JP3567123B2 - Singing scoring system using lyrics characters - Google Patents

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a singing scoring method in a karaoke apparatus, and more specifically, to a singing scoring method capable of appropriately evaluating a singing ability flexibly in response to a vocal utterance timing.
[0002]
[Prior art]
As a supplementary function of the karaoke apparatus, a singing scoring function is well known. In a typical singing scoring method, a pitch uttered by a singer is extracted by sampling a singer's voice signal input from a microphone. Using the main melody part of the karaoke accompaniment music in the karaoke data as the criterion for scoring, the correct pitch data at the correct timing to emit each note obtained from the score data of the main melody part and the pitch uttered by the singer at the correct timing Compare one by one. Then, a score according to the comparison result is given for each appropriate scoring section. When the main melody part to be sung ends, the scores are totaled to calculate a singing score.
[0003]
[Problems to be solved by the invention]
The goal is for karaoke beginners to sing along the score in order to sing well, but if they stay faithful to the score forever, they will only be able to expect a stylized musical expression. A person who has a singing ability sometimes intentionally delays the utterance timing by one tempo to use the ad lib. In addition, the model singing that is used as a criterion for singing a karaoke device is merely a singing at one point in the past, and even if the same singer sings, the vocal timing often changes with the times.
[0004]
However, in the conventional singing scoring method, the vocal pitch of the singer is compared with the correct pitch data of the karaoke data corresponding to the timing one by one. For this reason, if the singing timing of a certain phrase deviates even a little, even if the transition of the utterance pitch is not different from the transition of the correct pitch, the utterance pitch at the correct timing is different from the correct pitch, so the score will be greatly reduced Is done. However, in reality, if the change in pitch is accurate, even if there is a slight shift in the utterance timing, there is no musical inferiority, and in some cases, it can be felt as an excellent singing expression. Still, if you can only get a low score, it is hard to say that the correct scoring was done.
[0005]
Accordingly, an object of the present invention is to provide a singing scoring method capable of appropriately evaluating a singing ability from a musical point of view by evaluating a transition of an utterance pitch flexibly in response to a singer's utterance timing.
[0006]
[Means for Solving the Problems]
A singing scoring device using lyric characters according to the present invention is specified by the following items (1) to (10).
(1) A singing and scoring device including accompaniment music performance means, voice recognition means, frequency analysis means, and scoring control means. (2) The accompaniment music performance means plays accompaniment music based on karaoke data. (3) The karaoke data includes lyric characters corresponding to the note data of the main melody part. (4) The voice recognition means outputs recognition characters that are converted into characters by recognizing phonemes of voice signals from the microphone ( 5) The frequency analysis means analyzes the frequency of the audio signal from the microphone and outputs pitch data. (6) The scoring control means performs the association processing, the identification processing, the score calculation processing, and the score total processing. (7) In the associating process, the pitch data output from the frequency analysis unit is stored in association with the recognition character output from the voice recognition unit. (8) The identification process includes The correspondence between the recognized character and the note data of the main melody part is specified by specifying the correspondence between the lyric character and the note data. The pitch data corresponding to the recognized character and the pitch data of the note data are compared, and a pitch matching score representing the degree of similarity between the two is calculated. (10) The score comprehensive processing calculates a pitch matching score for each recognized character. Calculating and outputting a total score based on the total score
The present invention can be carried out on the basis of the above-mentioned items, with the following items (11) to (13) added.
(11) In the associating process, the duration is associated with the recognized character output by the voice recognition means and stored. (12) In the score calculation process, in the note data in which the correspondence between the recognized character and the corresponding character is specified, The duration corresponding to the recognized character is compared with the note length of the note data, and a note length matching score representing the degree of similarity between the two is calculated. Calculating and outputting a total score based on the long matching score;
BEST MODE FOR CARRYING OUT THE INVENTION
=== Basic Functions of Karaoke Device ===
FIG. 1 shows a configuration example of a karaoke apparatus employing a singing score system according to an embodiment of the present invention. A central control unit 1 including a CPU, a RAM, and a ROM controls and controls each peripheral component via a control bus.
[0010]
The hard disk drive 2 stores a large number of karaoke data including accompaniment music of karaoke music and lyric character data used for singing score. When receiving the request via the operation control unit 3, the central control unit 1 processes the corresponding karaoke data in the order of inputting the request. That is, the karaoke accompaniment music is retrieved by extracting the karaoke data corresponding to the requested music from the hard disk drive 2 and transferring the music score data of the included karaoke accompaniment music part to the synthesizer 5 via a MIDI (Musical Instrument Digital Interface) signal processing unit 4. Is generated. The generated karaoke accompaniment music and the singing voice collected from the microphone 6 are mixed by the mixing amplifier 7 and output to the speaker 8. The display 12 displays a background image input via the image control unit 13 and a lyric image with a color change of a portion to be sung.
[0011]
In the karaoke apparatus, the singing voice sung along with the karaoke music and collected from the microphone 6 is processed by the voice recognition processing unit 9 and the frequency analysis processing unit 10 and temporarily stored. This is transferred to the scoring unit 11 and compared with the correct answer data of the karaoke data and scored.
[0012]
=== Karaoke data structure ===
Karaoke data used in the karaoke apparatus according to the present invention conforms to the MIDI standard. That is, the minimum unit representing a pitch is a semitone (= 100 cents), and the minimum unit representing a note length is quantized data of a 32nd note. The karaoke data includes the musical score data of the karaoke accompaniment music part for generating the karaoke accompaniment music generated by the synthesizer 5, the musical score data of the main melody part to be sung, and each note data of the main melody part. Lyric character data. FIG. 2 shows an example of a main melody part with lyric character data in karaoke data used in the karaoke apparatus according to the present invention.
[0013]
=== Speech recognition means ===
The singing voice signal collected from the microphone 6 is input to the voice recognition processing unit 9 and the frequency analysis processing unit 10 of the karaoke apparatus. First, when the speech recognition processing unit 9 recognizes a certain pronunciation (hereinafter referred to as a first pronunciation), it converts the character into a character and writes it into a memory. At the same time, the frequency analysis processing unit 10 starts measuring the pitch of the sound. When recognizing a different pronunciation (the second pronunciation), it calculates the average value of the pitch in the first pronunciation interval from the start of the pitch measurement to that point, and corresponds to the recognized character written in the memory Then, the average pitch value of the first sounding section is written. At the same time, when the second pronunciation is recognized, it is converted into a character and written into the memory. Further, the pitch average value of the second sounding section is calculated and written in the memory in the same manner as in the case of the first sounding. Each time a different pronunciation is recognized, the process of characterizing and calculating the average pitch is repeated, and a table of average pitch data (in units of one cent) for the recognized character string is created in the memory.
[0014]
Simultaneously with the start of the pitch measurement, the speech recognition processing section 9 starts a timer for detecting the duration of sound generation, and starts measuring the time during which the first sound generation continues. In this case, when the second pronunciation is recognized, the timer is stopped, and the timer value is written in association with the recognized character written in the memory and the pitch average value. Thus, the table is extended.
[0015]
Further, the frequency analysis processing unit 10 monitors whether or not the sound generation is stopped. At regular intervals, for example, it is checked whether or not the karaoke data is generated at intervals corresponding to the 32nd note, which is the minimum unit of the note length, and when it is recognized that the utterance has stopped, a bit is added to the silence flag in the memory. Stand up. Until the next pronunciation is recognized, in this example, the silent flag is kept set at a time corresponding to the 32nd note.
[0016]
=== Identification of sung lyrics characters ===
In order to score a singing, it is necessary to first associate a lyric character to be sung included in the karaoke data with a recognized character string in the memory. When the recognized character is written to the memory by the above-described method, the scoring unit 11 compares the lyric character associated with the score data with the recognized character in the memory, and determines which lyric character of the score data corresponds to the recognized character. Is identified. Once identified, the correct pitch data at which the lyric character is to be pronounced is extracted from the karaoke data and added to the memory in association with the corresponding recognized character.
[0017]
When the pronunciation duration of the singing character is measured, the length of the note associated with the lyric character is also extracted from the karaoke data, and is added to the memory in association with the corresponding recognized character.
[0018]
=== Pitch matching score and note length matching score ===
The scoring unit 11 refers to the table written in the memory, compares the average pitch value (in units of 1 cent) and the correct pitch data (in units of 100 cents) for each recognized character, and calculates the degree of coincidence (error). I do. For example, when the difference is one cent, a rule such as minus one point is determined in advance, and the matching degree of all the recognized character strings in the marking section is integrated to obtain a pitch matching score.
[0019]
Also, the pronunciation duration and the length of the note for each recognized character are compared, and the degree of coincidence is calculated. For example, a rule such that a time corresponding to a 32nd note is set as an evaluation unit and an error of one unit time is set at minus one point is determined in advance, and the matching degrees of all recognized character strings in the marking section are integrated. Score length matching score.
[0020]
=== Process based on rest of main melody part ===
When the rest timing of the music data of the main melody part of the karaoke data is longer than a certain length, for example, one beat or more, the scoring unit 11 checks whether a bit is set in the silence flag. During the rest, the confirmation of the silence flag is continued at intervals equal to the interval of the silence flag (in this example, at intervals corresponding to the 32nd note), and the degree of coincidence of the rest timing is calculated. For example, if there is no bit of the silence flag that should have been set during the rest of one beat, a rule such as minus one point is determined in advance, and all the rests to be scored are set. The rest matching score is obtained by summing the degrees of coincidence for.
[0021]
=== Comprehensive singing score ===
The total score obtained by integrating the above-described pitch matching score, note length matching score, and rest matching score is used as the score for singing. The manner of integration differs depending on where the score is focused on pitch, note length, and timing. For example, when priority is given in the order of pitch> note length> timing, as an example, it is conceivable to apply a calculation formula of total score = 3 × pitch matching score + 2 × note length matching score + rest rest matching score. The total score calculated here is expressed in an appropriate format and displayed on the display 12, for example.
[0022]
【The invention's effect】
ADVANTAGE OF THE INVENTION According to the karaoke apparatus concerning this invention, it becomes possible to obtain the appropriate singing score result corresponding also to the singers' own singing timing by evaluating the pitch of the singing separately from the singing timing. On top of that, by taking into account the evaluation of the pronunciation duration for each pronunciation character and the evaluation of the silent section of the singing corresponding to the rest, the person who actually listens to the song feels It is possible to calculate a musical evaluation value corresponding to "the goodness of the song".
[Brief description of the drawings]
FIG. 1 is a configuration example of a karaoke apparatus according to an embodiment of the present invention.
FIG. 2 is an example of a main melody part with lyrics characters in karaoke data used in the karaoke apparatus according to one embodiment of the present invention.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 1 Central control unit 2 Hard disk drive 3 Operation control unit 4 MIDI signal processing unit 5 Synthesizer 6 Microphone 7 Mixing amplifier 8 Speaker 9 Voice recognition processing unit 10 Frequency analysis processing unit 11 Scoring unit 12 Display 13 Video control unit

Claims (2)

Accompaniment music performance means, voice recognition means, frequency analysis means, a singing scoring device provided with scoring control means,
The accompaniment music playing means plays the accompaniment music based on the karaoke data,
The karaoke data contains lyrics characters corresponding to the note data of the main melody part,
The voice recognition means recognizes the phoneme of the voice signal from the microphone and outputs a character recognition character,
The frequency analysis means analyzes the frequency of the audio signal from the microphone and outputs pitch data,
The scoring control unit performs an association process, an identification process, a score calculation process, and a score total process,
The associating process associates and stores pitch data output by the frequency analysis unit with a recognition character output by the voice recognition unit,
The identification processing specifies the correspondence between the recognized character and the note data of the main melody part by specifying the correspondence between the recognized character and the lyrics character,
In the score calculation process, the pitch data of the recognized character and the note data in which the correspondence relationship is specified are compared with the pitch data of the note data, and a pitch matching score representing the degree of approximation between the two is calculated. Calculate,
The score totaling process is a singing scoring device that calculates and outputs a total score based on a pitch matching score for each recognized character. The associating process associates and stores the duration of the recognized character output by the voice recognition means,
In the score calculation processing, in the note data in which the recognition character and the correspondence relationship are specified, the duration corresponding to the recognition character is compared with the note length of the note data, and a note length matching score representing the degree of approximation of both is calculated. And
The singing scorer according to claim 1, wherein the score totaling process calculates and outputs a total score based on a pitch matching score and a note length matching score for each recognized character.

Images