) (as shown in FIG. 14 L2) are first obtained. The energy level of MicVol (η) is adjusted to be the same as MelVol (η). As shown in Fig. 14, the average RMS sequence of the mth scale is obtained according to the length if of each scale. AvgMelVol (m) (as shown in Figure 15 L3), AvgMicVol (m) (as shown in Figure 15 L4), as shown in Figure 14 201034000 - 15, set the weighting factor, and thus can calculate SOET 2 98. 33 j sentence SOES(j) (as shown in Figure 16 L5) and SOEA (j) (as shown in Figure 16), j = 0, 1, 2, ... L-1, total number of sentences L = 6, as number 16 The S0ES = 97.2, and S0EA = 95. 67 are shown in the figure. After weighting, ScoreOfEmotion (S0E) = 97. 24 Finally, the weighting coefficient p = 0. 6, q = 0. 2, r = 0. 2, weighted total score: AES 2 p · SOP + q . SOB + r . SOE = 9Ί.55 ❹ Advantages of the present invention: The Karaoke Κ singer automatic scoring method of the present invention mainly relies on the pitch of the singer ', the position and volume of the beat and the main theme of the song The position and volume of the pitch point are respectively obtained by the pitch score, the rhythm score and the emotion, and the innovative unique design of the weighted total score is calculated by weighted scoring. The present invention can accurately calculate the singer in comparison with the prior art. The pitch, the position of the beat and the volume error of the song passage, and the display effect of the Φ curve and the volume curve can be used to make it easy for the singer to know where the sing is not accurate enough and which place needs to be strengthened to achieve the same teaching and entertainment. The practical advancement of the effect. The above embodiments are intended to be illustrative of the present invention, and are not to be construed as limiting the scope of the present invention, and those skilled in the art can understand the principles and principles of the present invention. Changes and modifications are made to the equivalent purpose, and such changes and modifications are to be included in the scope of the patent application as described later. In the picture L6, the score is obtained by comparison and shooting, so that each pitch has a special essence in the text and this is the boundary 15 201034000 - [Simplified illustration] Figure 1: The sound score of the present invention is obtained Method text block diagram 1. Fig. 2 is a block diagram 2 of the method for obtaining the pitch score of the present invention. Fig. 3 is a block diagram 3 of the method for obtaining the pitch score of the present invention. Figure 4: The method for obtaining the rhythm score of the present invention is shown in Figure 5: Figure 5: The method for obtaining the rhythm score of the present invention is shown in Figure 2. Figure 6 of the method for obtaining the rhythm of the present invention. Fig.: The rhythm score obtaining method of the present invention is shown in the figure block diagram 4. Fig. 8'· The block diagram of the emotion score obtaining method of the present invention. Figure 9: Text block diagram of the automatic score estimation method of the present invention. Figure 10: An example of the implementation of the present invention is illustrated in Figure 1. Figure 11: An example of the implementation of the present invention is illustrated in Figure 2. Figure 12: An example of the implementation of the present invention is illustrated in Figure 3. Figure 13: A practical example of the invention is illustrated in reference to Figure 4. Figure 14: A practical example of the invention is illustrated in Figure 5. ^ Figure 15: ❿: An example of the implementation of the present invention is illustrated in Figure 6. Figure 16: An example of the implementation of the present invention is illustrated in Figure 7. [Main component symbol description Note: No component symbol 16

Claims (1)

201034000 - VII. Patent application scope: 1. A method for automatic scoring of Karabok songs, mainly by comparing the pitch of the singer, the position and volume of the beat and the pitch of the main melody of the music, the position of the beat and the volume. The scores of the scores of the sensation scores, the rhythm scores, and the emotion scores are respectively obtained. Finally, the weighted total scores of the score items are calculated by weighted scoring to obtain the scores of the automatic scores. 2. The method for automatically scoring a Karaoke song accompaniment according to Item 1 of the patent application scope, wherein the sound score is obtained by estimating the singer's voice by the microphone sound sung by the singer every other time. High, the estimation of the pitch is obtained by the fundamental frequency of the human voice, and then the fundamental frequency is first converted into a relative scale by a sound estimator, and then the same is found in the scale and the main melody of the music. The degree of matching of the scales 'and gives the scale a sensation © score, so that the θ sense scores of all scales are calculated, and the sings are output, and the average pitch score is output. 3. The karaoke 0K song accompaniment automatic scoring method according to item 2 of the patent application scope, wherein the method for estimating the pitch height can be obtained by using a method based on an autocorrelation function. 4, according to the scope of the patent application, the karaoke 0K song accompaniment automatic 17 201034000 • scoring method, wherein the rhythm sensation score is the calculation of the vocal singer beat and the music main melody scale and the end of the vocal beat The point is determined by the degree of matching with the end time of the musical main melody scale. 5. The Karaoke 0K song accompaniment automatic scoring method according to Item 1 of the patent application scope, wherein the emotion score is determined by using a degree of matching between an average amplitude of the calculated human voice and an average amplitude of the music main melody; The average amplitude of the vocal is obtained by calculating the RMS (Root of Mean Square) value of each individual voice chord. The average amplitude of the music main melody can be calculated by calculating the RMS value of each main melody sound segment or directly by synthesis. The amplitude parameter acquirer in the music information. 18