TW559782B - Real-time music composition method - Google Patents

Abstract

The present invention uses the computer to real-time compose the music, which first records some basic word sound, which is the sound database of the system; after entering the score and lyrics for a song by the user, the music synthesizer produced according to the present invention can real-time synthesize the music. The major functions of the synthesizer include: arbitrarily rising or falling the tone, adjusting the volume, and scaling the tone lengths.

Description

559782 A7

V. Description of the invention (2) N Day-to-day fluctuations L Zanming overview j The instant singing voice synthesis method of the present invention is specified by the user as arbitrary- 首 i # 使 _ ”4⑽ can be verbal, systematic person + * Qu Zhi Ge Pu Gang Guang : The first day is synthesized into a singing voice (this synthesizing unit is a clean male or female voice); ⑵ user village borrows it: the voice data file is used as the synthesizing unit (this refers to 4n): 'The system can automatically According to the words recorded by the user;; = Compose the song sheet and lyrics of the input song, and complete the real-time singing voice :: system :: based on the original voice data _ 'Computer can stand alone to match the line with the mother-paragraph The words of the lyrics: After the input of two or eight hundred songs, the system instantly converts this into a singing voice. This function allows the user to record a complete 4I i mandarin monosyllable. The main purpose of the present invention is' healthy supply — a computer automatic Combined into a Japanese singing voice: Use the voice data file in the system database and cooperate with the ^ song's song score and lyrics' to make the computer synthesize the ^ and lyrics entered by the user into a singing voice in real time. Purpose, to enable users to translate their own voice Although the data or the voice data file of a certain singer is built in the database database of the system, when you use ^ to enter the score and lyrics of the song, you can use the computer to calculate: Really say a song with personal flavor , Whether it is a folk song or a popular song; or the user's voice and the singing voice of a certain singer by the present invention can be adjusted to the paper size ^^ i ^ CNSMoiiTilO X 297 public fl 559782 A7 B7 V. Description of the invention (

I Staff Consumer Printing Co., Ltd. printed a beautiful chorus together. Another aspect of the present invention is the simultaneous use of the voice of the eye and the synthesis of the singing voice: the present invention can be used to synthesize, synthesize, advance, and sing the sound of the song. Only scheng completed a song synthesized by a computer and used it for the karaoke mainly for coffee broadcasts. Κ software "Song contains Wei Da = ::, the present invention provides a method for the instant synthesis of a song." It is to make the user input the setting type · Tuan Er = Lang and store the input voice as electronic data. (.S is to use the silent or silent sound of the front and rear sections of the recording sound to test the truth.) The position of the starting point and ending point of the sound. · The two-position program uses the aforementioned endpoint detection program: the search for the basic cycle of the sound, and the location of each starting point of the sound technique; the singing synthesizer program , Department of Erkang = The basic cycle of sound extraction and labeling procedures, to get the correct basic cycle position and size of the sound of each word, and then according to the score and lyrics reversed by using ϋ, the Each, adjust the length, pitch, Volume, vibrato, echo, connection sound ;;:; ^ and song score input program 'After the aforementioned song synthesizer program is completed, the user can synthesize the entered lyrics and song score through computer calculations-a song ; The result output program 'plays immediately based on the results produced by the aforementioned program' and saves it as an electronic data format building, so that the paper size of the item sheet is adapted to the Chinese national standard (CNSM4 specification (210 X 297 public love 559782 A7) B7 V. Description of the invention (Printed by the Ministry of Intellectual Property Employees' Consumer Corporation, users can obtain more [main component symbol comparison instructions], 1 works and fun 11 voice data file recording procedures through this singing voice real-time synthesis system] 2 —_ endpoint彳 Chast test program 13 --- sound basic period extraction and labeling program 14 --- singing voice synthesizer program 15 lyrics and score input 16 result output program 3 1 --- sound source signal 32 --- range value of sound basic period 33-- -The range value of the basic period of the search sound 34 —- the position of the sampling point 3 5 —- the maximum value of the local section 36 the pitch [detailed description of the invention] Please read the formula containing the preferred embodiment of the present invention for a full description, but before this description, it should be understood that those familiar with the bank can modify the invention described in this article and obtain the effect of the invention at the same time. The description is a broad disclosure for those who are familiar with the skills of the bank, and its content is not intended to limit the present invention. Please refer to Fig. 1, which shows the architecture diagram of the instant singing voice synthesis system of the present invention. The present invention is- A method for enabling users to obtain instant singing voice by inputting lyrics and scores of songs. This method includes: a recording program for voice data broadcasting u: enables users to load a single syllable of input-language by input, and The input voice is stored as an electronic data box.

Printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs: (point her = use the silent period (Sllence) at the front and back of the recorded sound to measure the position of the origin and end of the sound. The basic cycle of sound extraction and labeling procedures 13 : Use the sound file data obtained by the aforementioned 洌 program I2 to perform sound beauty, and find, and mark the starting point of each basic sound cycle thereafter; Huer 2 :: Find = Synthesizer Program 14 •• According The above-mentioned sound base spot "Private Sequence 13", get the ^ = and big :, 'of each single-voice sound of each stroke of speech, and then synthesize according to the song score and song selected by the user; Base: :: individual characters in the include, including Adjust the characteristics of pitch, vibrato, echo, connection sound, etc .; in the sequence: and: = into the program 15: when the aforementioned song synthesizer 14 passes, it can synthesize a song in real time; and 电 the calculation result output program 16: Played according to the aforementioned program and stored as an electronic data cell π. For further brilliance, please continue to refer to FIG. 1. Please come to show a voice endpoint detection diagram of the present invention and the present invention; two = two since : Closed algorithm flow chart. The present invention can make use of A song from Baizaki, which is synthesized by a computer and played back. Two: six: system program 11, endpoint detection program 12, base earphone private sequence 13, song synthesizer program 14 t program 15 and result output program 16 Different; step = voice data ㈣_program u: before synthesizing the singing voice. ΦΜ --------- 丨 -order --------- (Please read the note on the back? Fill in the matter before filling in (This page). ^ / 82 ^ / 82 A7 B7 V. Description of the invention (6 ^ Private sequence 'needs to be established-a database of sound files, this data can be used, the sound file of the sigh = or the user records himself Sound program A,, f = Japanese branch, when recording sound files other than the sound file preset by the system, in terms of the pronunciation of :: 国 ί 吾 早 字音, the various combinations of the pronunciation are about 411 ΡΓΜ The second effect can be Monophonic or stereo sound effects, the format of the stored electronic data can be, "shame" files or other electronic files of sound effects; and the method used in this procedure is the energy value and '. 凊 Continue to Figure 2 As shown in Figure 1, the main method is to use two critical values (τ η ::::: The beginning and end of a single phonetic syllable;,% sequence is also enough to separate the consonant and final vowel ^ Generally speaking, the initial consonant can distinguish between voiced and non-voweled vowels. · Low zero-crossing rate (low ergy) is considered as sound. · Knife M (Log View: None) Zero !:-When a high 'sound decibel is low, 3 is considered as sound; The main reason for the Zoufen ::: voice and unvoiced part in the * rotation (―) is that the final singing part and the basic adjustment of the sound day are only for the sound part, and the part without reason is independent. Adjust for the basic sound cycle. ", Ear Basic Period Extraction and Marking Procedure 13: Please refer to the figure below for a description of a basic periodical map of speech and sound of the present invention, and this paper is suitable for China National Standard (CNS) A4 Specifications (210 xlW Male i "

--------------------- Order ----- (Please read the notes on the back before filling out this page) -III-559782 A7 V. Description of the invention ( 7) Refer to Figure 3 together. The basic period of sound is extracted using an autocorrelation algorithm. The autocorrelation algorithm, first, when the audio source signal 31 is input, it presents a sine wave signal as shown in Figure 4. From this sine wave signal, the basic period of the sound can be obtained. The range of the maximum and minimum values of 32 is 32. Therefore, the maximum and minimum values of the basic period of the sound can be used as a search range value 33 for processing. For example, the maximum value of the basic period of the sound is 200, and the minimum value is 100. If the position 34 of a certain sampling point is found to be 125 through the maximum value method, then the left and right sides should be searched for the next basic period of sound. The position should be between 1350 and U50. Find a local area in this section. The position of the segment maximum 35 (Local maximum), that is, the position of a basic period of the sound; using the aforementioned autocorrelation algorithm, iteratively repeated, that is, as shown in Figure 4, the position of the circle represents the basic sound The position of the cycle. Singing synthesizer program 14 · Please refer to Fig. 5, which shows the intention of the method of synchronizing and accumulating the basic period of sound in the speech synthesizer of the present invention. After pre-processing of the basic sound cycle extraction and labeling program, Dian Yue Hung can use this information to arbitrarily adjust the melody required in the song score; in adjusting the pitch parameters, the basic sound cycle is used to synchronize and accumulate (Pitch Synchronous 〇verLap and Add (ps〇LA) technology, this technology can make flexible adjustments to the basic period of sound and leaves different, more efficient, as shown in Figure 5, the basic period of sound is the original γ times 鬲 ' Then the distance between the original sound basic period marks is reduced to 1/2, otherwise, if it is the original 1/2 pitch, the distance between the sound basic period marks is doubled, and every time The basic sound of the two voices is to use a Hamming window to reshape the consumption of the employees of the Bureau of Intellectual Property Bureau. M 559782

Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs. 5. Description of the invention (8) A new sound source waveform, and then the waveform added by the Hamming window is added up in an overlapping manner to form a new waveform. The Hamming window formula is:

W㈣ = m〇s [j ^, Q < m < N Refer to FIG. 6 for a schematic diagram of adjusting the sound length of the present invention. For the adjustment of the sound length, a linear mapping method is used; each synthesis unit has its own sound length, which is desirable. The score of f will have beat information for each sound. The number of seconds can be calculated according to the multi-beat support. The length of one beat is calculated as follows: the number of seconds of a beat (t) x the sampling frequency (f) = one beat In terms of general speed, one beat (ΐ) is about 0.66 seconds. If the sampling rate is set to 2〇ΚΗζ⑴, then the length 2 of a beat is about 0.66 X 200. 00 = 132. The number of points, when the song score is input, and compared with the synthetic single reading in the shell database, you can know that the sound length needs to be expanded or contracted. When the aforementioned sound length determines the money, and then apply the principle of linear mapping to adjust to get the final Sound length. "In addition to the long change of θ, sometimes it is necessary to change the size in time. Generally speaking, the intro of popular songs is usually low, and the volume of the chorus part is higher, and it must be controlled. The degree of change in the overall volume cannot be changed suddenly. The unit for measuring the volume is decibel (dB), and its formula is:, l〇 * l〇gl〇 (l ^^ (/)) The quality of the synthesized singing voice is similar to that of the real singing voice. The present invention uses ------------- 11111.11 ^ --- I I--ΙΛ (Please read the note on the back before filling in this note page)

559782 A7 B7

Printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs

…, Explanations of Jin and echo. The advantage of connecting sounds is to avoid synthesizing more real s-images. In the implementation system of the present invention, it is taken from (Cross-Fading), and needs to be emphasized; 乂 and the addition should be based on the characteristics of each sound But the size, in principle, can be used as a connecting sound, there is a 4b single character Ziribe injury wound eight # said the old consonant part is an important special emblem injury, the right consonant part as the connection sound, it will destroy it For example, most of the silent consonants, such as 勹, female,: ,,, and shell, such as 丄, Θ, worm, p, Liu Xun, are not consonant.凊 Refer to Fig. 7, which shows the result output diagram of the present invention. The interpretation of the talented singers—the song usually trembles in the treble or bass, this is the vibrato = system database also designed this function, when the song score is entered into the system ㈣ library 'Please refer to the part indicated by the system index in Figure 7. This can depict the high and low parts of the song score. When the song reaches the high and low, the line database can actively follow the melody of the syllable at this time ^ The phenomenon of a vibrato makes the computer-synthesized singing sound more humane. As for the echo (Ech.) Part, a finite impulse response chirper (― 仏 I_Se Response Filter '_' Response) is a limited interval, that is, there is a non-disaster solution only in a limited time domain. After the waveform is shaped by a finite impulse response filter, it is synthesized into two results like KTV singing. 1.> Lyrics and song score input program 15: After the foregoing procedure is completed, the user enters lyrics and song scores, and a song can be synthesized through computer calculations. The lyrics input procedure can be a text file, and then Advance

This paper size applies to China National Standard (CNS) A4 (210 X 297 public love)

Please read the precautions on the back before filling out this page): --Order ---------

5597¾ Five A7 invention description (1.) Yuan Yuanhe: also singing voice: it can also be recorded by the user. In addition, it can also allow users to control each song for different songs and use the aforementioned program to synthesize singing voices in real time. Worker, mother, show, turn-out program 16: After being processed by the aforementioned program, the computer sends the result to ^ "and moon dagger_broadcast or other sound effects block." # The storage cost is prepared in another embodiment. Using the aforementioned voices of male and female users to record in the computer database, the method will sing the content and synthesize the deduction of male and female shouts in accordance with the user-defined performance command. Completion song 'If the song has a chorus part, the song method is completed. In addition to the above, the present invention uses the "song voice to synthesize x objects in real time, except for the implementation of the fan, sometimes m will inevitably have voice and Singing the sound, make the computer use the voice in 1 text, or i input the score in the text part, and then the computer can synthesize the singing voice; when making the second song, the computer can use it System database :: articles = more humane performance. After Bian and early people are familiar with the preferred embodiment, they will be familiar with the technology and understand the following. Please make the following changes: make various changes and changes, and the present "" = ί = implementation of the embodiment, in which the sound is basically _Take method can benefit; mean size difference function (八 州 响 Ｍ Coffee tUde Difference

Function) or using non-time domain methods, such as: to find the size of the base cycle. The aforementioned song synthesizer program, and the method of adjusting the basic period of the sound may be: the resampling method in the time domain

This paper size applies to China National Standard (CNS) A4 specification (210x 297 public love T ------. 丨 丨 Order ----------- / Please read the precautions on the back before filling this page > Printed by the Employees' Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs 559782 A7 B7 V. Adjustment method on the invention description (Domain); and the basic period of the residual error signal is synchronized and re-accumulated (Residual signal with ps〇LA) chords Wave sampling (SinUSOidal) and the like belong to the adjustment method in the frequency domain (Frequency D⑽ain). [Invention effect] According to the instant singing voice synthesis system and method implemented according to the present invention, the original voice synthesis system can only advance To achieve the synthesis of singing voice, when the user uses the instant singing voice synthesis system * method of the present invention, it is necessary to enter the song into the computer and then enter the computer into the computer, which can be used for brain calculations, or a moving song, or By making the princes stay, Li You, you use the test to record a basic early child of 0, and sing through the input of the cricket and score-a song that belongs to your own singing. --------- Order --- ----- (Please read the notes on the back before filling This page) Ministry of Economic Affairs Intellectual Property Office employees consumer cooperatives printed

559782 A7 V. Description of Invention ()

[Simplified description of the diagram J =-This is the instant voice synthesis system of the present invention. Figure 1 is a voice endpoint detection diagram of the present invention: a diagram. "Figure 3 is the white flow chart of the basic period of the sound of the present invention. V is taken from the phase algorithm. Γ is a basic figure of a speech sound of the present invention. Β Figure 5 is the basic sound of the speech synthesizer of the present invention. Schematic diagram of the method of stacking and accumulating j-terms. Figure 6 is the schematic diagram of adjusting the sound length of the present invention. Figure 7 is the output diagram of the result output procedure of the present invention. Paper size applies to China National Standard (CNS) A4 (210 X 297 public love)

Claims (1)

559782, patent application scope of Jibei Smart Employees Consumer Press 1. A method for real-time synthesis of singing voices, including: type; early ... and save the input voice as the end point of the electronic data grid_programs ·· 系 彻 录 声 ㈣ 后 # Measure the position of the starting and ending points of sounds during the sound period;…, ear or basic period of sound extraction and labeling procedures: The sound m obtained by the measurement procedure is m; Where the starting point is marked = find, the AHE program: it is based on the basic cycle of sound mentioned above & bit private order 'Paid to the mother-pen voice of the single-word sound as per-word = The "song score and lyrics" of the lyrics f song score input program: when the aforementioned song synthesizer program is completed = the user synthesizes the lyrics and song scores entered over time, and puts it on the back. The result output program: The results produced according to the aforementioned procedure are immediately 2. As the instant singing voice input device described in the item of the scope of the patent application can be a microphone or a keyboard. 彳 32 二; ^: 1 method of instant singing voice synthesis History ... It can be said that the input can be a variety of Mandarin monophone characters, and the input characters, Wave format files, or other electronic data formats can be entered. 4. The singing voice as described in the item in the scope of the patent application is in real time. T-shirts are averaged by using a different function or a field method to set the line. The size of the paper rule is * Guo ^ Voting standard ⑵ Specifications 〇〇 Norwegian public hair 559782 6. The scope of patent application 5 Gan Rushen. Please sing the song mentioned in item 1 of the scope of the patent in real time. The mid-point detection procedure can be used to detect the position of the starting point. The method is to determine whether the detection is wrong. Manual inspection to 6 · If the scope of patent application! The method of real-time singing voice described in the above item is a method in which the endpoint detection program can separate the initials or finals. 7. In the song real-time device program described in item i of the patent application scope, the composition of the lyrics and text is called tone, volume, Features such as vibrato, echo, or connection. i said long, 8 · such as the scope of patent application! The real-time and second-song of the song mentioned in the item: Synthesizer program, which mainly adjusts the square of the basic period of the sound. The second is: the resampling method, the sound of the residual error signal is basically heavy :: addition or sine wave sampling and other methods. Less heavy and rich 9. The singing voice as described in the scope of the patent application, which is the sound output file of the result output program, can be -or: he :: 10 · The singing voice as described in the scope of the patent application, i. The output program can be an electronic data file. Method, 11 · The singing voice described in item 1 of the scope of Chinese Patent No. 4 captures the same day. In the lyrics and score input program, the lyrics input part can be input by voice.疋 Youyu or I2 · — A method for real-time synthesis of voice and singing voice, including ... A voice depends on _ Procedure: The person who borrows the costumes = a single syllable of clean speech, and stores the input voice as electronic Information: Endpoint system program ... Using the silence of the front and rear sections of the sound recording floor (Please read the precautions on the back before filling this page) -------------- Printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs Paper size_CNS Standard A4 (297)