KR20030061241A - Conference service system and method using voice recognition and voice synthesis - Google Patents

Abstract

PURPOSE: A conference service system using voice recognition and synthesis functions and a method therefor are provided to input necessary information from a user by using a voice recognition function, and to supply necessary information to the user by using a voice synthesis function, thereby remarkably increasing the number of simultaneously accessible users. CONSTITUTION: A conference service supplying system(100) has a trunk line interface(210). The trunk line interface(210) receives a user's signal, and transmits the signal to an external PSTN/mobile phone network or an IP network. The trunk line interface(210) is connected to a call processing controller(212) for controlling processing of the transmitted call signal. The call processing controller(212) is connected to an H.110 bus controller(214), an internal bus controller(216), and a DSP controller(218). The H.110 bus controller(214) controls an H.110 bus for processing connections and data transceiving between other boards. The internal bus controller(216) controls an internal bus for processing connections and data transceiving between modules. The DSP controller(218) processes digital signals.

Description

Conference service system and method using speech recognition and synthesis {Conference service system and method using voice recognition and voice synthesis}

The present invention relates to a conference service system and method using voice recognition and speech synthesis, and more particularly, to increase the number of users that can be simultaneously connected, and to receive necessary information from a user using speech recognition, and to synthesize speech. The present invention relates to a conference system and a method for providing a user with necessary information.

In order to send and receive information using multi-way voices in remote areas, a voice conference system using a telephone or a voice conference system using an internet telephone or a personal computer has been developed and used.

For the conference call, a voice message from each user to be called must be delivered to all other users. In order to do this, a telephone conference call system is used separately, and a service system that provides a conference service may provide such a function.

However, most of these conventional methods are limited to the number of users who can be connected simultaneously to a certain number or less, not only because of limitations on the system, but also due to the fact that the conference itself is difficult when more than a certain number of users are connected at the same time. do.

Meanwhile, a method of controlling various devices using a voice interface, which is a user-friendly interface, or providing various services using a voice interface, has been spreading. In the conference service providing system and method thereof, if a user inputs information to be input by voice to recognize and process the information and synthesizes necessary information into a voice, the service will be very convenient.

An object of the present invention is to provide a more advanced service than the prior art, and to provide a conference service system and a method thereof, which can greatly increase the number of simultaneous access users.

Another object of the present invention is to provide a conference service system and method that can be used using a voice interface that is most familiar to a user by acquiring information desired by a user through voice recognition and providing information required by the user through voice synthesis. It is.

Another object of the present invention is to provide a system and method for providing various voice information services such as voice information supplementary service and two-way voice answering service using a specific tone.

1 is a diagram schematically showing an overall network configuration for providing a conference service using speech recognition and speech synthesis according to an embodiment of the present invention.

2 is a block diagram showing the configuration of a conference service providing system according to an embodiment of the present invention.

3 is a diagram schematically showing an example of a hardware configuration of a conference service providing system according to an embodiment of the present invention.

4 is a diagram illustrating a user configuration in a conference service according to an embodiment of the present invention.

5 is a view showing another example of a hardware configuration of a conference service providing system according to an embodiment of the present invention.

6A and 6B are block diagrams illustrating the configurations of the relay line substrate and the voice processing substrate of FIG. 5, respectively.

7 is a flowchart of a conference service according to an embodiment of the present invention.

8 is a diagram schematically illustrating a process of obtaining a conversion function for speech conversion used for an information providing service of a specific speech according to an embodiment of the present invention.

FIG. 9 is a diagram schematically illustrating a process of converting an original voice into a target speaker's voice using a conversion function obtained by the method shown in FIG. 8.

10 is a diagram illustrating an overall network of a system for providing a two-way voice answering service according to an embodiment of the present invention.

11 is a flowchart of a two-way voice answering service according to an embodiment of the present invention.

In order to achieve the above object, in the present invention, two or more voice processing boards are connected by using an H.110 bus so that a plurality of users can simultaneously use a conference service, and some of the plurality of users are set as active users. It is possible to both transmit and receive voice signals, but the remaining users can only receive voice signals from the active user. Information input by the user through the voice is recognized through the voice recognition unit, and information to be delivered to the user is synthesized and provided through the voice synthesizer.

That is, the conference service providing system according to the present invention includes a relay line interface unit for receiving a user's signal from an external communication network and transmitting a signal to the external communication network, a call processing control unit for controlling the processing of the transmitted call signal, and another substrate. H.110 bus controller for controlling the H.110 bus for handling connection and data transmission and reception, internal bus controller for controlling the internal bus for handling connectivity and data transmission and reception between modules on its own board, used in conference services It includes a DSP control unit for processing the digital signal, and comprises at least two substrates connected via the H.110 bus.

Here, the voice recognition unit is connected to the DSP control unit, and recognizes the information transmitted to the voice from the user, the voice synthesis unit is connected to the DSP control unit, and synthesizes the information to be provided to the user to the voice and / or the It is connected to the DSP control unit, and may further include a voice recording unit for recording the conversation content, which is voice information transmitted and received through the conference service to the recording device, and is also connected to the DSP control unit, each receiving data for the conference service The apparatus may further include a reception processor, a transmission processor, and a mixing processor capable of processing data transmission and data mixing.

In the conference service providing system of the present invention, the call processing control unit and the DSP control unit work together, and the voice signal transmitted from one of the first users, which is part of all users who are connected to the conference service providing system and uses the conference service, is a corresponding voice. The voice signal transmitted from one of the second users other than the first user among all users may be processed to be transmitted to all users except the user who has transmitted the signal.

Meanwhile, the method for providing a conference service according to the present invention includes a relay line interface unit receiving a user's signal from an external communication network and transmitting a signal to the external communication network, a call processing controller for controlling the processing of the transferred call signal, and another substrate. H.110 bus controller for controlling the H.110 bus for handling connection and data transmission and reception, internal bus controller for controlling the internal bus for handling connectivity and data transmission and reception between modules on its own board, used in conference services A conference service providing method using a conference service providing system including a DSP control unit for processing a digital signal, wherein a call signal from the user passes through the relay line interface unit as a user connects to the conference service providing system through a communication network. The call processing control unit Determining whether the number of connected users exceeds the number of simultaneous accessible users, and if the number of connected users does not exceed the number of concurrently accessible users in the determining step, the active user is allowed to send and receive signals. And determining which of the passive users is allowed only to receive a signal. If the user is determined to be an active user that is allowed to transmit and receive signals, the user may receive a voice signal transmitted from the user. And processing to deliver to all other users except for, and in the determining step, ignoring the voice signal transmitted from the user without passing the signal to another user when the user is determined to be an active user who is allowed to receive a signal. It is done by

The method may further include storing a voice signal transmitted from the active user to a recording device, or the conference service providing system is connected to the DSP control unit and receives data, transmits data, and mix data for a conference service, respectively. The apparatus may further include a reception processing unit, a transmission processing unit, and a mixing processing unit that process the voice signal, and in the processing of the voice signal transmitted from the user, signal reception, signal transmission, and signal mixing are performed by the reception processing unit, the transmission processing unit, and the mixing processing unit. It can be done through.

Voice information service providing method of a specific tone according to the present invention, a relay line interface unit for receiving a signal from an external communication network and transmitting a signal to a user through the external communication network, call processing control unit for controlling the processing of incoming or outgoing call signal A speech synthesis processor for processing a synthesizer's voice having a specific tone from a source voice using a conversion function, a DSP controller for processing digital signals used in a voice information service, information provided by a user, and the like. A method of providing voice information of a specific tone by using a voice information service device including a memory device for storing information necessary for providing a voice information service, wherein a user accesses the device through a communication network using a user terminal. Voice information to be provided by The first step in which the voice information service contents provided by the user are stored in a storage device in the device as input of the voice information service contents including the contents of the information, the tone, and the time of the voice information is inputted, is provided by the user. And a third step of synthesizing contents of voice information into voices of a specific tone selected by the voice synthesis processor, and a third step of delivering the synthesized voice to the user at the time of providing voice information input from the user. . Here, the content of the voice information that the user wants to be provided may be a wake-up call or scheduling information.

Here, the conversion function used for synthesizing the voice of the desired tone by the voice synthesis processor is configured to extract spectral information of each voice by analyzing a recorded voice and noise model (HNM) of the original speaker and the target speaker. Step, corresponding position of the same pronunciation of each spectral information using DTW (Dynamic Time Warping) method, expressing the relationship between the two spectral information in a Gaussian Mixture Model (GMM) and progressively using the Expectation Maximization (EM) method Training the GMM, and optimizing the trained GMM using a least square method, wherein the second step includes the HNM of the original voice signal containing the content of the voice information to be provided by the user. Step 2-1 of analyzing and separating the spectral envelope and sound source information of the voiced sound part by using the transform function. Step 2-2 of converting the desired spectral envelope to noise, step 2-3 of performing noise removing processing on the converted spectral envelope, and mapping of the sound source information to sound source information of a desired tone. And a second to fifth step of synthesizing the voice of a desired tone by HNM synthesis using the mapped sound source information and the noise canceled spectral envelope.

In the first step, the user terminal may be a wired or wireless telephone, and the communication network may be a public telephone network or a mobile phone network. In this case, in the third step, the user terminal used by the user to access the device in the first step. The synthesized voice can be delivered.

In the first step, the user terminal may be a personal computer, and the communication network may be an internet network.

Meanwhile, the content of the voice information service provided by the user in the first step includes user terminal information specifying a user terminal to receive the voice information service, and in the third step, the user specified by the user in the first step The synthesized voice may be transmitted to the terminal.

After the second step, the method further comprises storing the voice synthesized in the second step in the form of a file in the storage device of the device. In the third step, the call processing control unit and the After establishing the connection of the outgoing call to the terminal of the user through the relay line interface, the synthesized voice may be delivered to the user, or the synthesized voice may be delivered in the form of a voice message to the terminal of the user.

In addition, the apparatus for providing a two-way answering machine according to the present invention includes a relay line interface for receiving a user's signal from an external communication network and transmitting a signal to the external communication network, a call processing control module for controlling the processing of the transmitted call signal, and from the user. A voice recognition module for recognizing an input voice signal, a key word extraction and processing module for extracting a key word from the recognized voice and processing the key word, and a database storing information for connecting a user to a desired service.

The method for providing a two-way answering service according to the present invention is provided by using the apparatus as described above, and as a user connects to the device through a communication network, a call signal from the user passes through the relay line interface. The voice signal inputted by the user is recognized by the voice recognition module as the user inputs the contents of the service desired by the user. The key word is extracted from the recognized voice signal to the keyword extraction and processing module. And determining the type of service to be provided to the user by referring to the database using the extracted key words, and connecting the service to the determined service.

Here, in the voice signal recognition step, the user inputs a content to consult with a voice. In the service type determination step, a counselor to which the user is connected is determined by using key words extracted from the user's counseling content. In the step, the user and the counselor may be directly connected by telephone.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a schematic diagram showing the configuration of an entire network provided with a conference service according to the present invention.

As shown in FIG. 1, the conference service providing system 100 of the present invention is connected to a public switched telephone network / mobile digital cellular network (PSTN / MDCN) 110 and E1 / PRI, R2. It is also connected to the IP network 120 using the Internet protocol using the HTTP / H.323 protocol. A user who uses the landline 112 or the cordless telephone 114 as the user terminal can access the conference service providing system 100 of the present invention through the PSTN / MDCN 110, and the personal computer 116 or the Internet telephone. A user using 118 or the like as the user terminal may access the conference service providing system 100 of the present invention through the IP network 120.

FIG. 2 is a block diagram illustrating in more detail the configuration of the conference service providing system 100 of FIG. 1.

As shown in FIG. 2, the conference service providing system 100 receives a user signal from an external public telephone network / mobile telephone network 110 or an IP network 120 and receives an external public telephone network / mobile telephone network 110 or IP. The relay line interface unit 210 transmits a signal to the network 120. The relay line interface 210 is connected to a call processing control unit 212 that controls the processing of the transferred call signal. The call processing control unit 212 is an H.110 bus control unit 214 which controls an H.110 bus for processing connection and data transmission and reception with another substrate, and processing connection and data transmission and reception between each module in the corresponding substrate. It is connected to the internal bus control unit 216 for controlling the internal bus and the DSP control unit 218 for processing the digital signal used in the conference service.

Unlike general one-to-one calls, a conference service is a service that allows three or more users to talk together. These conference services require that multiple users can share the signals sent and received. That is, signals transmitted by one of the plurality of users who use the conference service must be transmitted to each other user who uses the conference service. To illustrate this concept, a virtual conference room can usually be assumed. The number of users who can participate in one conference room is limited. In general, when using one board, 12 users may use the conference system. In the present invention, in order to increase the number of users who can participate in one conference room, a plurality of boards are connected to each other. Several boards exchange data using an H.110 bus. As such, the H.110 bus controller 214 is responsible for transmitting and receiving data through the H.110 bus connecting the plurality of substrates.

Meanwhile, data transmission and reception between modules in one substrate is performed using an internal bus, and control thereof is performed by the internal bus controller 216.

That is, a plurality of substrates 100_1, 100_2, .., 100_n having the components as shown in FIG. 2 are connected to each other by using the H.110 protocol as shown in FIG. A conference service providing system can be implemented. However, when implementing a conference service providing system using a plurality of substrates, not all substrates need to have a configuration as shown in FIG. 2, and each functional module may be selectively included for each substrate. Such an example will be described later.

When a connection signal from a user is transmitted to the call processing control unit 212 through the relay line interface unit 210, the call processing control unit 212 first determines in which substrate the call signal should be processed. According to the determination result, the call signal is connected to the substrate by the H.110 bus controller 214, and the call signal connected to the substrate is connected to the desired module by the internal bus controller 216 in the substrate. However, in the case of a call signal that can be processed in the substrate receiving the call signal, the control passes directly to the internal bus controller 216 in the own substrate, out of the control of the H.110 bus controller 214.

The DSP controller 218 is connected to an Rx processor 226, a Tx processor 228, and a mixing processor 230 that can process data reception (Rx), data transmission (Tx), and data mixing, respectively, to provide a conference service. It is. In addition, the voice recognition controller 220 for controlling the information transmitted by the voice from the user to control it, the voice synthesis control unit 222 for controlling the speech synthesis to deliver the information necessary to the user and the conference service It is connected to the voice recording and broadcasting processing unit 224 for performing a process of directly recording or broadcasting voice information provided through the system.

Theoretically, no matter how many users join one conference room, the conference service is available. However, in reality, if a certain number of users join one conference room, the conference itself may be difficult. In particular, in the case of the meeting room service of the star to be described as an example in the present invention, at least 1,000 users simultaneously participate in one conference room, and at this time, more than 1,000 users who participated in the conference room talk together. It would be impossible to do.

In order to solve this problem, in the conference service providing system of the present invention, only some users among all users participating in one conference room are set as transmittable users, while others only receive information transmitted by some transmittable users. Provides a way to set. This method is implemented within the conference service providing system of the present invention and not the user device.

4 illustrates a user configuration in a conference service according to an embodiment of the present invention. That is, signals transmitted from an active user or talker (denoted by "T"), which are determined in a certain manner, are respectively transmitted to all remaining users. However, signals from other passive users (Passive users or listeners) (indicated by "L") that are not active users are ignored without being passed to other users.

Selecting an active user from among a plurality of users may use various methods, such as selecting an order of paying a large amount of usage fees, selecting a first-come-first-served basis, or selecting a lottery. In addition, if necessary, a method of allowing an active user to change within a conference room at a predetermined time interval or as a specific event occurs may be used. This will be described later.

As briefly described above, the configuration as shown in FIG. 2 does not necessarily have to be implemented on one hardware substrate. In other words, the configuration of the entire system does not necessarily consist of connecting the same plurality of substrates as shown in FIG. 3, and each functional module can be formed by dispersing the plurality of substrates as necessary. In particular, when providing additional services other than the conference service together, the function for providing the additional services does not need to be included in all the boards. In such a case, the entire system in the form as shown in FIGS. It may be advantageous to construct.

That is, the relay boards 510_1 and 510_2 for providing a general access function and an additional service function and the conference service boards 520_1,..., 520_n for providing a conference service are configured.

The relay line board 510 includes a relay line interface unit 610, a call processing control unit 612, and an H.110 bus control unit 614, and a voice recognition control unit 620 and a voice synthesis control unit that can be used for providing an additional service. 622, voice recording and broadcasting processing unit 624 is included.

In contrast, the conference service board 520 includes an internal bus control unit 614 for controlling an internal bus in the board, the DSP control unit 618 which performs signal processing for providing the conference service, and the Rx processing unit 626 connected thereto. ), The Tx processor 628 and the mixing processor 630.

Thus, the conference service method will be described in detail by taking a process of performing an actual service using the conference service providing system as an example. The conference service of the present invention is implemented as a so-called "room of meeting" service of stars and fans.

That is, at a certain time, a star accesses the conference service providing system, and a plurality of users who want to call the star participate in the virtual conference room where the star is connected. Due to the characteristics of the service of meeting stars and fans, the number of concurrent users will be very large, and the number of users who can be connected simultaneously may vary depending on the hardware configuration, but more than 1,000 users can simultaneously connect to the "room of meeting". The service will be available.

At this time, the star is basically set as the active user (Talker), a certain number of users are set as the active user, the remaining users are set to the passive user (Listener). A certain number of users set as active users can enjoy a multi-party call service with the star, which is also an active user, and passive users can listen to the conversation between the star and the active user. That is, the contents of the conversation are broadcast to the passive user. Message control from the active and passive users is performed by the DSP controllers 218 and 618, Rx processors 226 and 626, Tx processors 228 and 628, and mixing processors 230 and 630 connected thereto.

The selection of the users who can use the room service of the meeting with the star and the selection of the active users can be done by selecting the first-come-first-served basis and the user who pays the most fee. For convenience, such as payment of the fee for using the service, the meeting room service is preferably operated on a membership basis. In addition, although the hourly billing method used in the ordinary ARS service may be used for payment of the fee for using the service, a different payment method is provided in view of the fact that discriminatory services are provided according to the fee (determining whether or not to participate in the conversation). It is preferable that there is. To this end, a certain amount of money may be charged in advance to use the service, and the service may be used. You can use a variety of payment methods, such as bank transfer, credit card, mobile phone payment, and telephone bills and consolidated bills. For this purpose, although not shown in the drawings, a billing server may be provided, and may include a database storing member information and charge amount information.

7 shows a flowchart of the meeting room service of the present invention.

First, the user accesses the conference providing service system of the present invention using a wired / wireless phone, a computer, an internet phone, etc. (S710) and selects a meeting room service according to the voice answering guide (S720). In the case of a membership-based service, the member authentication process is performed according to the voice guidance. In addition, if necessary, membership may be registered through voice guidance. Program information, such as which stars appear in the meeting room at any time, may be promoted on / offline in advance through other methods, and program information may be heard through voice guidance, depending on the user's choice. Programs can be organized by time zone, and in some cases multiple conference rooms can be opened in one time zone. In the latter case, you can access the meeting room service and select the meeting room of the star you want to participate again.

When the user selects the meeting room service, the conference providing service system determines the number of users connected to the meeting room to determine whether the currently connected user can access the meeting room (S730).

If the access is possible, the user enters the meeting room, i.e., accesses the meeting room service (S750). If the user cannot access more than the number of concurrently connected users, a voice prompt is provided to the user. (S740).

When the user connects to the meeting room, the user determines and inputs a fee to be paid (S760). That is, the amount of money to be paid to become an active user in the room service of the meeting is determined. The input of the amount of money can be recognized by DTMF by pressing the numeric keypad of the phone, or by recognizing the amount of money desired by the user. When the user inputs the amount of money, it is preferable to use the synthesized voice by the voice synthesis control unit 222 to guide the input amount.

After recognizing the input amount, it is determined whether the corresponding user becomes an active user or a passive user (S770). The determination of the active / passive user may use a method of determining a certain number of users as active users in order of users who input a large amount of money by comparing the amounts input from each user.

The determined user status is voiced back to the user, and if the user status is determined to be an active user, the user can conference with the star (S780), and if the user status is determined to be a passive user, the user S hears the contents of the multi-party call between the star and the active user (S790).

The user status can be kept the same while the meeting room service continues, or can change the user status at regular intervals, and can be linked to the control of the service operator or to specific events of the stars participating in the meeting room. It is also possible to change it.

Meanwhile, the call may be automatically recorded in the conference service system. For this purpose, the conference service system has a recording processing unit, and records the conversation contents in a storage device such as a hard disk provided for recording processing. The recorded conversation contents may be produced on a CD or the like and provided to the user later.

According to the present invention, there is also provided a voice wake-up call and scheduling service using a specific tone and its system. This is implemented with so-called "star wakeup call" and "star scheduling" services.

First, the star wakeup call service will be described.

The user accesses the voice wake-up service system and selects a star to which he / she receives a wake-up service and a time to receive wake-up service. In addition, the user enters a phone number to receive the morning call service. The telephone number to receive the wake-up call service may be automatically set by the system without a separate input when the user accesses the voice wake-up call service system by telephone. However, the user may choose to use the wake-up service with a telephone number other than the telephone number used to connect. In addition, if necessary, a user may select a phrase to be used for a wake-up service.

According to the information input by the user, a voice call module is used to synthesize a morning call message using the voice of the corresponding star. The synthesized wake-up message is preferably stored in a voice file in preparation for repeated use.

When the wake-up call time set by the user is reached, the user calls the wake-up call number set by the user and delivers the morning call message synthesized above.

In order to synthesize a wake-up message with the voice of the corresponding star, a conversion function for converting the original speaker's voice into the voice of the target speaker (star) is obtained from the voices of the two speakers with the same contents. Use a method of converting a message into a star's voice. Now let's take a closer look at these methods.

The purpose of voice conversion is to change the voice of a specific speaker A (e.g., the operator of the star morning call service in the present invention) to the voice of the desired speaker B (e.g., a particular star in the present invention).

To convert speech, one must first quantify various characteristics that characterize that person from the person's utterance. In order to fully express the characteristics of a person's voice, not only the characteristics of the person's saints themselves (voices), but also many things such as pronunciation habits, intonation, rhyme information, etc., must be quantified. can do.

However, if you consider only very short utterances, for example a single utterance, a lot of extra-word information, such as average speed, pitch, and intonation, is used to distinguish that voice is a specific person's voice. It doesn't have a big impact. Instead, the characteristic information of a person's saints plays a large role in distinguishing that the voice is that person.

Therefore, in the case of short speech, a relatively satisfactory result can be obtained even if the speech conversion is performed using only the spectrum envelope information representing the vocal characteristics of the person.

Therefore, for speech conversion, it is necessary to find an appropriate model that represents the probabilistic relationship between the spectral envelopes of two speakers with the same content. Based on such a model, a conversion function for converting between two spectral envelopes is obtained. This requires training the model through experimental data.

8 schematically illustrates a process of obtaining a conversion function for speech conversion. First, the voices of the original speaker and the target speaker (810_s, 810_t) are recorded with the same contents, and then spectrum spectrum (spectrum envelope) (Harmonic and Noise Model) analysis method (820_s, 820_t) is used. 830_s and 830_t). Thereafter, the same pronunciations are mapped to the same position by using a dynamic time warping (DTW) method (840).

To obtain the transform function, we need to train a probabilistic model that represents the relationship between the two spectra. To this end, the relationship between the two spectral information for each corresponding pronunciation is represented by a Gaussian Mixture Model (GMM) and trained in a progressive manner using an Expectation Maximization (EM) method (860). Next, it is matched again by pronunciation using the alignment path obtained from DTW.

Here, the Gaussian Mixture Model (GMM) is a classic model used in many pattern recognition techniques. This model assumes that the distribution of observed factors follows a certain normal distribution, and that each of the observed factors is independent of each other. It is suitable for use in situations where the order of observations is not important. This model can be thought of simply as a Hidden Markov Model (HMM) whose distribution function is a Gaussian and has the same probability of progressing to each state.

This function can be used to create a speech conversion function because the input of the conversion function is not related to time but only to the spectral envelope of the original speech at a given time.

In the GMM model, eventually, each phoneme is characterized by its average characteristic and degree of distribution. Assuming that the GMM model is trained, the probability that a particular input vector is a given phoneme is given as a conditional probability.

The parameters of the GMM model can be obtained from the given training input (final set of vectors) through the Expectation Maximization (EM) algorithm.

Expectation Maximization (EM) algorithm is a general method of taking a vector from a set of training inputs and changing the model's parameters by repeating it in a direction that maximizes the expected value of the vector into the correct phoneme model. By iterating until convergence, a transform function is obtained.

The transformation function thus obtained is again optimized using least squares (860).

Using the method described above, a conversion function for converting a desired original voice to a desired speaker's target voice using a small amount of recorded voices of about 4 minutes can be obtained. This method is particularly useful for short speech in syllable units. Do.

FIG. 9 schematically illustrates a process of converting an original speech into a target speaker's speech using a conversion function obtained by the method shown in FIG.

First, the original speech signal 910 is separated into a spectral envelope 930_v and a sound source specification 930_p of the voiced sound unit through the HNM analysis 920, and the separated spectral envelope 930_v is converted into a transform function 940. 950 is transformed into the desired spectral envelope. The converted spectral envelope is not perfect, so synthesizing the signal using it results in a noisy synthesized sound. To avoid this, noise processing is performed using a noise filter for additionally removing noise (970).

Meanwhile, separately, sound source information 930_p separated from the original speech signal 910 is mapped to convert the height of the sound (960), and merged with the processed spectral envelope information as described above to HNM synthesis (980). ) Is converted into the speaker's voice (990).

Next, the star scheduling service will be described.

The user accesses the voice scheduling service system and selects a star and scheduling contents for which he / she wants to receive the scheduling service. In addition, the user enters a phone number to receive the scheduling service. As in the wake-up service, the telephone number to receive the scheduling service may be automatically set by the system without a user input. However, the user may, of course, choose to use the scheduling service with another phone number in addition to the phone number used when he connects. In addition, when the phone to which the scheduling service is intended is a mobile phone, the scheduling service may be selected to be provided through voice and / or text message rather than a direct call. In addition, as in the morning call service, the user may select a phrase to be used in the scheduling service.

According to the information input by the user, a scheduling message is synthesized using the voice of the corresponding star using the voice synthesis module. Synthesis of the scheduling message may be performed immediately after the user inputs a schedule, or may be performed at a time when a user inputs a schedule. When synthesizing the scheduling message in advance, it is preferable that the synthesized scheduling message is created and stored in a voice file. Synthesizing the scheduling message by the voice of the star is performed in a similar manner as described above with respect to the star morning call service.

When the time set by the user is reached, the telephone number set by the user is dialed to deliver the synthesized scheduling message, or the user's scheduling message is transmitted by voice and / or text message.

The voice wakeup call system and the voice scheduling system of the present invention may be implemented using a conference system described above with respect to its structure and function. That is, the process according to the user's connection is performed by the relay line interface unit, and the service selection and input of the service contents are performed by the call processing control unit, the DSP control unit, and the voice processing module (voice synthesis control unit, voice recognition control unit, etc.) connected to the DSP control unit. The morning call and the scheduling service may be provided through a relay line interface unit, a call processing controller, a DSP controller, a voice synthesis controller, and the like. However, it is of course possible to configure a separate system for voice wake-up call and / or scheduling service.

In the voice wake-up call and / or scheduling service system of the present invention, when a web server that a user can access through the Internet is provided, more convenient and various services can be provided to the user.

For example, if a user enters information for using a star wakeup call and / or scheduling service, that is, a desired star, a telephone number to be provided with a service, a wakeup call time, a schedule, and the like, the system uses the information. Star wakeup call and / or scheduling service can be provided. As such, the method of inputting necessary information using the Internet has an advantage that it can be used at a convenient and low cost compared to a method of inputting corresponding information using a telephone.

In addition, in the case of a conference service system, if a web server accessible through the Internet is provided, various additional services can be provided. For example, a room service program guide for a star meeting can be provided through the Internet, and a service for re-listening the recorded contents or a service for ordering and selling a CD containing the recorded contents can be provided.

According to the present invention, there is also provided a two-way voice answering service and a system thereof.

10 shows an example of the configuration of a two-way voice answering system of the present invention. The system 1000 of the present invention includes a relay line interface 1010 for controlling a connection with a user 1060 connecting to a system through a communication network, a call processing module 1020 for controlling the processing of a transmitted call signal, and an input from a user. A voice recognition module 1030 for recognizing a voice signal, a key word extraction and processing module 1040 for extracting a key word from the recognized voice and processing accordingly, and a database storing information for connecting a user to a desired service ( 1050). As described above, the user 1060 can access the two-way voice answering service system of the present invention through a communication network such as a public telephone network or a mobile telephone network, and the system of the present invention is also provided through the system of the present invention. It is also connected to the expert's terminal 1070 that directly provides services.

An example of a service provided through the two-way voice answering system of the present invention is a two-way consultation service. This service flow is shown in FIG.

First, the user accesses the two-way voice answering system of the present invention by using a user terminal such as a wired / wireless telephone (S1110).

Next, the user talks his / her counseling contents over the phone according to the guidance of the system (S1120).

The voice information transmitted from the user is recognized through the speech recognition module (S1130), and a key word is extracted from the recognized sentence (S1140). When the key word is extracted, the category of the professional counseling service required by the user and the corresponding counselor may be determined from the key word information (S1150).

When the counselor is determined, the system of the present invention directly connects the user who requested the consultation to the counselor and provides a counseling service (S1160).

Unlike the existing one-way service, the two-way voice answering service detects and connects the appropriate service contents from the voice information provided by the user on the system, which is convenient and difficult to accurately determine the category of the service desired by the user. In addition, there is an advantage that the desired service can be provided.

The present invention has been described in detail with reference to preferred embodiments, but the scope of the present invention is not limited thereto, and should be interpreted by the following claims. In addition, it will be understood by those skilled in the art that various modifications or changes may be made without departing from the spirit of the present invention.

As described above, using the conference service system according to the present invention, more than 1,000 users can simultaneously access a conference room and share a voice message, which is an active user who can only speak and listen to a specific user. By setting this up, you can solve the problem of too many users accessing a conference room, and use it appropriately for certain applications, such as a fan club service with a star. Since the information input by the user is delivered through voice recognition, it is convenient, and the contents of the conversation can be recorded directly in the system, and the information can be applied to various additional services.

In addition, it is possible to provide a wake-up call and / or scheduling service with a specific tone by using voice synthesis and modulation, and to provide a convenient and effective service to the user by analyzing and connecting the user's desired contents with the voice. Can be.

Claims (23)

A relay line interface unit for receiving a user's signal from an external communication network and transmitting the signal to the external communication network, a call processing controller for controlling the processing of the transmitted call signal, an H.110 bus for connecting to another board and processing data transmission and reception. H.110 bus control unit for controlling the internal bus control, the internal bus control unit for controlling the internal bus for processing data transmission and reception between each module in its own board, DSP control unit for processing digital signals used in the conference service, And at least two substrates connected via an H.110 bus. The method of claim 1, And a voice recognition unit connected to the DSP control unit and recognizing information transmitted from the user by voice. The method of claim 1, And a voice synthesizer connected to the DSP control unit and synthesizing information to be provided to a user into voice. The method of claim 1, And a voice recording unit connected to the DSP control unit and recording a conversation content, which is voice information transmitted and received through a conference service, to a recording device. The method of claim 1, And a reception processing unit, a transmission processing unit, and a mixing processing unit, each of which is connected to the DSP control unit and can process data reception, data transmission, and data mixing for a conference service. The method of claim 1, The call processing control unit and the DSP control unit operate together so that a voice signal transmitted from one of the first users, which is a part of all users who are connected to the conference service providing system and uses the conference service, is all other users except the user who transmitted the voice signal. And a voice signal transmitted from one of the second users other than the first user among all the users. A relay line interface unit for receiving a user's signal from an external communication network and transmitting the signal to the external communication network, a call processing controller for controlling the processing of the transmitted call signal, an H.110 bus for connecting to another board and processing data transmission and reception. H.110 bus control unit for controlling the internal bus control unit for controlling the internal bus for processing connection and data transmission and reception between each module in its own board, DSP control unit for processing digital signals used in conference services As a conference service providing method using a conference service providing system, Transmitting a call signal from the user to the call processing controller via the relay line interface unit as the user accesses the conference service providing system through a communication network; Determining whether the number of connected users exceeds the number of simultaneous accessible users, If the number of connected users does not exceed the number of concurrently accessible users in the determining step, determining, by the user, which of the active users is allowed to both transmit and receive signals and a passive user who is only allowed to receive signals; In the determining step, if the user is determined to be an active user who is allowed to both transmit and receive a signal, the voice signal transmitted from the user is transmitted to all other users except the user, and in the determining step, the user receives only the signal. If it is determined that this is an allowed active user, processing to ignore the voice signal delivered from the user without passing it to another user. The method of claim 7, wherein And storing the voice signal transmitted from the active user to a recording device. The method of claim 7, wherein The conference service providing system is connected to the DSP control unit, and further includes a reception processing unit, a transmission processing unit, and a mixing processing unit, each of which can process data reception, data transmission, and data mixing for a conference service, And in the processing of the voice signal transmitted from the user, signal reception, signal transmission, and signal mixing are performed through the reception processor, the transmission processor, and the mixing processor. A relay line interface unit for receiving a signal from an external communication network and transmitting a signal to a user through the external communication network, a call processing control unit for controlling incoming or outgoing call signal, and an object having a specific tone from an original voice using a conversion function Speech synthesis processing unit for processing the speaker's voice synthesis, DSP control unit for processing digital signal used in voice information service, memory for storing information provided from user and other voice information service A method of providing voice information of a specific tone using a voice information service device comprising a device, Voice information service provided by the user as the user inputs the voice information service content including the content of the voice information to be provided, the tone, and the time to be provided with the voice information by accessing the device through a communication network using the user terminal. A first step in which content is stored in the storage device in the device, A second step of synthesizing contents of voice information to be provided by a user into voices of a specific tone selected by the user by the voice synthesis processor; And a third step of delivering the synthesized voice to the user at the time of providing the voice information input from the user. The method of claim 10, A conversion function used for synthesizing a voice of a desired tone by the voice synthesis processor, Extracting spectral information of each voice by analyzing the recorded voice of the original speaker and the target speaker, HNM (Harmonic and Noise Model), Mapping positions of the same pronunciation of each spectrum information by using a dynamic time warping (DTW) method, Expressing the relationship between the two spectral information as a Gaussian Mixture Model (GMM) and gradually training using an Expectation Maximization (EM) method; A method of providing a voice information service of a specific tone obtained by optimizing a trained GMM using a least square method. The method of claim 10, The second step, Step 2-1 of HNM analysis of the original speech signal containing the contents of the speech information to be provided by the user and separation into spectral envelope and sound source information of the voiced sound; Step 2-2 of converting the spectral envelope into a desired spectral envelope using the transform function; Performing a noise canceling process on the converted spectral envelope; A second to fourth mapping process for converting the sound source information into sound source information of a desired tone; And a second to fifth step of synthesizing a voice of a desired tone by HNM synthesis using the mapped sound source information and the noise canceled spectral envelope. The method of claim 10, The content of voice information to be provided is a wake-up call or scheduling information. The method of claim 10, In the first step, The user terminal is a wired or wireless telephone, The communication network is a public telephone network or a mobile telephone network voice information service providing method of a specific tone. The method of claim 14, And in the third step, delivering the voice synthesized to the user terminal used by the user to access the device in the first step. The method of claim 10, In the first step, The user terminal is a personal computer, And the communication network is an internet network. The method of claim 10, The voice information service contents provided by the user in the first step includes user terminal information specifying a user terminal to receive the voice information service. And in the third step, delivering the voice synthesized to the user terminal specified by the user in the first step. The method of claim 10, After the second step, And storing the voice synthesized in the second step in the form of a file in the storage device of the device. The method of claim 10, In the third step, And establishing a connection of an outgoing call to the user's terminal by the call processing control unit and the relay line interface, and transmitting the synthesized voice to the user. The method of claim 10, In the third step, Voice information service providing method of a specific tone to deliver the voice synthesized in the form of a voice message to the user terminal to the user. A relay line interface for receiving a user's signal from an external communication network and transmitting a signal to the external communication network, a call processing control module for controlling processing of the transmitted call signal, a voice recognition module for recognizing a voice signal input from the user, Apparatus for providing a two-way answering service comprising a keyword extraction and processing module for extracting a key word from the voice and processing accordingly, and a database storing information for connecting a user to a desired service. A relay line interface for receiving a user's signal from an external communication network and transmitting a signal to the external communication network, a call processing control module for controlling processing of the transmitted call signal, a voice recognition module for recognizing a voice signal input from the user, A method for providing a two-way answering service using a device including a keyword extraction and processing module for extracting a key word from voice and processing accordingly, and a database storing information for connecting a user to a desired service, Transmitting a call signal from the user to the call processing control module via the relay line interface as the user connects to the device through a communication network; Recognizing a voice signal input from the user by the voice recognition module according to a voice input of contents of a service desired by the user, Extracting a key word from the recognized voice signal by the key word extracting and processing module, and determining a type of service to be provided to the user by referring to the database using the extracted key word, And the user is connected to the determined service. The method of claim 22, In the voice signal recognition step, the user inputs the contents to consult by voice, In the service type determination step, a counselor to which the user is connected is determined using key words extracted from the counseling content of the user. In the service connection step, the user and the counselor are connected via a direct telephone method.