KR20050004755A - Wireless Communication Devices with a Built-in Speech Recognition Function - Google Patents

Abstract

PURPOSE: A wireless communication device embedded with a voice recognition function for credit card transactions is provided to secure convenience/promptness/security for actual use at the minimum expense while maximally utilizing a merit of mobility, when an existing portable terminal is used for credit card settlement. CONSTITUTION: A speaker(300) receives voice information of a client holding the wireless communication device. A voice recognition interface(310) analyzes features of inputted voice data. A user voice data storing part(340) stores voice data features analyzed through the voice recognition interface. A voice data comparing part(330) analyzes/compares the voice data inputted for financial transaction with the voice data features of the client stored in the user voice data storing part after receiving financial information from the client through the speaker. A text generating part(350) converts the inputted voice data into text data in case that the voice features of two data are identical with each other.

Description

Wireless Communication Devices with a Built-in Speech Recognition Function}

The present invention relates to a wireless mobile communication payment system for requesting a credit card transaction using a wireless communication device and receiving a transaction approval for a requested message and performing a credit card payment process wirelessly. Voice information input means using the wireless communication device owning client's own voice to input necessary information (sales price, merchant number, etc.), and voice recognition means for analyzing the characteristics of voice data inputted by the client more accurately. Analyzing and comparing the client voice quality storage means for storing and managing the input voice data characteristics in advance in the wireless communication device with the voice data of the client inputted for a transaction at the time of credit card payment and the previously input client voice data characteristics. Voice data comparison means And a text converting means for converting the input voice data into text (numeric) data required for payment information and the generated payment information data when the two data match with each other as a result of the comparison through the voice data comparison unit. And payment information data transmission / reception means for transmitting to a payment processing authentication financial network and receiving payment approval data from the financial network. The convenience and security of the client using a voice recognition function during wireless card payment. The castle has been greatly improved.

In general, the voice is superior to other interfaces can be divided into three.

The first voice is the most natural means of communication for human beings. If such communication with a machine is possible as humans communicate with each other, the convenience will not be comparable to any interface. In particular, in the case of keyboards, keyboards, and other interfaces, separate training for use is required, but in the case of voice, training is unnecessary, and thus, there is an advantage in that the system is highly adaptable.

Second is parallelism.

Simultaneous processing, such as typing on the keyboard while talking by voice, allows for free movement rather than from a fixed position for using or entering information.

Thirdly, it is possible to speed up data input and remote input.

Voice is said to be four times faster than writing. Therefore, it is possible to input data at high speed when continuous speech recognition is possible. Remote input through telephone line or wireless communication is possible, and direct data input enables improved operation reliability and operation efficiency by eliminating intermediate results such as data key-in.

Anyone can predict that the flow of payment in the future will go to payment using credit card rather than cash payment. In this case, the credit card payment is a form in which a credit card company (payer) is involved between a customer (buyer) and a merchant (seller) to pay to the merchant instead of the customer and later receive payment from the customer. Of course, credit card companies charge a certain fee to customers and merchants. The credit card company registers the merchant as a merchant and assigns a merchant number.

The customer establishes a transaction using the card checker installed at the merchant. That is, when the card is read in the checker, the transaction details (item price, installment months) are sent to the card company, and the card company accepts or rejects the transaction according to the credit rating of the customer, or the transaction price exceeds the card usage limit. It sends the card through the merchant's card check.

Nowadays merchants pay a small fee to credit card providers and receive payments by direct debit. Direct debit means that the seller submits the bank account number to receive the credit card payment when the merchant requests the card company, and the payment is made regardless of the card company's submission of the sales slip 4-5 days after the transaction is approved.

Sales slips are usually collected once a month and submitted to the credit card provider. The transaction for the transaction consists of reading the card into the checker and entering the transaction price and the installment months. It is much simpler and more convenient to use than the wireless electronic payment system currently being developed. In addition, since it uses a wired network, it can be said to be superior in security without a separate device than a wireless network.

At present, various inventions have been prepared to use the convenience of mobile / mobile for mobile payments in credit card payment or online financial transactions.However, when implementing the service, this can provide some convenience compared to the existing method. It is also true that the customer is questioning whether to use it, and it is necessary to reconsider overcoming the security problem, which is the disadvantage of wireless over wires.

It should not be overlooked that the most important factor in making the service available to customers is convenience, stability and economy.

In view of the existing inventions related to this, only the use of a mobile terminal as a card checker is a good idea. However, the customer's credit card number / deposit account number must be entered into the handset keyboard, and the merchant number (8-11 digits) that the seller has received from the credit card company has to be manually entered.

Compared to a system that reads a credit card at the checker and inputs only the transaction amount / monthly installment in the current credit transaction, it lags behind in convenience and security except for the advantage of using a portable terminal owned by the user. can do.

Credit card payments using wireless handset currently under development or completed have various security methods (mainly notary key cryptography), but it is still unknown how much the security weaknesses of wireless can be solved.

In addition, convenience / promptness of use remains a challenge. Customers who use shortcut keys because they don't want to press the 9-11 digits of the phone number will read the card to the current card checker and bypass the system where all credit card transactions are closed. The question is whether you should try the service, but the answer is probably NO. For a single transaction, you will need to press 16 card numbers, merchant number 8-11 digits, transaction amount (usually 4-6 digits based on credit card transactions) and enter 6-8 digits of password directly. I doubt it.

An object of the present invention for solving the above problems is to make the most of the advantages of mobility when using the existing mobile terminal for credit card payment, voice to secure convenience / speed / security at a minimum cost in actual use It is to develop and utilize recognition system for service.

That is, according to the present invention, voice information input means using the voice of the client owning the wireless communication device and voice data characteristics input by the client can be used to input information (sales price, merchant number, etc.) required for the wireless credit card payment request. Voice recognition means for accurate analysis, client voice feature storage means for storing and managing the voice data quality inputted by the client in the wireless communication device in advance and voice data of the client inputted for transaction at the time of credit card payment After comparing the voice data comparing means and the voice data comparing unit for analyzing and comparing the characteristics of client voice data, if the two data match, the input voice data is converted into character (numeric) data necessary for payment information. Character Switching Means to Generate Built-in voice recognition function characterized in that it comprises a payment information data transmission and reception means for transmitting the generated payment information data to the payment processing authentication financial network via a wireless communication network and receiving payment approval data from the financial network. The purpose is to develop a wireless communication device,

To this end, the client voice data storage unit and the voice to store in advance the voice recognition interface for receiving the voice data of the client and the voice data (voice number data from 0 to 9) of the client input through the voice recognition interface in advance A low pass filter (LPF) for removing or reproducing noise or distortion of the voice data of the client input through a recognition interface, and a collection of voice data coming out of the low pass filter and a feature of voiced sound are extracted. A real speech section that removes unnecessary silence parts from the speech extracting unit and the pitch extracting unit that extracts voiced and unvoiced speech from the formant extracting unit and the speech data from the formant extracting unit, and detects only necessary speech portions. Detection section and the real sound section test Pre-emphasis unit for extracting the consonant features of the voice data coming out through the unit, voice feature parameter detection unit for comprehensively detecting the characteristics (usually expressed as a number) of the voice data passed through each component unit and the voice feature parameter detection unit The voice data comparing unit comparing the characteristics of the voice characteristic of the client with the characteristics of the client client previously stored in the voice data storage unit and the voice data comparing unit are compared. A character generation unit for converting voice data into character (numeric) data required for payment processing, and transmitting the generated payment information data to a payment processing authentication financial network through a wireless communication network and receiving payment approval data from the financial network. Have a payment information data sending and receiving section It is an object of the present invention to provide a wireless communication device with a voice recognition function, characterized in that made.

1 is an embodiment showing the entire flow of the card payment system using a conventional wired network.

2 is a schematic diagram of a wireless card payment system using a mobile terminal having a voice recognition system.

3 is a configuration diagram of a system additionally configured in a mobile terminal for a wireless payment service.

4 is a configuration diagram of a voice data comparison unit.

5 is a diagram illustrating a process of selecting a credit card payment mode from a menu of a mobile terminal.

Figure 6 is an embodiment showing a form for creating a transaction history for credit card payment.

7 is an overall flowchart of how voice data for transaction is processed in a voice recognition system.

8 is a flowchart illustrating a process of generating and storing client voice data.

[Description of main part of drawing]

300: voice data input unit (speaker) 310: voice recognition interface

320: Low Pass Filter 330: Voice Data Comparator

340: Client voice data storage unit 350: Character generation unit

360: vernier 370: display screen

400: Formant extraction unit 410: Fitch extraction unit

420: real sound section detection unit 430: pre-emphasis unit

440: voice parameter detection unit 450: comparison unit

The present invention provides a voice recognition system and a wireless credit card using the system in order to secure convenience, speed and security at a minimum cost in actual transactions while maximizing the advantage of mobility when using an existing mobile terminal for credit card payment. A payment method, comprising: inputting information (sales price, merchant number, etc.) required for a wireless credit card payment request, using voice information input means using the voice of the client owning the wireless communication device, and voice data characteristics input by the client. Voice recognition means for more accurate analysis, client voice feature storage means for storing and managing the voice data quality input by the client in the wireless communication device in advance, and voice data and pre-input of the client input for transaction at the time of credit card payment Client voice days Characters that are generated by converting the input voice data into character (numeric) data required for payment information when the two data match, as a result of comparing the voice data comparison means for analyzing and comparing characteristics and the voice data comparison unit. And a payment information data transmission / reception means for transmitting the switching means and the generated payment information data to a payment processing authentication financial network through a wireless communication network and receiving payment approval data from the financial network. The client voice data storage unit and the voice recognition interface for storing and managing voice data (voice number data from 0 to 9) of the client input through the voice recognition interface and voice recognition interface to receive voice data in advance. The climax input Low-pass filter unit (LPF) for removing or reproducing noise or distorted parts of the voice data of an unwitting voice, and a formant extraction unit for extracting the collection of voice data and voice characteristics of the voice data from the low-pass filter unit and the formant extraction A pitch extractor for discriminating and extracting voiced and unvoiced voice data from the voice data, and a real speech section detector and a real speech section detector for removing unnecessary silence parts from the speech data output from the pitch extractor and detecting only necessary speech parts. Pre-emphasis unit for extracting the consonant features of the voice data coming out through the voice feature parameter detection unit for comprehensively detecting the characteristics (usually expressed as a number) of the voice data passed through each component and the voice feature parameter detection unit Voice characteristics of the client and the voice data storage The voice data comparison unit comparing the characteristics of the voice quality of the client previously stored in the unit and the voice data comparison unit, and the character data (numeric) data required for the payment processing when the two data match, And a payment information data transmitting / receiving unit for transmitting the generated text information and the generated payment information data to a payment processing authentication financial network through a wireless communication network and receiving payment approval data from the financial network. The present invention relates to a wireless communication device having a voice recognition function.

In addition, for the credit transaction according to the present invention, the characteristics of the voice tone of the client must first be matched until the transaction statement is completed (usually undergoing a process of checking whether the client is 17-18 in one transaction. -8-11 digits of merchant number, 5-6 digits of transaction value, installment months, card number, etc.) The use by third parties is blocked, and the second security is realized by entering a password at the end of the transaction.

In addition, the present invention is a credit card payment transaction is possible only by numeric input (ie card number, merchant number, transaction amount, installment month, transaction approval number received from the card company) desired input in a separate credit payment mode added to the mobile terminal Select the desired credit card (credit card information is pre-populated in the mobile terminal-credit card number, client personal information for security), and enter the merchant number, transaction price, and installment months by voice. After reviewing the transaction details displayed on the display screen, enter the password, send it, and receive the transaction approval number from the card company.

In addition, since the present invention uses non-consecutive numbers from 1 to 9 for credit card payment, the system can be made simpler by excluding continuity of voice recognition and can give certainty in input.

In addition, the speech recognition method, which is the core of the present invention, is generally divided into various types according to the number of words, whether or not the speaker depends on the continuity of the uttered sound and the recognition unit.

First, the number of words is divided into small (10-100), medium (100-500), and large (1000 or more).

Since the present invention uses only 10 numbers, it is possible to further refine the feature extraction for each phonetic character, and input the same phonetic character more than five times when receiving the phonetic character information of the client to be used for the previously input voice comparison. By extracting the commonality of the features of the client's voice in more detail, the error rate of the voice recognition of the existing mobile phone can be reduced by 100% even if the client is not recognized by the surrounding noise or the state of the client's voice. (The speech recognition system used in existing mobile terminals has to store data for many vocabularies, so the recognition rate is deteriorated.)

In the case of second speaker dependency, the speaker-dependent is the same when the speaker used for learning is the same as the speaker used for recognition, and the speaker-independent is called when the speaker used for learning is different from the speaker used for recognition. It is known that the speaker dependency has a higher recognition rate than the speaker independence. (The speech recognition system has security in itself if only the characteristics of the client speech pattern are completely extracted.) The present invention will select a speaker-dependent method, and this selection itself has some security.

According to the continuity of the third voice, it is divided into independent voice, connected voice, continuous voice, and free speech. In the case of independent speech, the beginning and end points of the speech are clear, which means that a clear pause is inserted at the beginning and end of the word when the words are spoken. Connected speech refers to voices uttered so that a short pause exists between words, and continuous speech sounds refers to voices uttered such that an extremely short pause exists between words. Free speech is a voice that is spoken in everyday conversations, and refers to both phonological sounds with non-grammatical syntactic structures in addition to grammatically correct sentences (eg, independent school "school", / connected school "I", "school"). "I go to school", "to go" / continuous "I go to school" / "I go to school. With Hongseok.")

Fourth, in the case of a recognition unit, a large recognition unit may be used like a word, but a subword may be used such as a phoneme or a syllable. To simplify the implementation of the recognition system and recognize the large vocabulary, the recognizer is constructed using units such as phonemes rather than units such as words or syllables.

Since the present invention will use a voice recognition system for credit card payment using a voice (numbers only) as a wireless portable terminal, a simple independent voice is sufficient, and the vocabulary number is about 10 characters. Will be used.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. First, in adding reference numerals to components of each drawing, it should be noted that the same components are denoted by the same reference numerals as much as possible even if they are displayed on different drawings. In addition, in the following description, in order to explain the embodiments of the present invention, specific details such as specific elements are illustrated, which are provided to help a more general understanding of the present invention, and the present invention may be practiced without these specific details. It will be apparent to those skilled in the art. In describing the present invention, when it is determined that a detailed description of a related known function or configuration may unnecessarily obscure the subject matter of the present invention, the detailed description thereof will be omitted.

Figure 1 shows the overall flow of the card payment system using a conventional wired network.

The card checker 100 of the merchant is connected to a credit card transaction exclusive network 110 to which a credit card provider 120 (or a card inquiry provider, also known as a VAN company) is subscribed, and a credit card transaction of the credit card provider 120 is performed. Dedicated network 110 is connected to the card company 130 and the financial common network (140).

That is, when the merchant requests credit card payment, the card checker 100 connects the information of the credit card to the credit card inquiry service provider 120 through the credit card transaction exclusive network 110, and the credit card inquiry service provider 120 receives The customer's credit card information is sent to the card issuer 130 to confirm the credit of the card, and the merchant's card checker 110 notifies whether the transaction is possible or not. When a transaction is possible, a transaction approval number is assigned.

2 is a schematic diagram of a wireless card payment system using a mobile terminal 200 having a voice recognition system.

A mobile terminal 200 having a voice recognition system for embedding credit card information (card number and client personal information) and inputting transaction details by voice and a wireless communication network 210 to which the portable terminal 200 as described above is subscribed. And the wireless electronic transaction gateway 220 which checks whether the mobile terminal 200 and the client correspond to the transaction request message transmitted from the data transmission system and connects to the existing credit card inquiry service provider 120, and the credit card. It is composed of the card issuer 130 and the financial common network 140 of the account holders connected with the inquiry service provider 120 to determine the credit of the credit card to determine the approval.

It is the same as the system of FIG. 1 except for using the wireless communication network 210 instead of the wired credit card transaction dedicated network 110 and the transaction being made through the wireless electronic transaction gateway 220 in the middle.

When the transaction is completed, the card company 130 sends the transaction approval number to the display screen of the transaction requesting mobile terminal 200, and the seller sends the payment price, the installment month number, and the transaction approval number to a separate form paper by the card company instead of the card sales slip. Fill in and get the buyer's handwritten signature to close the deal.

3 is a block diagram of a voice recognition system to be added into a portable terminal for service.

If you select the credit card payment mode with your mobile device, you do not need to enter your credit card number separately, because your credit card information (credit card number, name, social security number, credit information) This is because the usable limit, expiration date, etc. of the credit card are already entered in the package form.

During the credit transaction, the voice information for credit card payment is transferred to the Low-Pass Filter unit (LPF) 320 through the speaker and the voice recognition interface 310 to reproduce noise or distortion other than necessary voice information. The voice data comparison unit is then transferred.

The voice data delivered to the voice data comparator 330 is compared with the voice data comparator 330 to extract comparison data for confirming whether or not the voice of the client has been previously input to the client voice data storage 340. After extracting the characteristics of the input vowels and voiced sounds by analyzing the single vowels of vowels or similar voiced sounds in the formant extractor 400, the quasi-periodic characteristics of the vowels and voiced sounds and waveforms are maximum in the pitch extractor 410. Use to distinguish between voiced and unvoiced sounds. The unvoiced sound is non-periodic in waveform and relatively small in the analysis section.

The speech data passing through the pitch extracting unit 410 removes unnecessary silence portions from the real speech region detecting unit 420 and detects only the necessary speech portions. This is an important process for extracting client voice qualities. In order to ensure the characteristic of the consonant in the real speech section detector 420, the pre-emphasis process 430 is performed, and then the speech characteristic parameter detector 440 comprehensively detects the speech feature (usually represented by a number).

After the voice feature parameter detection is completed, the voice data of the client, which is previously input to the client voice data storage unit 340, is transmitted to the comparator 450. The comparator 450 compares the input voice data from which the voice feature parameter is detected with the data previously input to the client voice data storage unit 340 with the client's identity. If the two signals match, it is sent to the text generation unit 350 to generate characters suitable for the voice data, and output them in sequence to the display screen 370 of the mobile terminal credit card payment mode as shown in FIG. At this time, the character for each voice is linked to the database of the existing memory.

If the two data are inconsistent, the correction unit 360 outputs a correction message on the screen 370 and simultaneously outputs a simple sound through the voice recognition interface 310.

One transaction requires multiple inputs of numeric voices (usually 17-18), which requires continuous voice data input until a transaction statement is completed.

The role of the voice data comparator 330, which is the core of security, is to input only the voice data by a voice other than the client's own voice, so that only the error message is continuously output and the statement for payment itself as shown in FIG. 5 cannot be created. That is, as shown in FIG. 5, only an error message is output and the screen remains empty and no transaction is performed.

The speech recognition methods currently being studied or implemented are the most widely used, including Hidden Markov Model, Dynamic Time Warping, and Neural Network, which enable large-scale continuous speech recognition.

Since voice data required for credit card payment can be set to 0-9 only, the dynamic matching method is used. In addition, it is possible to compare the client voice characteristics more clearly by extracting the formant and pitch of the client voice to more accurately match the client. Of course, if a better voice recognition method emerges in the future, it is natural to adopt a better one as needed.

* Dynamic matching method is to compare the representative pattern with the given input pattern and determine the similarity between the two. Even when the same word is spoken, it has different durations according to the speaker's emotion, surroundings, and noise. Therefore, the non-linear optimization of the mismatches of the durations is based on the partial optimization. Has

Due to the dynamic programming, there is a large amount of calculation, and there is a constraint that we need to know the starting point and ending point of the voice in advance, but the speech data we use is composed of numeric characters from 0-9. First of all, dynamic matching, which can easily create a reference pattern, is suitable for services.

Since the voice data used is not at the level of small vocabulary (10-100), and each number is isolated one by one like calling a phone number when inputting the voice data for payment (Example 712342: seven, seven, two, 3) There is an advantage that the amount of voice database is significantly smaller than the voice recognition method used in other situations. .

The dynamic matching method uses the same voice when making the basic voice patterns of the clients from 0 to 9, and even if the same speaker has different durations depending on the speaker's emotions and the surrounding environment, the discrepancy between these durations is avoided. It is a nonlinear optimization method.

In order to secure the security, that is, inputting data by the voice of another client, as described above, the number of input data (0-9) that is dealt with in order to prevent a transaction from being established is utilized. Formant [Vocal tract fundamental frequency] for each data in the pattern: The band in which the spectral energy of a sound source is concentrated, or a group of bands, that specifies a certain tone. It is the peak of energy generated in several specific frequency ranges due to resonance when the vowels or the close voices of voiced sounds are analyzed by frequency. The center frequency of this region is called the formant frequency and is called low to 1st, 2nd, .formant.

The distinction between monolingual and verbal sounds depends heavily on the formant configuration. In addition, by making a database, the voice data comparator compares together to determine whether or not the speaker is identified. Also, the pitch part uses quasi-periodic characteristics of vowel and voiced sound and maximum value of waveform to distinguish between voiced and unvoiced sound. The unvoiced sound is aperiodic in waveform and relatively small in the analysis section.

4 is a block diagram of the voice data comparison unit 330.

The speech data comparator 330, which is a key part of solving the security of the present invention by applying a basic speech recognition method, includes a formant extractor 400, a pitch extractor 410, and a real speech section detector 420 (EndPoint Detection). ), A pre-emphasis unit 430, a voice parameter characteristic detector 440, and a comparator 450.

Formant extractor 400 analyzes the vowels or similar voices of single voices by frequency and extracts the characteristics of the input vowels and voiced sounds, and then extracts 410 the quasi-periodic characteristics of the vowels and voiced sounds and the maximum value of the waveform. Distinguish between voiced and unvoiced sounds. The unvoiced sound is aperiodic in waveform and relatively small in the analysis section.

The real voice section detecting unit 420 removes unnecessary silence from the recorded voice and detects only the necessary voice.

The pre-emphasis 430 process is performed to ensure the characteristics of the consonants in the real speech section.

As described above, vowels, voiced sounds and unvoiced sounds close to each other are separately patterned for formant of each voice data.

The comparator 450 compares the data of the real voice section detection unit 420 (EndPoint Detection) and the pre-emphasis 430 process with the voice data of the previously input client to determine whether the client is the user, If the input data is displayed on the display screen 370 through the voice recognition interface 310, and if there is a mismatch, a warning message is output through the correction unit 360 and the voice recognition interface 310.

5 is an exemplary diagram illustrating a process of selecting a credit card payment mode from a menu of a mobile terminal and outputting an error message when a client mismatch occurs.

6 is an exemplary view showing a form for creating a transaction history for credit card payment.

7 is an overall flowchart of how voice data for transaction is processed in a voice recognition system.

When the voice data for the transaction is input (700), the LPF 320 reproduces noise or distortion other than necessary voice information (705). The input voice data passing through the LPF 320 extracts the formant feature (710) by analyzing the vowels or similar single voices of the voiced voices by frequency and extracts the input vowel and voiced features, and then extracts the pitch feature (715). We distinguish between voiced and unvoiced by using the quasi-periodic characteristics of the overshoot and the maximum value of the waveform. The unvoiced sound is aperiodic in waveform and relatively small in the analysis section.

The voice data passing through the pitch extractor 410 removes unnecessary silence parts from the real voice interval detection unit 420 and detects only necessary voice parts (720). This is an important process for extracting client voice qualities. After the pre-emphasis process is performed in order to ensure the characteristic of the consonant in the real voice section detector 420 (725), the voice characteristic parameter detector 440 comprehensively detects the characteristic of the voice (usually represented by a number) (730). . After detecting the voice characteristic parameter, the voice data of the client, which is pre-inputted in the client voice data storage unit, is sent to the comparator.

The comparator 450 compares the input voice data from which the voice feature parameter is detected with the data previously input to the client voice data storage unit 340 with the client's identity. If the two signals match (740), the character generator 350 sends the text to the screen to generate a character suitable for the voice data (750) and outputs it to the screen (755). At this time, the character for each voice is linked to the database of the existing memory.

A single transaction requires multiple inputs of numeric voices (usually around 17-18). If the transaction details are completed or checked and completed (765), they are terminated. Otherwise, additional input is required. .

FIG. 8 is a flowchart illustrating a process of generating and storing client voice data regarding how the client's voice characteristics are inputted and managed in advance in a mobile terminal to check whether the voice data inputted for a transaction when using the voice recognition system. .

The customer who wants to use the service repeatedly inputs the voice of 1-9 to the portable terminal about 5 times for the number one after the other. The system can average this to extract more robust client voice qualities.

When the voice data of the client is input (800), the noise and the distorted portion are recovered as much as possible through the LPF 320 (805), and N = 0 is input to the COUNTER 803 (810). (This is a system necessary for repetitive input several times.) The input voice data is extracted formant feature (815) and the vowel or similar voiced sound is analyzed by frequency to extract the input vowel and voiced sound quality and then the pitch feature is extracted. By using the quasi-periodic characteristics of the vowel and voiced sound and the maximum value of the waveform, the voiced sound and the unvoiced sound are distinguished (820).

The unvoiced sound is aperiodic in waveform and relatively small in the analysis section. The speech data passing through the pitch extracting unit 410 removes unnecessary silence portions from the real speech region detecting unit 420 and detects only necessary speech portions (825). This is an important process for extracting client voice qualities.

After performing the pre-emphasis process in order to ensure the characteristic of the consonant in the real speech section detector 420 (830), the speech characteristic parameter detector 440 comprehensively detects the characteristic of the speech (usually represented by a number) (835). . When the voice characteristic parameter detection is finished, the N assigned to the input voice data is examined to determine whether it is greater than or equal to four. If it is less than 4 (850), the analyzed voice quality is stored in the temporary storage memory 812 (860), and the same input data is requested again as a client voice (855).

If N> = 4 is satisfied (840), the data input into the temporary storage memory and data satisfying N> = 4 are sent to the voice characteristic synthesis unit, and the average value of five voice characteristics of one voice data is synthesized (845). . When the voice data characteristic synthesis is completed, the synthesized data is sent to the client voice data storage unit 340 (865). The data stored in the client voice data storage unit 340 is used to determine whether or not the client is the voice data to be input during the credit card payment using the mobile terminal.

According to the present invention, by using a voice recognition system that is the most convenient and friendly interface to a person in a wireless credit card payment service using a wireless communication device, it is possible to simultaneously solve the convenience and security of the wireless card payment service. There is an advantage to a more sound credit settlement society.

In addition, by solving the ease of use, it can be a great help to generalize the use of wireless credit card payment service that emphasizes future mobility.In terms of security, the input data is about 20 times until the transaction history is written. It is superior to any security method because it continuously checks whether it is own.

Claims (2)

In a wireless communication device, Voice information input means for receiving voice information of the client owning the wireless communication device; Voice recognition means for analyzing voice data characteristics inputted through the voice information input means; Speech feature storage means for storing speech data features analyzed by the speech recognition means; In a financial transaction using the wireless communication device, after receiving predetermined financial information from the client through the voice information input means by voice, the voice data received for the financial transaction and prestored through the voice quality storage means. Voice data comparison means for analyzing and comparing client voice data characteristics for matching; Text conversion means for converting the voice data input for the financial transaction into text (or numeric) data when the voice characteristics of the two data match as a result of comparing the voice data comparison means; And And information data transmitting / receiving means for transmitting the financial transaction information data converted by the text switching means to the financial transaction server through a wireless communication network. According to claim 1, The wireless communication device, And a payment means storage and management means for storing at least one payment means information data of the client.