Voice separation device and method for seperating voice signals based on speaker

Abstract

A voice separation device is disclosed. The voice separation device includes a voice signal receiving circuit configured to receive voice signals, a trigger signal generating circuit configured to generate a speaker registration trigger signal, a processor configured to receive voice signals and identify the speaker of the voice signals using the voice signals, and a memory. The processor, in response to the speaker registration trigger signal, generates speaker location information indicating the location of the speaker using the voice signals, generates a speaker identifier for identifying the speaker using the voice signals, stores the speaker location information in memory as reference speaker location information, stores the speaker identifier in memory as reference speaker identifier, and the reference speaker location information and the reference speaker identifier are matched and stored with each other.

Description

Technology Field

[0001] Embodiments of the present invention relate to a voice separation device and a voice separation method for separating voice signals according to the speaker. Background Technology

[0002] A microphone is a device that recognizes speech and converts the recognized speech into an electrical signal called a voice signal. When a microphone is placed in a space where multiple speakers are present, such as a conference room or classroom, the microphone receives all the speeches from the multiple speakers and generates voice signals corresponding to their voices. Therefore, when multiple speakers speak simultaneously, it is necessary to separate their voice signals. Furthermore, it is necessary to identify which speaker each separated voice signal is from. Prior art literature

[0003] Korean Published Patent Application No. 10-2016-0091725 (Aug. 3, 2016) The problem to be solved

[0004] The problem that the present invention aims to solve is to provide a voice separation device and method capable of identifying the location of a speaker using a plurality of input voice signals and separating and recognizing the plurality of voice signals by speaker.

[0005] The problem that the present invention aims to solve is to provide a voice separation device and method capable of separating and recognizing a plurality of input voice signals by speaker, and storing the plurality of voice signals by matching them with a corresponding speaker identifier. means of solving the problem

[0006] A voice separation device according to embodiments of the present invention comprises a voice signal receiving circuit configured to receive voice signals, a trigger signal generating circuit configured to generate a speaker registration trigger signal, a processor configured to receive voice signals and identify the speaker of the voice signals using the voice signals, and a memory. The processor, in response to the speaker registration trigger signal, generates speaker location information indicating the location of the speaker using the voice signals, generates a speaker identifier for identifying the speaker using the voice signals, stores the speaker location information in memory as reference speaker location information, stores the speaker identifier in memory as reference speaker identifier, and the reference speaker location information and the reference speaker identifier are matched and stored with each other.

[0007] A voice separation method for separating voice signals according to embodiments of the present invention comprises the steps of: generating a speaker registration trigger signal; receiving voice signals; generating speaker location information indicating the location of a speaker using the voice signals in response to the speaker registration trigger signal; generating a speaker identifier for identifying a speaker using the voice signals; storing the speaker location information as reference speaker location information; and storing the speaker identifier as a reference speaker identifier, wherein the reference speaker location information and the reference speaker identifier are matched and stored with each other. Effects of the invention

[0008] A voice separation device according to embodiments of the present invention can identify the location of a speaker using a voice signal and distinguish which speaker's voice the voice signal corresponds to through the location of the speaker. Accordingly, even if multiple speakers speak simultaneously, the voice separation device has the effect of separating the voices by speaker. Brief explanation of the drawing

[0009] FIG. 1 shows a voice separation system according to embodiments of the present invention. FIG. 2 shows a voice separation device according to embodiments of the present invention. FIG. 3 is a flowchart for explaining the speaker registration mode of a voice separation device according to embodiments of the present invention. FIGS. 4 to 6 are drawings for explaining the speaker registration mode of a voice separation device according to embodiments of the present invention. FIG. 7 is a flowchart for explaining the voice separation mode of a voice separation device according to embodiments of the present invention. FIGS. 8 to 10 are drawings for explaining the speaker registration mode of a voice separation device according to embodiments of the present invention. FIG. 11 is a flowchart for explaining a speaker movement mode according to embodiments of the present invention. FIGS. 12 and FIGS. 13 are drawings for explaining a speaker movement mode according to embodiments of the present invention. FIG. 14 is a diagram illustrating the operation of a voice separation device according to embodiments of the present invention. Specific details for implementing the invention

[0010] Hereinafter, embodiments of the present invention will be described with reference to the attached drawings.

[0011] FIG. 1 illustrates a voice separation system according to embodiments of the present invention. Referring to FIG. 1, the voice separation system (10) may include a plurality of microphones (100) and a voice separation device (200).

[0012] Multiple microphones (100) receive the voice of speakers (SPK1~SPK4) located at each position (P1~P4) and can convert the voice of the speakers (SPK1~SPK4) into voice signals (VS1~VSn) which are electrical signals.

[0013] A plurality of microphones (100) can output voice signals (VS1 to VSn). According to embodiments, a plurality of microphones (100) can provide voice signals (VS1 to VSn) to a voice separation device (200). For example, a plurality of microphones (100) can transmit voice signals (VS1 to VSn) to a voice separation device (200) according to a wired or wireless method.

[0014] A plurality of microphones (100) are configured as beamforming microphones, so that voice can be received from multiple directions. According to embodiments, a plurality of microphones (100) may be spaced apart from each other to form a single microphone array, but embodiments of the present invention are not limited thereto.

[0015] Each of the plurality of microphones (100) may be a directional microphone configured to receive sound from any specific direction, or an omnidirectional microphone configured to receive sound from all directions.

[0016] The voice separation device (200) receives voice signals (VS1~VSn) transmitted from a plurality of microphones (100) and can separate the voice signals (VS1~VSn) by speaker.

[0017] The voice separation device (200) may be a computing device having a computational processing function. According to embodiments, the voice separation device (200) may be implemented as a computer, laptop, mobile device, smartphone, or wearable device, but is not limited thereto. For example, the voice separation device (200) may include at least one integrated circuit having a computational processing function.

[0018] The voice separation device (200) can determine the location of the sound source (i.e., the location of the speaker) using voice signals (VS1~VSn) and separate each of the voice signals (VS1~VSn) by speaker according to the determination result. For example, the voice separation device (200) can distinguish and separate voice signals corresponding to the voice of the first speaker at the first location and voice signals corresponding to the voice of the second speaker at the second location.

[0019] The voice separation device (200) can generate a speaker identifier to identify each speaker from the voice signals (VS1~VSn). The voice separation device (200) can store the voice signals (VS1~VSn) separated by speaker by matching them with the corresponding speaker identifier.

[0020] In this specification, the statement that a voice signal and a speaker identifier (SID) are stored in a matched manner may mean that the speaker identifier can be searched (or read) using the voice signal, or that the voice signal can be searched (or read) using the speaker identifier.

[0021] The voice separation device (200) can process voice signals (VS1 to VSn). According to embodiments, the voice separation device (200) can convert the voice signals (VS1 to VSn) into text data and store the converted text data by matching it with a speaker identifier. Additionally, the voice separation device (200) can convert text data of a first language into text data of a second language.

[0022] According to embodiments, the voice separation system (10) according to embodiments of the present invention may further include a server (300), and the voice separation device (200) transmits voice signals (VS1~VSn) to the server (300), the server (300) converts the voice signals (VS1~VSn) into text data, transmits the converted text data to the voice separation device (200), and the voice separation device (200) may store the converted text data by matching it with a speaker identifier. Additionally, the server (300) may convert text data of a first language into text data of a second language and transmit it to the voice separation device (200).

[0023] According to embodiments, the voice separation system (10) according to embodiments of the present invention may further include a speaker (400). The speaker (400) may receive a voice signal and convert the received voice signal into voice and output it. According to embodiments, the speaker (400) may output voice into space by generating vibration in response to the voice signal. The voice separation device (200) may convert text data into a voice signal and transmit the converted voice signal to the speaker (400). The speaker (400) may output voice corresponding to the text data using the transmitted voice signal.

[0024] FIG. 2 illustrates a voice separation device according to embodiments of the present invention. Referring to FIG. 2, the voice separation device (200) may include a voice receiving circuit (210), a trigger signal generating circuit (220), a processor (230), and a memory (240).

[0025] The voice receiving circuit (210) can receive voice signals (VS1~VSn) transmitted from a plurality of microphones (100). According to embodiments, the voice receiving circuit (210) includes an analog-to-digital (ADC) converter, converts analog voice signals (VS1~VSn) into digital voice signals, and outputs the converted voice signals. However, for convenience of explanation, in FIG. 2, the reference numerals for the analog voice signals (VS1~VSn) and the digital voice signals are shown as identical.

[0026] According to embodiments, the voice receiving circuit (210) includes an input port that is wired to each of a plurality of microphones (100), and can receive voice signals (VS1~VSn) through the input port.

[0027] According to embodiments, the voice receiving circuit (210) includes a communication circuit capable of communicating with each of the plurality of microphones (100), and can receive voice signals (VS1~VSn) through the communication circuit.

[0028] The trigger signal generation circuit (220) can generate trigger signals (TRS1 and TRS2) in response to an external input. The trigger signals (TRS1 and TRS2) may be signals that cause the processor (230) to perform a specific action. The trigger signals (TRS1 and TRS2) may include a speaker registration trigger signal (TRS1) and a speaker movement trigger signal (TRS2). In this case, the input conditions for generating the speaker registration trigger signal (TRS1) and the speaker movement trigger signal (TRS2) may be different.

[0029] According to embodiments, the trigger signal generating circuit (220) includes an input unit capable of detecting physical input from the outside, such as a touch panel or a button, and can generate trigger signals (TRS1 and TRS2) in response to the physical input. For example, the trigger signal generating circuit (220) can generate trigger signals (TRS1 and TRS2) when a user's touch is detected.

[0030] According to embodiments, the trigger signal generation circuit (220) may receive voice signals (VS1 to VSn) from the voice signal receiving circuit (210) and recognize a trigger word included in the voice signals (VS1 to VSn) to generate trigger signals (TRS1 and TRS2). For example, the trigger signal generation circuit (220) may generate trigger signals (TRS1 and TRS2) when it receives voice signals (VS1 to VSn) containing a specific phrase such as "speaker registration".

[0031] The trigger signal generation circuit (220) can transmit the generated trigger signals (TRS1 and TRS2) to the processor (230).

[0032] The processor (230) receives voice signals (VS1 to VSn) and can process the voice signals (VS1 to VSn). According to embodiments, the processor (230) may be implemented as an integrated circuit having computational processing capabilities. For example, the processor (230) may include a CPU (central processing unit), MCU (micro controller unit), DSP (digital signal processor), GPU (graphics processing unit), ASIC (application specific integrated circuit), or FPGA (field programmable gate array), but embodiments of the present invention are not limited thereto.

[0033] The processor (230) can determine the source location (i.e., speaker location) of the voice signals (VS1~VSn) using the voice signals (VS1~VSn). According to embodiments, the processor (230) can determine the source location of the voice signals (VS1~VSn) based on the magnitude of the voice signals (VS1~VSn), the phase difference between the voice signals (VS1~VSn), and at least one of the locations of the plurality of microphones (100).

[0034] The processor (230) can generate speaker location information (SPI) corresponding to the determined sound source location. That is, since the voice signals (VS1~VSn) correspond to voice output from the speaker, the sound source location of the voice signals (VS1~VSn) can represent the location of the speaker.

[0035] The processor (230) can generate a speaker identifier (SID) to identify the speaker using voice signals (VS1~VSn).

[0036] The processor (230) can match speaker location information (SPI) and speaker identifier (SID) and store them in memory (240).

[0037] In this specification, the statement that speaker location information and speaker identifier (SID) are stored in a matched manner may mean that the speaker identifier can be searched (or read) using the speaker location information, or that the speaker location information can be searched (or read) using the speaker identifier.

[0038] The memory (240) may include at least one of non-volatile memory and volatile memory.

[0039] According to embodiments, the voice separation system (10) may further include a voice signal output circuit (250). The voice signal output circuit (250) may output voice signals (VS1 to VSn).

[0040] According to embodiments, the voice signal output circuit (250) includes a digital-to-analog (DAC) converter, converts digital type voice signals (VS1 to VSn) into an analog type, and outputs analog type voice signals (VS1 to VSn) to an external audio device.

[0041] According to embodiments, the voice signal output circuit (250) includes a communication circuit and can transmit voice signals (VS1 to VSn) to a server (300).

[0042] FIG. 3 is a flowchart for explaining the speaker registration mode of a voice separation device according to embodiments of the present invention. The operation of the voice separation device (200) to be described with reference to FIG. 3 can be implemented with instructions stored in a computer-readable storage medium.

[0043] The speaker registration mode is a mode for registering and storing each speaker's speaker identifier and speaker location information in an electronic device (200) using each speaker's voice.

[0044] Referring to FIG. 3, the trigger signal generation circuit (210) can generate a speaker registration trigger signal (TRS1) (S110). According to embodiments, the trigger signal generation circuit (210) can detect a trigger word included in physical inputs or voice signals (VS1~VSn) from the outside and generate a speaker registration trigger signal (TRS1) based on the detection result.

[0045] The processor (230) may enter a speaker registration mode in response to a speaker identification trigger signal (TRS1) (S120). According to embodiments, the speaker registration mode may be defined as a predetermined interval from the time when the speaker registration trigger signal (TRS1) is received or as a interval while the speaker registration trigger signal (TRS1) is received, but is not limited thereto.

[0046] In speaker registration mode, the processor (230) can determine the source location (i.e., speaker location) of the voice signals (VS1~VSn) using the voice signals (VS1~VSn) and generate speaker location information (SPI) corresponding to the determined source location (S130).

[0047] The processor (230) can generate a speaker identifier (SID) to identify a speaker using voice signals (VS1~VSn) (S140). The speaker identifier (SID) is intended to distinguish a speaker and may represent the speaker's name, but is not limited thereto.

[0048] According to embodiments, the processor (230) can determine whether the voice signals (VS1 to VSn) contain a pre-specified speaker identification phrase and generate a speaker identifier (SID) from the speaker identification phrase. For example, the processor (230) can generate a speaker identifier (SID) corresponding to "Alice" from a speaker identification phrase such as "I am Alice."

[0049] According to embodiments, the processor (230) recognizes an identification start phrase and an identification end phrase included in voice signals (VS1 to VSn), and can recognize the speaker identification phrase based on the identification start phrase and the identification end phrase. For example, the processor (230) recognizes an identification start phrase such as "I am" and an identification end phrase such as "is," and can recognize the phrase from "I am" to "is" as a speaker identification phrase.

[0050] The processor (230) can match speaker location information (SPI) and speaker identifier (SID) and store them in memory (240) (S150). The speaker location information (SPI) stored in memory (240) can be referred to as reference speaker location information, and the speaker identifier (SID) stored in memory (240) can be referred to as reference speaker identifier.

[0051] The processor (230) can store a speaker identifier (SID) and speaker location information (SPI) in association. According to embodiments, multiple speaker location information (SPI) may be matched and stored for a single speaker identifier (SID), and the multiple speaker location information (SPI) matched with a single speaker identifier (SID) may represent a range of the speaker's location.

[0052] That is, the voice separation device (200) according to the embodiments of the present invention can store a speaker identifier (SID), such as the speaker's name, in association with the speaker's location, so that the speaker of the voice signal (VS1~VSn) can be identified by determining the speaker's location of the voice signal (VS1~VSn).

[0053] FIGS. 4 to 6 are drawings for explaining the speaker registration mode of a voice separation device according to embodiments of the present invention.

[0054] In speaker registration mode, the voice separation device (200) receives voice signals (VS1~VSn), generates speaker location information (SPI) and speaker identifier (SID) using the voice signals (VS1~VSn), and can store the speaker location information (SPI) and speaker identifier (SID) by matching them.

[0055] Referring to FIG. 4, when Alice utters the voice “I am Alice,” voice signals (VS1 to VSn) corresponding to “I am Alice” are transmitted from the microphone (100) to the voice separation device (200). The voice separation device (200) can use the voice signals (VS1 to VSn) to generate speaker location information (SPI) indicating Alice’s location (P1) and generate a speaker identifier (SID) indicating Alice’s name (Alice).

[0056] The voice separation device (200) can store speaker location information (SPI) and speaker identifier (SID) by matching them. For example, Alice's location (P1) and Alice's name (Alice) can be matched and stored in the voice separation device (200). That is, the speaker for location "P1" can be registered as "Alice".

[0057] Referring to FIG. 5, similar to FIG. 4, when Bob utters the voice “I am Bob,” the voice separation device (200) can generate speaker location information (SPI) indicating Bob’s location (P2) from voice signals (VS1~VSn) corresponding to “I am Bob,” generate a speaker identifier (SID) indicating Bob’s name (Bob), and store the speaker location information (SPI) and the speaker identifier (SID) by matching them. For example, Bob’s location (P2) and Bob’s name (Bob) can be matched and stored in the voice separation device (200). That is, the speaker for location “P2” can be registered as “Bob.”

[0058] Referring to FIG. 6, the voice separation device (200) can store speaker location information (SPI) and speaker identifier (SID) for Dave by matching them. For example, Dave's location (P4) and Dave's name (Dave) can be matched and stored in the voice separation device (200). That is, the speaker for location "P4" can be registered as "Dave".

[0059] A voice separation device (200) according to embodiments of the present invention can store a speaker identifier (SID) and speaker location information (SPI) in association based on voice signals (VS1~VSn).

[0060] FIG. 7 is a flowchart for explaining the voice separation mode of a voice separation device according to embodiments of the present invention. The operation of the voice separation device (200) to be described with reference to FIG. 7 can be implemented with instructions stored in a computer-readable storage medium.

[0061] Referring to FIG. 7, the processor (230) can enter a voice separation mode (S210). According to embodiments, the processor (230) can enter a voice separation mode after the speaker registration mode is terminated. The voice separation mode is a mode for separating voice signals (VS1~VSn). The processor (230) can separate the voice signals (VS1~VSn) using a pre-stored speaker identifier (SID) and speaker location information (SPI).

[0062] The processor (230) can determine the source location (i.e., speaker location) of the voice signals (VS1~VSn) using the voice signals (VS1~VSn) and generate speaker location information (SPI) (S220).

[0063] The processor (230) can determine a speaker identifier (SID) corresponding to the generated speaker location information (SPI) (S230). According to embodiments, the processor (230) can identify the speaker of the voice signals (VS1~VSn) using the generated speaker location information (SPI) and the speaker location information stored in memory (240) (i.e., reference speaker location information).

[0064] According to embodiments, the processor (230) can determine reference speaker location information corresponding to the speaker location information (SPI) of the voice signals (VS1 to VSn) and determine a speaker identifier corresponding to the reference speaker location information. For example, the processor (230) can determine a speaker identifier (SID) corresponding to the speaker location information (SPI) generated from the voice signals (VS1 to VSn) by referring to the memory (240), and can identify the speaker of the voice signals (VS1 to VSn) through the determined speaker identifier (SID).

[0065] According to embodiments, the processor (230) determines a speaker identifier (SID) corresponding to speaker location information adjacent to the generated speaker location information (SPI) among the speaker location information stored in memory (240), and can separate voice signals (VS1~VSn) based on the speaker identifier (SID).

[0066] Accordingly, the voice separation device (200) according to the embodiments of the present invention has the effect of determining which of the registered speakers each of the voice signals (VS1~VSn) corresponds to by determining a speaker identifier (SID) corresponding to the received voice signals (VS1~VSn).

[0067] The processor (230) can match and store voice signals (VS1~VSn) with a determined speaker identifier (SID) (S240).

[0068] In this specification, the statement that a voice signal and a speaker identifier (SID) are stored in a matched manner may mean that the speaker identifier can be searched (or read) using the voice signal, or that the voice signal can be searched (or read) using the speaker identifier.

[0069] According to embodiments, the processor (230) may associate voice signals (VS1 to VSn) with a speaker identifier (SID) and output them to a voice signal output circuit (250), and the voice signal output circuit (250) may output the voice signals (VS1 to VSn) and the associated speaker identifier (SID) to an external device such as a speaker or display. Accordingly, the speaker or display may output the voice of each speaker visually or audibly.

[0070] FIGS. 8 to 10 are drawings for explaining the speaker registration mode of a voice separation device according to embodiments of the present invention. In the examples of FIGS. 8 to 10, it is assumed that a speaker identifier (SID) and speaker location information (SPI) corresponding to each speaker are stored in advance in memory (240).

[0071] In voice separation mode, the voice separation device (200) receives voice signals (VS1~VSn), determines a speaker identifier (SID) corresponding to the voice signals (VS1~VSn), and can store the voice signals (VS1~VSn) and the corresponding speaker identifier (SID) by matching them.

[0072] Referring to FIG. 8, when a pre-registered speaker Alice utters the voice “aaa,” voice signals (VS1 to VSn) corresponding to “aaa” are transmitted from the microphone (100) to the voice separation device (200). The voice separation device (200) receives the voice signals (VS1 to VSn) corresponding to “aaa” and can store the voice signals (VS1 to VSn) by matching them with Alice’s speaker identifier (SID) “Alice”.

[0073] According to embodiments, the voice separation device (200) may determine speaker location information (SPI) from voice signals (VS1 to VSn), determine reference speaker location information (e.g., P1) corresponding to the determined speaker location information (SPI), and determine a speaker identifier (SID) (e.g., Alice) corresponding to the determined reference speaker location information (P1) as the speaker identifier (SID) corresponding to the voice signals (VS1 to VSn).

[0074] For example, the voice separation device (200) can receive voice signals (VS1~VSn) corresponding to "aaa" and store the voice signals (VS1~VSn) by matching them with Alice's speaker identifier (SID) "Alice". The voice separation device (200) can additionally store the time (e.g., t1) at which the voice signals (VS1~VSn) corresponding to "aaa" were received by matching them. That is, the speaker for the voice "aaa" can be identified as "Alice".

[0075] Referring to FIG. 9, when a pre-registered speaker Dave utters the voice “ddd,” voice signals (VS1 to VSn) corresponding to “ddd” are transmitted from the microphone (100) to the voice separation device (200). The voice separation device (200) receives the voice signals (VS1 to VSn) corresponding to “ddd” and can store the voice signals (VS1 to VSn) by matching them with Dave’s speaker identifier (SID) “Dave”.

[0076] For example, the voice separation device (200) receives voice signals (VS1~VSn) corresponding to "ddd" and can store the voice signals (VS1~VSn) by matching them with Alice's speaker identifier (SID) "Dave". That is, the speaker identifier (SID) for the voice "ddd" can be identified as "Dave".

[0077] The voice separation device (200) can additionally match and store the time (e.g., t2) when voice signals (VS1~VSn) corresponding to "ddd" are received.

[0078] Referring to FIG. 10, when a pre-registered speaker Bob utters the sound "bbb" and Charlie utters the sound "ccc", the voice signals (VS1~VSn) corresponding to "bbb" and "ccc" are transmitted together from the microphone (100) to the voice separation device (200).

[0079] The voice separation device (200) receives voice signals (VS1~VSn) and can determine speaker location information (SPI) of the voice signals (VS1~VSn). The voice separation device (200) can read a speaker identifier (SID) corresponding to each of the voice signals (VS1~VSn) from the memory (240) using the determined speaker location information (SPI). The voice separation device (200) can match and store the speaker identifier (SID) corresponding to the voice signals (VS1~VSn).

[0080] For example, the voice separation device (200) can determine the speaker identifier (SID) as "Charlie" in the case of voice signals (VS1~VSn) where the speaker location information (SPI) corresponds to location (P3), and can determine the speaker identifier (SID) as "Bob" in the case of voice signals where the speaker location information (SPI) corresponds to location (P2). That is, the speaker identifier (SID) for the voice "bbb" can be identified as "Bob", and the speaker identifier (SID) for the voice "ccc" can be identified as "Charlie".

[0081] After identification, the electronic device (200) can store voice signals corresponding to the voice “bbb” by matching them with Bob’s speaker identifier (SID) “Bob”, and store voice signals corresponding to the voice “ccc” by matching them with Bob’s speaker identifier (SID) “Charlie”.

[0082] That is, according to the voice separation device of the embodiments of the present invention, even if two or more speakers utter voices, there is an effect of determining the location of the voice signal and determining which speaker's voice each voice signal corresponds to.

[0083] FIG. 11 is a flowchart for illustrating a speaker movement mode according to embodiments of the present invention. The operation of the voice separation device (200) to be described with reference to FIG. 11 can be implemented with instructions stored in a computer-readable storage medium.

[0084] Speaker movement mode is a mode for updating existing registered speaker location information to changed speaker location information when the location of a registered speaker changes.

[0085] Referring to FIG. 11, the trigger signal generating circuit (210) can generate a speaker movement trigger signal (TRS2) (S310). According to embodiments, the trigger signal generating circuit (210) can detect a trigger word included in physical inputs or voice signals (VS1 to VSn) from the outside and generate a speaker movement trigger signal (TRS2) based on the detection result.

[0086] The processor (230) may enter a speaker movement mode in response to a speaker movement trigger signal (TRS2) (S320). According to embodiments, the speaker registration mode may be defined as a predetermined interval from the time when the speaker movement trigger signal (TRS2) is received or as an interval while the speaker movement trigger signal (TRS2) is received, but is not limited thereto.

[0087] In speaker movement mode, the processor (230) can generate speaker location information (SPI) and speaker identifier (SID) of the voice signals (VS1~VSn) using the voice signals (VS1~VSn) (S330).

[0088] The processor (230) can update the existing speaker location information (SPI) matched to the speaker identifier (SID) of the voice signals (VS1~VSn) (S340). According to embodiments, the processor (230) can refer to the memory (240) to determine the speaker identifier (SID) corresponding to the speaker location information (SPI) generated from the voice signals (VS1~VSn) among the speaker identifiers (SID) (i.e., reference speaker identifiers) stored in the memory (240). The processor (230) can update the speaker location information (i.e., reference speaker location information) that is matched to the speaker identifier (SID) and stored in the memory (240) with the generated speaker location information (SPI).

[0089] Accordingly, the voice separation device (200) according to the embodiments of the present invention has the effect of being able to update the speaker location information stored by matching it to the speaker identifier to the changed speaker location information even if the speaker's location changes due to the speaker's movement.

[0090] FIGS. 12 and FIGS. 13 are drawings for illustrating a speaker movement mode according to embodiments of the present invention. Referring to FIGS. 12 and FIGS. 13, a voice separation device (200) enters a speaker movement mode in response to a speaker movement trigger signal (TRS2).

[0091] Alice's location before movement is "P1", and accordingly, "Alice", which is Alice's speaker identifier (SID), and "P1", which is speaker location information (SPI), are matched and stored in the memory (240) of the voice separation device (200).

[0092] After moving, when Alice speaks the voice “I am Alice,” voice signals (VS1~VSn) corresponding to “I am Alice” are transmitted from the microphone (100) to the voice separation device (200). The voice separation device (200) can use the voice signals (VS1~VSn) to generate speaker location information (SPI) indicating the location (P5) and generate a speaker identifier (SID) indicating Alice’s name (Alice).

[0093] The voice separation device (200) can update the speaker identifier (SID) stored by matching it to the speaker identifier (SID) by referring to the memory (240). For example, the voice separation device (200) can update the speaker location information (SPI) "P1", which is previously matched and stored to the speaker identifier (SID) "Alice", to the speaker location information "P5" after moving.

[0094] Accordingly, the voice separation device (200) according to the embodiments of the present invention has the effect of being able to update the speaker location information stored by matching it to the speaker identifier to the changed speaker location information even if the speaker's location changes due to the speaker's movement.

[0095] FIG. 14 is a diagram illustrating the operation of a voice separation device according to embodiments of the present invention. Referring to FIG. 14, the voice separation device (200) receives voice signals (VS1~VSn) and separates and processes the voice signals (VS1~VSn) by speaker to generate a meeting minutes (MIN). The generated meeting minutes (MIN) may be stored in the form of a document file, an image file, or a voice file, but is not limited thereto.

[0096] The voice separation device (200) receives voice signals (VS1~VSn), determines a speaker identifier (SID) corresponding to each of the voice signals (VS1~VSn) using speaker location information (SPI) of the voice signals (VS1~VSn), and can store the voice signals (VS1~VSn) and the determined speaker identifier (SID) by matching them.

[0097] The voice separation device (200) can generate a meeting minutes (MIN) using voice signals (VS1~VSn) that are matched and stored with each other and a determined speaker identifier (SID). According to embodiments, the voice separation device (200) can generate a meeting minutes (MIN) by arranging the voices of each speaker in chronological order using the time when each of the voice signals (VS1~VSn) is received.

[0098] As illustrated in FIG. 14, speakers (Alice, Bob, Charlie, and Dave) sequentially utter "aaa1", "bbb2", "aaa3", "ccc4", "ddd5", "ccc6", and "bbb7". The voice separation device (200) can recognize that "aaa1" and "aaa3" correspond to Alice's voice, "bbb2" and "bbb7" correspond to Bob's voice, "ccc4" and "ccc6" correspond to Charlie's voice, and "ddd5" corresponds to Dave's voice by using voice signals (VS1 to VSn).

[0099] The voice separation device (200) can store voice signals corresponding to "aaa1" and "aaa3" and match Alice's speaker identifier, store voice signals corresponding to "bbb2" and "bbb7" and match Bob's speaker identifier, store voice signals corresponding to "ccc4" and "ccc6" and match Charlie's speaker identifier, and store voice signals corresponding to "ddd5" and match Dave's speaker identifier.

[0100] The voice separation device (200) can construct a meeting minutes (MIN) using voice signals and speaker identifiers that are matched and stored with each other, and at this time, the voice separation device (200) can arrange the voices based on the time when each voice signal is received. According to embodiments, the voice separation device (200) can convert the voice signals into text data and generate a meeting minutes (MIN) containing the speaker and the speaker's remarks using the text data and the matched speaker identifier. The remarks of each speaker included in the meeting minutes (MIN) can be arranged in chronological order.

[0101] A voice separation device (200) according to embodiments of the present invention can identify the location of a speaker using a voice signal and can distinguish which speaker's voice the voice signal corresponds to through the location of the speaker. Accordingly, even if multiple speakers speak voices simultaneously, the voice separation device (200) has the effect of separating the voices by speaker.

[0102] In addition, the voice separation device (200) has the effect of converting the voice separated by speaker into text and storing it, translating it into another language and storing it, or outputting it as voice through a speaker, etc.

[0103] Although the embodiments have been described above with reference to limited embodiments and drawings, those skilled in the art can make various modifications and variations from the description above. For example, appropriate results can be achieved even if the described techniques are performed in a different order than described, and / or the components of the described system, structure, device, circuit, etc. are combined or assembled in a form different from described, or replaced or substituted by other components or equivalents.

[0104] Therefore, other implementations, other embodiments, and equivalents to the claims also fall within the scope of the claims set forth below. Explanation of the symbols

[0105] 100: Microphones 200: Voice splitter 300: Server 400: Speaker 210: Voice signal receiving circuit 220: Processor 230: Trigger signal generation circuit 240: Memory 250: Audio signal output circuit

Claims

Claim 1 A voice separation device comprising: a voice receiving circuit configured to receive voice signals; a trigger signal generating circuit configured to generate a speaker registration trigger signal; a processor configured to receive the voice signals and identify the speaker of the voice signals using the voice signals; and a memory, wherein the processor enters a speaker registration mode in response to the speaker registration trigger signal, and in the speaker registration mode, generates speaker location information indicating the location of the speaker using the voice signals, generates a speaker identifier for identifying the speaker using the voice signals, stores the speaker location information in the memory as reference speaker location information, stores the speaker identifier in the memory as reference speaker identifier, and stores the reference speaker location information and the reference speaker identifier by matching each other, and the processor enters a voice separation mode after the speaker registration mode is terminated, and in the voice separation mode, generates speaker location information using the voice signals, determines a reference speaker identifier corresponding to the generated speaker location information, and stores the voice signals and the corresponding reference speaker identifier by matching them. Claim 2 A voice separation device according to claim 1, wherein the voice receiving circuit is configured to receive voice signals from a plurality of microphones, and the processor determines the source location of the voice signals based on at least one of the magnitude of the voice signals, the phase difference between the voice signals, and the location of the plurality of microphones, and generates speaker location information according to the determination result. Claim 3 A voice separation device according to claim 1, wherein the processor recognizes a speaker identification phrase included in the voice signals and generates a speaker identifier from the speaker identification phrase. Claim 4 In paragraph 3, the processor recognizes an identification start phrase and an identification end phrase included in the voice signals, and recognizes a speaker identification phrase based on the identification start phrase and the identification end phrase, a voice separation device. Claim 5 In claim 1, the trigger signal generation circuit receives input from a user and generates a speaker registration trigger signal according to the user's input, a voice separation device. Claim 6 In claim 1, the trigger signal generation circuit recognizes a trigger word included in the voice signals and generates a speaker registration trigger signal according to the recognition of the trigger word, a voice separation device. Claim 7 delete Claim 8 A voice separation device according to claim 1, wherein the processor outputs voice signals matched with the speaker identifier to a server, receives text data corresponding to the voice signals from the server, and stores the text data by matching it with the speaker identifier. Claim 9 A voice separation device according to claim 1, wherein the processor converts voice signals associated with the speaker identifier into text data and stores the converted text data by matching it with the speaker identifier. Claim 10 A voice separation device according to claim 1, wherein the trigger signal generation circuit generates a speaker movement trigger signal, and the processor, in response to the speaker movement trigger signal, generates speaker location information and a speaker identifier using the voice signals, determines a reference speaker identifier matched with the speaker identifier, and updates the reference speaker location information stored matched with the reference speaker identifier to the generated speaker location information. Claim 11 A voice separation method for separating voice signals comprises: a step of generating a speaker registration trigger signal; a step of receiving voice signals; a step of entering a speaker registration mode in response to the speaker registration trigger signal; a step of generating speaker location information indicating the location of a speaker using the voice signals in the speaker registration mode; a step of generating a speaker identifier for identifying a speaker using the voice signals in the speaker registration mode; a step of storing the speaker location information as reference speaker location information in the speaker registration mode; and a step of storing the speaker identifier as a reference speaker identifier in the speaker registration mode, wherein the reference speaker location information and the reference speaker identifier are matched and stored with each other; a step of entering a voice separation mode after the speaker registration mode is terminated; a step of generating speaker location information using the voice signals in the voice separation mode; a step of determining a reference speaker identifier corresponding to the generated speaker location information in the voice separation mode; and a step of matching and storing the voice signals and the corresponding reference speaker identifier in the voice separation mode. Claim 12 A voice separation method according to claim 11, wherein the step of generating the speaker identifier comprises recognizing a speaker identification phrase included in the voice signals and generating the speaker identifier from the speaker identification phrase. Claim 13 delete Claim 14 A voice separation method according to claim 11, further comprising: a step of generating a speaker movement trigger signal; a step of generating speaker location information and a speaker identifier using the voice signals in response to the speaker movement trigger signal; a step of determining a reference speaker identifier matched with the speaker location information; and a step of updating the reference speaker location information stored in matched with the reference speaker identifier with the generated speaker location information.

Images