KR102381117B1 - Method of music information retrieval based on brainwave and intuitive brain-computer interface therefor - Google Patents

Abstract

The present invention relates to an EEG-based music search method and an intuitive brain-computer interface device for the same. In particular, based on EEG acquired through music imagination, the user intuitively infers the music that the user wants to search and reproduce, and outputs the music scale and lyrics. It relates to an EEG-based music search method and an intuitive brain-computer interface device for the same.

Description

EEG-based music search method and intuitive brain-computer interface device for the same

The present invention relates to an EEG-based music search method and an intuitive brain-computer interface device therefor, and more particularly, based on the EEG acquired through music imagination, intuitively infers the music that the user wants to search for and reproduce, thereby performing musical scale and It relates to an EEG-based music search method for outputting lyrics and an intuitive brain-computer interface device for the same.

Brain-wave is an electrical signal that is expressed when information is transmitted between the nervous system and cranial nerves of our body. In particular, EEG using a non-invasive method can be measured through electrodes attached to the scalp without a separate surgical operation. And it can measure the real-time activity of the brain, so it is becoming an important tool to understand the intention of the user.

The non-invasive brain-machine interface is an interface technology that controls external machines by recognizing the user's intentions through brain waves measured from the scalp. It is usually used for device control or communication for patients who are not free or paralyzed in the medical field. has been used

However, in recent years, due to the development of EEG analysis technology, the above-mentioned non-invasive brain-machine interface is being applied in various fields including the development of daily life assistance services for ordinary people.

In particular, the music imagery-based brain-computer interface is highly useful because it is possible to intuitively recognize what a user wants to search for and play music without a separate external stimulus.

Such music imagination activates the lower left frontal cortex, left temporal cortex, premotor cortex, and auxiliary motor areas of the brain, and it is possible to grasp the user's intention according to the music imagination using this.

For example, by using brain waves generated by imagining, for example, saying lyrics, humming, and listening to music without actually making a sound, the user can recognize what he or she wants to search and play.

On the other hand, music information retrieval (MIR) is a technology that extracts various information of the music, such as genre, rhythm, tempo, and chord, using only a music file without any other information.

Since recent music services are produced, distributed, and enjoyed in a digital state, the aspect of music services desired by existing consumers has naturally changed. Typical examples include streaming, music recommendation, and user-participatory listening and editing. As a result, music information retrieval is also attracting attention thanks to the quantitative increase and diversification of music services.

Nevertheless, the currently used brain-computer interface-based user intention recognition method is related to communication elements such as motion imagination (Motor Imagery, MI) of a specific part of the body or Event Related Potential (ERP) caused by external stimuli. There is a problem of having a rather non-intuitive aspect by using features that are not yet available.

Republic of Korea Patent No. 10-1633743 Republic of Korea Patent Registration No. 10-1553256

The present invention is to solve the above-mentioned problems, and based on the brain wave obtained through the music imagination and including the user's intention, the user intuitively infers the music that the user wants to search and play, and outputs it as a scale and lyrics. An object of the present invention is to provide an intuitive brain-computer interface device for a music retrieval method based on the same.

To this end, the brain wave-based music search method according to the present invention is to search for music based on the brain wave including the user's intention, and the music sample providing unit provides a music sound for inducing the brain wave including the music imaginary intention. providing a music sample; an EEG measurement step of measuring a user's music imaginary EEG generated by listening to the music sound from a measurement sensor and storing the measured EEG in a measurement data storage unit; a model learning step of learning an EEG analysis model through feature analysis of the music imaginary EEG and sound waves of a music sound in the model learning unit; a search EEG input step of measuring an EEG including a user's intention to request a music search by a measurement sensor and recognizing the user's intended music in real time; a search decision step of specifying a music imagination-related brainwave measured in the search brainwave measurement step by using the brainwave analysis model in the music generation intention recognition unit; and a music search step of searching for user intention music corresponding to the specified brain wave associated with the music imagination in the music information search unit.

Preferably, the method further includes a music reproducing step of reproducing the searched music in the music reproducing unit.

In addition, when an event-related potential occurs during music reproduction in the music reproduction step, it is preferable to return to the search determination step to reproduce music included in the candidate group among the searched pieces of music.

In addition, the music sample providing step may include: a sample storage step of storing a music sound file to be presented to a user in a music sound DB; and a sample reproducing step of generating the music sound stored in the music sound DB by the music sound generating unit and providing it to the user, wherein a plurality of candidate music sounds to be provided to the user in the sample storing step are stored, the candidate music It is preferable that the sound includes any one or more of several music sounds specified by the user, several frequently used music sounds, and several arbitrarily set music sounds.

In addition, the EEG measuring step is a user's EEG related to the sound wave of the music sound provided to the user, and at least one of the EEG generated when the user listens to the music sound and the EEG generated when the user directly sings music is measured. desirable.

In addition, the model learning step includes a pre-processing step of removing the noise of the input sound wave and brain wave and extracting a set band; a feature extraction step of extracting waveform features of the pre-processed sound waves and brain waves; a classification model learning step of classifying by features using the extracted waveform features of sound waves and brain waves; and a generating model learning step of generating a sound wave of a scale unit having a waveform similar to the brain wave at the time of imagining music.

In addition, the search determining step may include: determining whether a music imaginary brain wave intended by the user exists in the database of the learned classification model; and generating a similar waveform through the generation model when the user's intended music imaginary brain wave does not exist in the database of the learned classification model.

On the other hand, the intuitive brain-computer interface device according to the present invention is for the EEG-based music search method as described above, comprising: a music sample providing unit providing a music sound for inducing an EEG including a music imaginative intention; a measurement data storage unit for storing the user's music imaginary brain wave data generated by listening to the music sound; a model learning unit for learning an EEG analysis model through feature analysis of the music imaginary EEG and sound waves of the music sound; a music generation intention recognition unit for specifying an EEG including a user's intention to request a music search by using the EEG analysis model; and a music information search unit that searches for user intention music corresponding to the specified music imagination-related brain waves.

In this case, the music sample providing unit includes: a music sound DB for storing a music sound file to be presented to the user; and a music sound generator for generating music sounds stored in the music sound DB and providing them to a user, wherein a plurality of candidate music sounds to be provided to the user are stored in the music sound DB, wherein the candidate music sounds are selected by the user. It is preferable to include any one or more of several specified music sounds, several frequently used music sounds, and several arbitrarily set music sounds.

In addition, the model learning unit includes a pre-processing unit for removing the noise of the input sound wave and brain wave and extracting a set band; a feature extraction unit for extracting waveform features of the preprocessed sound waves and brain waves; and a learning unit for learning a classification model that classifies each feature by using the extracted sound waves and waveform features of brain waves and a generation model that generates sound waves in a scale unit having a waveform similar to the brain waves when imagining music. it is preferable

As described above, the present invention provides an intuitive brain-computer interface device and method capable of transmitting a user's intention only by thinking by recognizing a music imagination using brain waves and searching for and playing music intended by the user.

In addition, high reliability and accuracy can be obtained because only the relevant brain waves that are expressed when imagining music without extra external stimuli are used. there is.

In addition, since the user intention recognition result can be output in the form of sound, it is possible to provide a neural entertainment brain-computer interface system that can be utilized in various fields with a high degree of freedom.

1 is a block diagram showing a brain-computer interface device for EEG-based music search according to the present invention.
2 is a block diagram illustrating a music sample providing unit according to the present invention.
3 is a block diagram showing a model learning unit of the present invention.
4 is a flowchart illustrating an EEG-based music search method according to the present invention.
5 is a block diagram illustrating a measurement target in the step of providing a music sample according to the present invention.

Hereinafter, an EEG-based music search method and an intuitive brain-computer interface device therefor according to a preferred embodiment of the present invention will be described in detail with reference to the accompanying drawings.

First, prior to the description of the brain wave-based music search method according to the present invention, an intuitive brain-computer interface device as a preferred embodiment for implementing it will be described.

The present invention provides an intuitive brain-computer interface for searching and playing music based on brain waves including the user's intention. In addition, based on the brain waves acquired through music imagination, the music that the user wants to search and play is intuitively inferred and output as scales and lyrics.

To this end, as shown in FIG. 1 , the intuitive brain-computer interface device 100 according to the present invention includes a music sample providing unit 110 , a measurement data storage unit 120 , a model learning unit 130 , and a music generation intention recognition unit ( 140) and a music information search unit 150.

In another embodiment, it further includes a music reproducing unit 160 , and the music sample providing unit 110 includes a music sound DB 111 and a music sound generating unit 112 . The model learning unit 130 includes a preprocessing unit 131 , a feature extraction unit 132 , and a learning unit 133 for a classification model and a generation model.

Here, the music sample providing unit 110 provides the music sound so that the user can induce brain waves including the intention of imagining music. In order to build the database, the user may include other third parties in addition to the search requester using brain waves.

As will be described again below, the present invention learns brain waves including the user's intentions, classifies them, and generates sound waves in scale units. use brain waves.

In addition, by using the brain waves when listening to music or singing music, these brain waves are compared with the sound waves of the music sound, the brain waves correspond to the sound waves, and decoded to generate effective data similar to the sound waves.

Accordingly, the music sound provided by the music sample providing unit 110 induces brain waves corresponding thereto, and furthermore, it is directly connected with the meaningful user's intention to recall or search for music, so that it can be learned.

More specifically, as shown in FIG. 2 , the music sample providing unit 110 generates a music sound DB (DataBase, 111) for storing a music sound file to be presented to the user, and a music sound stored in the music sound DB 111 for the user. and a music sound generating unit 112 provided to the user.

In particular, a plurality of candidate music sounds to be provided to the user are stored in the music sound DB 111 . As the candidate music sound, a plurality of music sounds designated by a user, several frequently used music sounds, and a plurality of arbitrarily set music sounds may be stored.

The measurement data storage unit 120 stores the user's music imagination brain wave data generated by listening to the music sound. To this end, the measurement data storage unit 120 is connected to the brain wave measuring device 121 such as an electrode sensor attached to the scalp of the human body and stores the signal-processed brain wave data.

The brain wave data stored in the measurement data storage unit 120 may be stored together with the sound wave of the music sound provided by the music sample providing unit 110 , and the stored brain wave data is related to or recorded as an EEG corresponding to the sound wave.

Therefore, along with the user's music imaginary brain wave, the sound wave of the music sound provided to the user and the brain wave when listening to the sound are measured and converted into a digital signal to build a database. However, when measuring the EEG, the EEG when the user directly calls it can also be measured, converted into a digital signal, and then stored.

The model learning unit 130 learns the brain wave analysis model through feature analysis of a music imaginary brain wave and a sound wave of a music sound. Therefore, it is possible to read whether the brain wave generated by the user is a brain wave generated by imagining a music search.

A user music creation intention classification model, which will be described later, is learned through the characteristic analysis of the acquired sound waves and brain waves. That is, the user's music creation intention classification model is learned based on the extraction of features from preprocessed sound waves and brain waves.

To this end, the model learning unit 130 includes the preprocessing unit 131, the feature extracting unit 132 and the classification model to generate the user's word generation intention classification model through the feature analysis of the acquired EEG and sound waves as shown in FIG. and a learning unit 133 for generative model learning.

Among them, the preprocessor 131 removes noise of the input sound wave and brain wave and extracts a set band. In order to remove noise from sound waves and EEG, signals other than the band containing important information are filtered, and only EEG of a significant band is extracted.

The feature extractor 132 extracts the waveform features of preprocessed sound waves and brain waves, and generates music (searching) through feature extraction techniques such as common space pattern (CSP), wavelet transform (Wavelet Transform), and principal component analysis (PCA). ) Extracts features related to sound waves of music sound from brain waves including intention.

The classification model learning unit of the learning unit 133 classifies for each characteristic by using the waveform characteristics of the extracted sound waves and brain waves. For example, the classifier is learned through classification methods such as linear discriminant analysis (LDA), random forest, and artificial neural network (ANN). The learned classifier makes it possible to classify which music is searched for when searching for music.

The generation model learning unit of the learning unit 133 generates a sound wave in a scale unit having a waveform similar to the brain wave when imagining music. That is, a similar waveform is generated through the correlation between the music imaginary brainwave and the envelope of the brainwave generated when listening to or singing music.

Similar waveforms generated in this way are replaced or converted by brainwaves into waveforms corresponding to sound waves of music sound, and the present invention allows music to be searched for and reproduced by using user intentional brainwave information converted similarly to sound waves.

The music generation intention recognizing unit 140 uses an EEG analysis model to specify an EEG including a user's intention to request a music search, and recognizes lyrics and scales from the EEG measured through the EEG analysis model.

Lyrics and scales recognized from EEG can be recognized by converting the EEG waveforms measured as described above into similar waveforms corresponding to sound waves and extracting contrastable lyrics and scales from them.

In this case, in the music generation intention classification model, the intention to create a word is detected through the classification model learned from not only the EEG provided and stored for the purpose of learning, but also the newly measured music imaginary EEG for music search.

As a result, if the user's intended music imagination does not exist in the classification model database, that is, if the lyrics and scale recognized from the brain waves do not exist in the classification model, the model cannot be applied. A similar waveform is created through the search, and the generated similar waveform is searched.

The music information search unit 150 searches for user intention music corresponding to the specified brainwave related to the musical imagination, and searches for music corresponding thereto using the lyrics and scales recognized by specifying the brainwave as described above.

In other words, music imaginary brain waves are classified through a classification model, and based on this, the user's intention is identified, and music is searched in the DB (DataBase). The DB includes not only the DB provided in itself, but also the DB provided in an external third server.

The music searched as above is reproduced by the music reproducing unit 160 . To this end, the music player 160 may include hardware for outputting the searched music, and outputs and provides music intended by the user through the music imagination brain wave.

In addition, in the music output, the music generation may include the music output generated by the music selection unit and the output of the music generated through the generator based on the analyzed music imaginary brain wave.

Hereinafter, an EEG-based music search method according to a preferred embodiment of the present invention will be described in detail with reference to the accompanying drawings.

The present invention infers the corresponding sound waveform from the EEG waveform that is generated when the user imagines speaking lyrics, humming, and listening to music.

In addition, the completed music is output by classifying and reconstructing the scale and lyrics generated from the brain waves. Therefore, it is possible to search and play music with only thoughts by decoding the brain waves including the user's intention.

To this end, the brain wave-based music search method according to the present invention includes a music sample providing step (S110), an EEG measurement step (S120), a model learning step (S130), a search EEG input step (S140), a search decision step (S150) and It includes a music search step (S160), and preferably further includes a music playback stage (S180) that determines whether it is suitable for the user's intention (S170) and plays music.

In this case, in the music sample providing step ( S110 ), a music sound for inducing brain waves including the intention of imagining music is provided. The music sound is provided by the music sample providing unit 110 of a brain-computer interface device implemented in a search terminal or a PC.

EEG measurement induced by the sound of music learns the EEG including the user's intention, classifies it, and generates sound waves in scale units.

Accordingly, as shown in FIG. 5 as an embodiment, the present invention measures EEG when imagining music with lyrics, humming, and listening to music.

In addition, the brain waves when listening to music or singing music are measured and stored together with the brain waves of imagining described above and the sound waves of music sounds. However, if it is used for patients who find it difficult to sing aloud, EEG at the time of direct speech can be selectively applied.

Accordingly, the music sound provided by the music sample providing unit 110 induces brain waves corresponding thereto. In particular, it is directly related to a meaningful user's intention to recall or search for music, so that it can be learned.

Specifically, the music sample providing step ( S110 ) includes a sample storage step of storing a music sound file to be presented to a user in the music sound DB 111 , and the music stored in the music sound DB 111 in the music sound generating unit 112 . and a sample playback step of generating a sound and providing it to a user.

In the sample storage step, a plurality of candidate music sounds to be provided to the user are stored. The candidate music sounds may include several music sounds specified by the user, several frequently used music sounds, and several randomly set music sounds.

Preferably, at least any one or more of several music sounds designated by the user, several frequently used music sounds, and several arbitrarily set music sounds may be provided, thereby inducing brain waves.

Next, in the EEG measurement step (S120), the measurement sensor 121 measures the user's music imaginary EEG generated by listening to the music and stores it in the measurement data storage unit 120 to build a database.

In order to measure the user's EEG induced by the provided music sound, the user first listens to the music while wearing the EEG measurement device and measures the EEG related to the music imagination containing the user's intention.

The brain waves related to the above music imagination are stored in the measurement data storage unit 120 as corresponding to the music imagination brain waves of the user generated by listening to the music sound.

The brain wave data stored in the measurement data storage unit 120 is stored together with the sound wave of the music sound provided by the music sample providing unit 110 , and the stored brain wave data is related to or recorded as an EEG corresponding to the sound wave.

Therefore, along with the user's music imaginary brain wave, the sound wave of the music sound provided to the user and the brain wave when listening to the sound are measured and converted into a digital signal to build a database. In this case, the EEG when the user directly calls may also be measured, converted into a digital signal, and then stored.

Next, in the model learning step (S130), the brain wave analysis model is learned through feature analysis of the music imaginary brain wave and the sound wave of the music sound. Learning is performed in the model learning unit 130 of the brain-computer interface device. Therefore, it is possible to read whether the brain wave generated by the user is a brain wave generated by imagining a music search.

The user's music generation intention classification model, which will be described later, is learned through the feature analysis of the acquired sound waves and brain waves, and features are extracted from the preprocessed sound waves and brain waves, and based on this, the user's music creation intention classification model is learned.

To this end, the model learning step (S130) can be implemented through the pre-processing step, the feature extraction step, the classification model learning step, and the generating model learning step.

In the pre-processing step, the noise of the input sound wave and brain wave is removed and the set band is extracted. That is, signals other than the band containing important information are filtered and only the EEG of the meaningful band is extracted. As an example, frequency filtering is performed in a band of 0.5 to 50 Hz in order to minimize the influence of noise caused by ambient noise or DC power.

In the feature extraction step, waveform features of preprocessed sound waves and brain waves are extracted. For feature extraction, features related to sound waves of music sound are extracted from brain waves including the intention to create (search) music through feature extraction techniques such as common spatial pattern, wavelet transformation, and principal component analysis.

In the classification model learning step, classification by features is performed using the extracted sound wave and EEG waveform features. For example, a classifier is learned through a classification method such as a linear discriminant analysis method, a random forest, and an artificial neural network. The learned classifier makes it possible to classify which music is searched for when searching for music.

In the generative model learning stage, a sound wave of a scale unit having a waveform similar to the brain wave when imagining music is generated. For example, the envelope of the brain wave generated when listening to music or singing music is extracted and a similar waveform is generated through a correlation between them.

Similar waveforms generated in this way are replaced or converted by brainwaves into waveforms corresponding to sound waves of music sound, and the present invention allows music to be searched for and reproduced by using user intentional brainwave information converted similarly to sound waves.

Next, in the search EEG input step ( S140 ), the EEG including the user's intention to request the music search is measured with the measurement sensor 121 , and the user's intended music is recognized. Accordingly, it is possible to grasp the intention of creating music, which will be described later.

To this end, the brain waves appearing when the user imagines the intended music sound are measured in real time, and then, the user's intended music is recognized based on the learned decoding model (ie, the analysis model).

Next, in the search determination step (S150), the brainwave related to the musical imagination measured in the search brainwave measurement step (S120) is specified using the EEG analysis model. The specification is performed in the music generation intention recognition unit 140 of the brain-computer interface device, and a search target is extracted from the brain waves.

EEG is specified by classifying EEG according to a classification model built through learning or generating sound waves in scale units of similar waveforms corresponding to EEGs through a generation module. That is, EEG specification is achieved through classification and/or generation.

Accordingly, the music generation intention recognizing unit 140 uses the EEG analysis model to specify an EEG including the user's intention to request a music search, and recognizes lyrics and scales from the EEG measured through the EEG analysis model.

In this step, the music that is presumably intended by the user is first classified by the learned classification model. At this time, it is determined whether the music intended by the user exists in the database of the learned classification model (S151), and if it does not exist as a result, music having a similar waveform from the user's intended music imaginary brainwave through the learned sound wave generation model generated (S152).

Lyrics and scales recognized from EEG can be recognized by converting the EEG waveforms measured as described above into similar waveforms corresponding to sound waves and extracting contrastable lyrics and scales from them.

To this end, the search decision step (S150) is a step of determining whether the user's intended music imaginary EEG exists in the database of the learned classification model, and when it does not exist in the database of the learned classification model, generating a similar waveform through the generation model. It includes steps.

Next, in the music search step ( S160 ), the music information search unit 150 having a search function searches for user intention music corresponding to the brain wave related to the music imagination specified above. For example, the search may be performed through a music-related application or a large-capacity music DB.

This music search step ( S160 ) is executed by the music information search unit 150 . The music information search unit 150 searches for music corresponding to the specified EEG by using the recognized lyrics and scale in order to search for the user's intended music corresponding to the specified EEG related to the specified musical imagination.

In other words, music imaginary brain waves are classified through a model, and based on this, the user's intention is identified and music is searched in the DB. The DB includes not only the DB provided in itself, but also the DB provided in an external third server.

Next, in the music reproduction step ( S180 ), the music searched for through the music reproduction unit 160 is reproduced. To this end, the music player 160 may include hardware for outputting the searched music, and outputs what the user intends through the music imagination brain wave.\

The music reproduction step (S180) may be performed after determining whether the search result is suitable for the user's intention (S170), and the music generation is generated through a generator based on the music output generated from the music selection unit and the analyzed music imaginary brain wave It may include the output of the music.

However, if an error-related potential occurs during music reproduction in the music reproduction step S180, it is preferable to return to the search determination step S150 and reproduce the music included in the candidate group among the searched music. .

In this case, the corresponding music being output to the user is replaced with the music found as a candidate. The music replacement is changed to the music with the second highest probability except for the music selected as the intended music by the user in the classification model.

At this time, based on the Bayes probability, the most suitable music is modified by considering the relationship between the front and rear scales and lyrics during the natural language processing process. Through such user feedback, correction and supplementation of music output is performed.

However, the step of correcting the music using the error-related induced potential is not an essential component of the present invention and may be selectively applied. After that, the final music completes the brain-computer interface through the search and playback stages.

In the above, specific embodiments of the present invention have been described above. However, the spirit and scope of the present invention is not limited to these specific embodiments, and various modifications and variations can be made within the scope that does not change the gist of the present invention. You will understand when you grow up.

Therefore, since the embodiments described above are provided to fully inform those of ordinary skill in the art to which the present invention belongs the scope of the invention, it should be understood that they are exemplary in all respects and not limiting, The invention is only defined by the scope of the claims.

110: music sample providing unit
111: music sound DB
112: music sound generation unit
120: measurement data storage unit
130: model learning unit
131: preprocessor
132: feature extraction unit
133: learning unit (classification model learning unit, generative model learning unit)
140: music generation intention recognition unit
150: music information search unit
160: music playback unit

Claims (10)

In the brain wave-based music search method for searching music based on the brain wave including the user's intention,
A music sample providing step (S110) of providing a music sound for inducing an EEG including a music imaginary intention in the music sample providing unit 110;
an EEG measurement step (S120) of measuring the user's imaginary music brainwave generated by listening to the music sound in the measurement sensor 121 and storing it in the measurement data storage unit 120;
a model learning step (S130) of learning the brain wave analysis model through the feature analysis of the music imaginary brain wave and the sound wave of the music sound in the model learning unit 130;
a search EEG input step (S140) of measuring an EEG including the user's intention to request a music search by the measurement sensor 121, and recognizing the user's intended music in real time;
a search decision step (S150) of specifying the brainwave related to the music imagination measured in the search brainwave measurement step (S120) by using the brainwave analysis model in the music generation intention recognition unit 140; and
and a music search step (S160) of searching for user intended music corresponding to the specified brainwave related to the musical imagination in the music information search unit 150;
Further comprising a music playing step (S180) of playing the searched music in the music playback unit 160,
In the music reproduction step (S180), when an error-related potential occurs during music reproduction, the process returns to the search decision step (S150) to reproduce music included in the candidate group among the searched music EEG based music search method. delete delete According to claim 1,
The music sample providing step (S110),
a sample storage step of storing the music sound file to be presented to the user in the music sound DB (111); and
A sample playback step of generating the music sound stored in the music sound DB 111 in the music sound generating unit 112 and providing it to the user; including,
Storing a plurality of candidate music sounds to be provided to the user in the sample storage step, wherein the candidate music sounds include any one or more of several music sounds designated by the user, several frequently used music sounds, and several arbitrarily set music sounds. An EEG-based music retrieval method. According to claim 1,
The EEG measurement step (S120) is,
EEG-based music search, characterized in that at least one of an EEG generated when a user listens to a music sound and an EEG generated when the user directly sings music is measured as the user's brainwave related to the sound wave of the music sound provided to the user method. According to claim 1,
The model learning step (S130) is,
a pre-processing step of removing noise from input sound waves and brain waves and extracting a set band;
a feature extraction step of extracting waveform features of the pre-processed sound waves and brain waves;
a classification model learning step of classifying by features using the extracted waveform features of sound waves and brain waves); and
A brain wave-based music search method comprising: a generating model learning step of generating a sound wave in a scale unit having a waveform similar to the brain wave at the time of imagining music. 7. The method of claim 6,
The search decision step (S150) is,
determining whether the user's intended music imaginary brain wave exists in the database of the learned classification model; and
and generating a similar waveform through the generation model when the user's intended music imaginary brainwave does not exist in the database of the learned classification model. a music sample providing unit 110 that provides a music sound for inducing an EEG including a music imaginative intention;
a measurement data storage unit 120 for storing the user's music imaginary brain wave data generated by listening to the music sound;
a model learning unit 130 for learning the brain wave analysis model through the feature analysis of the music imaginary brain wave and the sound wave of the music sound;
a music generation intention recognition unit 140 for specifying an EEG including a user's intention to request a music search by using the EEG analysis model; and
and a music information search unit 150 that searches for user intention music corresponding to the specified music imagination-related brainwave.
Further comprising a music playback unit 160 for playing the searched music,
The music reproducing unit 160 is an intuitive brain-computer interface device, characterized in that when an error-related potential (Event-related Potential) occurs during music reproduction, the music included in the candidate group among the searched music. 9. The method of claim 8,
The music sample providing unit 110,
a music sound DB (111) for storing a music sound file to be presented to a user; and
A music sound generating unit 112 that generates the music sound stored in the music sound DB 111 and provides it to the user; including;
The music sound DB 111 stores a plurality of candidate music sounds to be provided to the user, and the candidate music sounds include any one or more of several music sounds designated by the user, several frequently used music sounds, and several arbitrarily set music sounds. Intuitive brain-computer interface device, characterized in that it comprises. 9. The method of claim 8,
The model learning unit 130,
a pre-processing unit 131 for removing noise of input sound waves and brain waves and extracting a set band;
a feature extraction unit 132 for extracting waveform features of the pre-processed sound waves and brain waves; and
A learning unit 133 for learning a classification model that classifies by features by using the extracted sound waves and waveform features of brain waves and a generation model that generates sound waves in a scale unit having a waveform similar to the brain waves when imagining music; Intuitive brain-computer interface device, characterized in that it comprises.

Images