Summary of the invention
The present invention is directed to the deficiencies in the prior art, a kind of real-time methods of marking of performance accurate, that user experiences of marking is provided.
The technical scheme that technical solution problem of the present invention is taked is:
The inventive method is specifically:
Step 1, original singer's melody is marked;
Step 2, to sing pattern select, comprise simple mode or difficult pattern;
Step 3, show original singer's note and the lyrics in real time;
Step 4, to sing record;
Step 5, extraction singer's pitch;
Step 6, by the contrast of pitch and rhythm, calculate every frame scoring;
Step 7, according to ventilation frequency and degree-of-difficulty factor, scoring is proofreaied and correct;
Step 8, every score is accumulated;
Step 9, calculate the score situation of every;
Step 10, determine the final score of whole song.
Beneficial effect of the present invention: the present invention marks in real time, can obtain performance mark all the time in performance process, and then the cumulative mark of every, the last score that obtains whole song according to every mark of obtaining.The pitch of having realized high accuracy rate extracts, and supports the various innovative functions such as fast forwarding and fast rewinding scoring, ventilation corrections, degree-of-difficulty factor corrections, Combo weighting, has finally realized accurate, the user of marking and has experienced extraordinary real-time methods of marking.
Embodiment
Below in conjunction with accompanying drawing, the present invention is described in detail, be to be noted that described embodiment is only convenient to the understanding of the present invention, and it is not played to any restriction effect.
As shown in Figure 1, a kind of real-time methods of marking of performance accurate, that user experiences of marking is realized according to the following steps:
1. original singer's melody mark, has following characteristics:
By musical expert, the melodic informations such as the lyrics of pitch corresponding all original singer notes and beginning and ending time and correspondence marked and recorded, becoming standard melody, as the standard of scoring reference.
Be as the criterion with original singer's voice, instead of be as the criterion with song numbered musical notation.Like this, when performance, be to be as the criterion with original singer's voice, meet the needs of performance.Because original singer and numbered musical notation are at aspects such as note lengths conventionally, and not quite identical.Taking numbered musical notation as standard, do not meet real performance custom.
Artificial precision of demarcating is very high, is less than 30ms with respect to original singer's time error.
2. difficulty or ease are selected
Select simple mode or difficult pattern.New hand wishes, with simple mode, to be unlikely to mark too low.Master-hand wishes to use difficult pattern, challenges higher performance difficulty and obtains higher mark.
Simple mode, marks relatively loose, and general performance also can obtain good mark, but can not obtain extra high mark.
The scoring of difficulty pattern is stricter, and general performance is not easy to obtain high score, and good performance can obtain the mark higher than simple mode.
3. show in real time original singer's note and the lyrics
Sing according to the lyrics, using original singer's note as pitch and the standard of rhythm.When performance, show in real time original singer's note and the lyrics, as the standard of singing.Very directly perceived.
4. sing recording
In performance process, the sound recording that singer is sung gets off.Only record voice, do not record accompaniment music.
5. singing pitch extracts
With high-precision pitch extracting method, realize the accurate extraction of pitch.Step is as follows:
A, will sing a point frame, the length of a frame is 50ms, then to every frame signal, carries out pitch extraction by subsequent step;
B, quiet detection: the frame that volume is too low, as background noise elimination sound, does not carry out pitch extraction;
C, carry out the bandpass filtering of 60 ~ 900HZ frequency separation;
D, adopt improved auto-correlation algorithm, extract the pitch of every frame;
E, calculating pitch extract degree of confidence, and degree of confidence is low, is judged to be to extract unsuccessfully, greatly reduces wrong extraction result;
F, consecutive frame extract result, carry out median filter smoothness of image, obtain final extraction result.By medium filtering, make to extract erroneous results rate and greatly reduce;
G, by above step, obtain the very high pitch of accuracy rate and extract result.
Show and sing pitch with vernier.Because it is very accurate to sing pitch extraction, in performance process, user can sing with the control representative of fivebyfive the vernier of pitch, and it is very good to experience.Can see in real time that whether performance vernier is consistent with original singer's note.
6. calculate every frame scoring
By pitch and rhythm contrast, calculate every frame score.Method is as follows:
Pitch score: pitch and standard pitch that every frame is extracted are compared, and calculate the score of every frame.Singer can see that the performance of oneself is with the match condition between standard note in real time.
Rhythm score: only, in the time having original singer, singing could score.Like this, only have in the time that rhythm mates completely, just can obtain high score.
Such methods of marking, irrelevant with volume, meet the requirement of specialty scoring.Instead of the score that gives great volume is high.Meanwhile, the people of different tone colors, the performance of being on close level, score also approaches.Man sings female's song, female sings the situations such as man's song, as long as pitch standard also can secure satisfactory grades.
7. ventilation is proofreaied and correct
According to the ventilation frequency of current sentence, score is weighted.Such as the sentence frequently of taking a breath, score suitably improves and revises.
8. degree-of-difficulty factor is proofreaied and correct
According to the degree-of-difficulty factor of current sentence, score is carried out to degree-of-difficulty factor weighting.Such as the high sentence of difficulty, score is higher.
9. every score of accumulation
The score that every frame is sung adds up in real time.
10. fast forwarding and fast rewinding scoring
If in scoring process, F.F. or rewind down are carried out.Can automatically reorientate, then repeat above step 5 ~ 9 and mark.
11. calculate every score
In the time that this sentence finishes, calculate this PTS.Singer can see the performance situation of every like this.
12. whole first final scores
When song finishes, calculate the average mark of all sentences, obtain the score of final whole song.
If a lot of sentences are all sung finely continuously, mark award can be identified and carry out to scoring algorithm automatically.Combo in similar game.
This methods of marking, by the automatic analysis of above step, improves user's experience greatly, has solved original insurmountable difficulty from the following aspects:
A marks according to pitch and rhythm.
Analyze pitch and the rhythm difference of singing between original singer, mark.Meet the standard of specialty scoring.And other singing marking systems are normally marked according to volume: the score that gives great volume is high, and be not mainly to mark according to the order of accuarcy of pitch and rhythm.
B sings well and sings badly, and score has difference.
Sing well and sing badly, score has significant difference.Other singing marking systems, because be not to mark by pitch and rhythm, so the height of score and performance level does not have proportional relation.
C score is stable.
Different songs, the performance of different people, same person are sung repeatedly etc. in various situations, as long as it is few to sing level error, score also can be similar.Mark has really embodied performance level.
D can improve score by exercise.
New hand's score is low, and after exercise a period of time, score can obviously improve.
E vernier is very sensitive, can pass through the very sensitive control vernier of voice.
Such as singing trill when, can significantly see the small size shake of vernier.This operation sense, it is very good to experience.
Other type singing system divides two kinds, and one is not analyze pitch completely, so cannot show in real time performance vernier.Another is to adopt simple pitch extracting method, poor accuracy, and scoring is inaccurate, and vernier can randomized jitter.So all cannot well control slider position by singing, experience very poor.
The present invention, on high precision pitch extraction algorithm basis, marks according to pitch and the rhythm sung, is not subject to the impact of other factors such as volume.Sing for different songs, different people, as long as performance level is identical, score is all close.The score that performance level is high is high, and low point that performance level is low is low.The present invention is in several ten thousand songs, and people up to a million have obtained good evaluation after using.