MXPA96004051A - Method and system to register addresses in a database for recognition of ha - Google Patents

Abstract

The present invention relates to a method and system for inscribing speed dialing names, which include providing partner dependent templates and associated telephone numbers and providing a penalized garbage model for unrecognized speech. When a request for a new template is received, it is determined whether the list of names for speed dialing is filled (Step 201) and if not, it is determined if that name is too similar (Step 205) to a name that is already in the list of marking with speed. If so, that name is rejected, but if it is not, it is determined if the speed dialing name is too short (Step 302) and if it is not too short or if the user wishes to provide the short name, the system asks the user who repeats the name for marking with speed and if there is correspondence, is supplied. If there is no correspondence, the system will change the first and second mentions and compare to see if there is a correspondence

Description

METHOD AND SYSTEM FOR REGISTERING ADDRESSES IN A DATABASE FOR SPEECH RECOGNITION TECHNICAL FIELD OF THE INVENTION This invention relates to speech recognition and more particularly to registering addresses for speech recognition in a database with speech recognition. BACKGROUND OF THE INVENTION The registration of name addresses in a database for speech recognition is used in speed dialing. Speed dialing is when for example a certain number or bank of telephone numbers are pre-stored and the user only has to address that bank of numbers by mentioning a name, in order to get the telephone number called. It is highly convenient for the user to dial with speed by mentioning the addresses by name to the telephone and to dial the telephone number associated with that name in the bank of telephone numbers. It is therefore convenient to provide some improved method and system for registering the named addresses for speed dialing to the telephone system, so that the correct numbers are dialed when they are mentioned in the telephone system. REF: 23138 BRIEF DESCRIPTION OF THE INVENTION According to a preferred embodiment of the present invention, a method and system for entering addresses as names in a database for speech recognition is provided, by providing a "garbage" model penalized for speech not recognized, receive a new mention for registration from a user and generate a template of the new mention. A repetition of the mention is then compared with the template, to determine if the new template mentioned will be accessed in the database. According to another preferred embodiment of the present invention, a method and system for entering names in a database for speech recognition include a database with dependent templates of the speaker and a penalized "garbage" model and comparing the name to enter the names in the database to reject any names that are too similar. According to another preferred embodiment of the present invention, determine if the name to be entered in the database is too short, before accessing it to the database. DESCRIPTION OF THE DRAWINGS Figure 1 illustrates a simplified block diagram of a telephone system implementing a method of the present invention; Figure 2 illustrates a flowchart of a method for generating speed dial addresses mentioned by multiple users in the telephone system with speech recognition, • Figure 3 illustrates a general flow chart of registering and deleting a directory name within the telephone system; Figure 4 illustrates a flow chart of a method for registering and modifying a speed dialing list that corresponds to a directory name in the telephone system; Figure 5 illustrates a flow diagram of an input addition for voice dialing, according to an embodiment of the present invention; Figure 6 illustrates a simple "garbage" model; Figure 7 illustrates an input addition flow diagram for voice dialing, in Figure 5; Figure 8 illustrates an input update flow diagram for voice dialing of Figure 7; and Figure 9 illustrates a flow chart for input addition retry for voice dialing of the Figure 8. DETAILED DESCRIPTION OF THE INVENTION FIGURE 1 is a simplified block diagram of a telephone system 10. The telephone system 10 includes a telephone 11 that connects to a processor 12. A pick-up detection circuit 13 and a circuit of recognition and recording 14, connect the telephone 11 and the processor 12. The processor 12 also connects to a memory 15. In operation, the off-hook detection circuit 13 informs the processor 12 that the telephone 11 indicates a hang-up condition and allows that the processor 12 verify commands according to a program stored inside and executed by the processor 12. The program inside the processor 12, allows a user to generate a directory name address and a list of entry names for speed dialing and corresponding telephone numbers associated with the address of the directory name. The telephone system 10 stores partner-dependent templates of the directory name address and associated entry names and telephone numbers, so that each user can access only this specific directory name and speed dial list. FIGURE 2 is an initial flow diagram of a method for generating speed dialing directories mentioned by multiple users in the voice recognition telephone system 10. The processor 12 in a mode is programmed according to that flow diagram. The off-hook detection circuit 13 of the telephone system 10 checks the telephone 11 in step 16, to detect a hang-up condition in the specific telephone. Once detection of a hang-up condition occurs, the processor 12 signals a user to feed a command in step 17. In step 18, the processor 12 in conjunction with the recognition and recording circuit 14 which may include a processor, includes a comparator and memory 15, compares the response of the user to one of a plurality of memory-coded templates 15 of the telephone system 10. The flexibility of the telephone system 10 allows receiving in stage 18, either words spoken or mentioned by a user or in some cases, corresponding DTMF pressure button codes of telephone 11, which represent spoken command words. Through the drawings, an asterisk indicates that the telephone system 10 can recognize either spoken command words or corresponding DTMF pressure button codes that represent the command words. The asterisks also indicate that the telephone system 10 performs independent speech recognition of the speaker, to fit a model to the response of a user. For illustrative purposes only, the description of the preferred embodiment will proceed as if the telephone system receives spoken responses instead of representative commands through corresponding DTMF oppression button codes.

In step 18, the telephone system 10 can recognize one of several command phrases and proceed according to the requested command. The telephone system 10 can recognize a telephone number in the stage 19 that is received from a user as a first command. The telephone system 10 informs the user in stage 20 of the received number and the sequence will continue to step 22, where the telephone number will be automatically dialed, in order to make the requested call. The telephone number 10 can also recognize an emergency command in step 24 such as "help", as a second command phrase received from the user. The telephone system 10 notifies the user in step 26 that the emergency telephone number such as 911 is dialed, and the sequence proceeds in step 22, where the call will be made again. The telephone system can also recognize a third command phrase, CANCEL of the user, which automatically returns the telephone system to step 17, interrupting any currently running script. For example, as illustrated in FIGURE 1, the user can stop sending a telephone call before a connection is made at the other end of the telephone line. Although illustrated only in a point in Figure 1, the recognition of a CANCEL command (CANCEL) in step 28, can occur anywhere within the method and telephone system described with reference in the subsequent figures. The telephone system 10 may recognize a fourth command word in step 30 where the user requests that the user's directory list be provided. When the telephone system 10 recognizes this command, the process flows to step 32, to allow the user to provide the directory option. FIGURE 3 illustrates a flowchart of the process steps in the directory option part of the telephone system program. Upon recognition of command, the user accesses the directory option in step 34. To ensure that only authorized users can access the directory option, the telephone system 10 implements security measures in steps 36 and 38, requiring the user provide a verification of the authority to access the directory option. The verification can be an authorization code that supplies the user to the system or can be speech recognition templates dependent on the speaker, so that they match the user's speech patterns to verification templates stored within the telephone system 10. In step 36, the telephone system 10 warns the user of proper verification and recognizes the user verification response in step 38. The telephone system 10 can perform stages 36 and 38 more than once as part of the process of check. If the telephone system does not recognize the verification code given by the user, the process flow returns to stage 17 of FIGURE 2 in a manner similar to the CANCEL command. If the telephone system 10 recognizes a valid verification code, the process flow continues to step 40, wherein the telephone system 10 notifies the user to feed one of four commands for the directory option. Also see Kero in U.S. Pat. No. 5,369,685 for user verification. Once the user has provided the appropriate verification and provides the directory address, the telephone system 10 can recognize a first word sub-command in step 42 to add a user directory name to the system. The telephone system 10 registers the user in step 44, by requesting a directory name and saves the user's response in a template in step 46, to be stored within the telephone system. By subscribing a user directory name, the telephone system 10 may repeat steps 44 and 46 in order to create the template and save it with the existing list of user identification templates already registered for that account or telephone. Once a template is saved, the process flow returns to step 40, where the telephone system 10 asks the user for another command word.

The telephone system 10 can recognize a second sub-command word in step 48, to delete a user directory name. When recognized, the telephone system 10 notifies the asuario in stage 50 of the name of the user directory to be deleted. The telephone system 10 recognizes the directory name given by the user in step 52 and requests that the user confirm the deletion of directory name in step 54. If the user does not confirm deletion of the directory name, the process flow returns to stage 40, where the telephone system 10 asks the user for a command phrase. If the user confirms deletion of the directory name in step 54, the telephone system 10 removes the template in step 56 created for that directory name and any telephone list entries corresponding to that directory name. Once deleted, the process flow returns to step 40, where the telephone system 10 asks the user for a new command phrase. The telephone system 10 can recognize a third sub-command phrase in step 58, to review the list of directory names. When recognized, the telephone system 10 plays the user directory list in step 60 before returning to step 40 to request a new command phrase. The telephone system 10 may also recognize a fourth command phrase in step 62, determining that the user has completed the request for directory option. When recognized, the process flow returns to stage 17 of FIGURE 2, where the telephone system 10 asks the user for a telephone number. Returning to FIGURE 2, the telephone system 10 can recognize a directory name in step 64, as a fifth command phrase. When the telephone system 10 recognizes a user directory name, the process flow proceeds to step 66 where the telephone system enters a speed dialing list option. FIGURE 4 is a flow diagram of the speed marking list option process of the present invention. The telephone system 10 enters the speed dialing list option in step 68 and subsequently asks the user in step 70 whether to request a name to dial or enter the list. When the telephone system 10 recognizes an entry name in step 72, a signal is given to the user in step 74, indicating the requested name to be called by the telephone system 10. The telephone system 10 then makes the call in step 22 of FIGURE 2 as previously described. The telephone system 10 can also recognize and access a list command in step 76 and signals the user of 1 of 5 list command phrases in step 78.

The telephone system 10 can recognize a first list command phrase in step 80, to add an entry name and telephone number to the speed dialing list under the user's directory name. When recognized, the telephone system 10 asks the user in step 82 to enter the entry name in the speed dialing list. The telephone system 10 warns the user in step 83 to enter a telephone number corresponding to the newly registered entry name in step 82. The telephone system 10 creates and saves a template corresponding to the registered name and telephone number. by the user in step 84. The telephone system 10 may repeat steps 82, 83 and 84 in order to verify and create a valid template of the entry name and telephone number for the speed dialing list. Once saved, the process flow returns to the start of the speed dialing list option routine in step 70. The telephone system 10 can recognize a second list command in step 86, to modify a telephone number which corresponds to an entry name. When it is recognized, the telephone system 10 warns the user in step 88 to provide the name whose telephone number is to be modified. The telephone system 10 recognizes the name given by the user in step 90 and allows the user to modify the telephone number corresponding to that name in step 92. The telephone system 10 keeps a template of the modified number, which corresponds to the entry name with which modification of step 94 was requested. Telephone number 10 may repeat steps 92 and 94 to ensure valid creation of the telephone number template. Once the template is saved, the process flow returns to step 70, as previously described. The telephone system 10 may recognize a third list command in step 96 to remove a name from the speed dial list. When recognized, the telephone system 10 notifies the user in step 98 of the name to be removed from the speed dialing list. The telephone system 10 recognizes the name in step 100 and requests that the user confirm deletion of the name in step 102. If the user does not wish to delete the entry name, the process flow returns to step 70. If the user does not confirm deletion of the entry name, the telephone system removes the entry name template and the corresponding telephone number template in step 104 before directing the process flow back to stage 70. The telephone system 10 it can recognize a fourth command phrase list in step 106, to allow a user to check his speed dial list. When recognized, the telephone system 10 reproduces the dialing list with user speed in step 108 and returns the process flow, back to step 70. The telephone system 10 can also recognize a fifth command phrase of list in step 110, indicating that the user has completed the speed dialing list option. When recognized, the process flow returns to stage 17 of FIGURE 2, and the method is repeated as previously described. In summary, a telephone system can generate a separate directory for each authorized user of the telephone system. Each user can create a speed dialing list that contains names and phone numbers under the user's own directory. When using dependent characteristics of the interlocutor, nobody can have access to the directory of an authorized user or speed dialing list. The above is by way of background to list speed dialing names using speech recognition. The processor 12 in FIGURE 1 according to one embodiment of the present invention is programmed to operate in accordance with the flow chart of FIGURE 5 to register or inscribe speed dialing names in a speed dialing list. Processor 12 includes the ULPC and ELPC counters and allows a subscriber to create a base phrase and then update it. The subscriber has three opportunities to mention the name to access it in a list, in order to better recognize the name mentioned. The system also addresses the subscriber problem that adds a name to the list that is already in the list or adds a name very similar to the list. It also addresses the problem that the subscriber pronounces the name too differently as it is registered and updated. Recent developments in the use of trash model to determine speech outside of vocabulary, have led to new recognition processes that provide an ability to recognize outside of vocabulary as well as retain a high speed of recognition in vocabulary. This new recognition process uses a penalized garbage model, in parallel with mentioned speed dialing names, to discriminate out-of-vocabulary speech. This approach applies to recognition with speed marking inscription to address the problem of entering names that are already on a speed dial list and too much variability during registration. A "garbage model" is defined as a model for any speech that may be words or sounds for which there is no other model within the recognition system. There are several possibilities for ways to build garbage models. A single garbage model commonly used in recognition of the state-of-the-art, illustrated in FIGURE 6, models a collection of broad phonetic classes of speech sounds, which are also linked to form sounds that constitute a word. As illustrated in FIGURE 6, the circles represent the broad acoustic phonetic classes. Solid lines indicate transitions that can be made in any direction from one broad phonetic class to another. Dotted lines indicate that the model can cycle in a particular state. Transitions are weighted by probabilities based on temporary phonatactic constraints. These restrictions require that the more a particular phonics class is used to explain speech, the less likely the class will be used to explain speech, the less likely the class will be used to explain subsequent speech, the more likely subsequent speech will be explained by other phonics classes different The model can begin to explain speech when accessing or leaving in any state. During verification of a similar name, recognition is performed with the new name that is added to the list. The new name can already correspond to an existing name in the list, or correspond to the garbage model in parallel. If the name corresponds to an existing name, the user is then informed that the name or a similar name is already in the list, and that the name will not be added. If the new name corresponds to the garbage model in parallel, then it is considered that the name is not in the list and the addition of the name continues. The penalty in the garbage model can be adjusted to accept the sensitivity to correspond to either a name on the list or the garbage model. The trash model, preferably, may be similar to the phonotactic garbage model discussed in the Patent Application Serial No. entitled "Enrollment and Modeling Method and Apparatus for Robust Speaker Dependent Speech Models" (Method and Apparatus for Registration and Modeling for Robust Speech Dependent Speech Models) by Lorin Netsh et al., Presented concurrently with this. This model has language restrictions. The subscriber when attempting to enter a new name to the dialing list with speed, enters a menu entitled "Handling Voice Dial List" or position 40 in FIGURE 3 or 78 in FIGURE 4, "and provides or mentions" Add Entry n- (Add Entry) when this command is recognized, the system first verifies in step 201 whether the list is full or not. If it is full, the system notifies the user that it is full. This can be done by a synthesized voice command of memory 15 and synthesizer 15a which states "Your list is full, you must delete a name before adding a name before adding a new one". The user can return to the list manager, and delete a name in the list. See steps 96 to 104 in FIGURE 4. If the list is full or after deleting a name in the list and returning to "add LPCCNT entry", the ULPC and ELPC counters are set to zero (step 202). The system will then maintain the LPC enrollment account (linear predictive code) or ELPC and the Linear Predictive Coding Update (ULPC) accounts. The LPC is a speech sample represented by linear prediction parameters. The LPC is considered linear. For more about LPC, for example see pages 81 to 124 in "Linear Predictive Coding of Speech "(Linear Predictive Speech Coding) by Bishnu S. Ital (chapter 4) in" Computer Speech Processing "(Computer Speech Processing) edited by Frank Fallside and William Woods, Prentice Hall (ISBN 0-13- 163841-6) .This is incorporated herein by reference.If the ELPC and LPC account are both 0 (stage 203 ) indicating that nothing has been previously added, a tutorial, a synthesized signal declaration ("yes" decision (yes) in step 203) is reproduced.) The synthesized statement may read, "The system requires learning as you say the name. There will be a long pause after you say the name the first time, and then the system will ask you to repeat the name between one and four times. In the future, you can skip this message by dialing the number sign. After the signal, please mention the name for voice call. " (An audible signal (Beep) is heard at the end for the user to mention the name). The user mentions the name. A check is made in step 405, to determine if the name is already in the list (for example) of fifteen names listed or close to a name or corresponds to the garbage model. If there is a correspondence to that in the list (indicating that a similar name already exists in the list) the system enters Retry Add Entry from step 206. In the case that the attempts are finished it is "yes" (yes) ( step 207), then the system returns to the voice dial list management menu or position 40 in figure 3 or 78 in figure 4. If the attempts were not finished ("no" in step 207) the system plays by synthesizer 15a a message "too similar" ("... is too similar to another name in your list, please choose a different name") and increase the counter ELPC 202 illustrated in Figure 5 and use the shortest signal ("after the signal, please say the name for voice dialing") to try again. If the user does not give a new name in time, the system reaches the end of the interval and increments the counter and requests a new name. If a key is pressed on the keyboard, which is not valid and / or after reaching the end of the interval, the system increments the counter in step 202 and asks for a name to be mentioned. If the system fails to enroll after three attempts or a fifth invalid DTMF key, the system disconnects (step 209). If the aforementioned name does not correspond, the system changes to "Add Entry Instruction" of Figure 7. After a successful mention of a name that has no correspondence in Enroll Add Entry in Figure , the system follows the flow chart of Figure 6. The mention is stored when online registration is initiated.

The saved mention is used to create a template (stage 301) when making an offline registration. If the duration of the mention is not too short such as greater than or equal to a minimum threshold such as for example 10 data frames ("no" in step 302) the system advances to step 305 to update add entry and follow the flow diagram of the Figure 8. If the message is too short or less than the minimum threshold (less than 10) data tables for the example ("yes" in step 302), the system requests by means of the synthesizer, if the user wishes to use the template and even if it has not been used before. In the case of "if", which means that it is less than the minimum threshold, the signal message may say: "The name [given name] is shorter than the recommended name length. To use this name in any way, say CORRECT (OKAY). To have this name canceled, say CANCEL. This is followed by an auditory signal. If CORRECT (OKAY) is received in response stage 307, the system proceeds to the Update Add Entry of Figure 8. If "CANCEL" is received, a synthesized declaration is generated and provides such as "name not added "and the system proceeds back to the Voice List Management menu or position 40 of Figure 3 or 78 in Figure 4. If nothing is mentioned (the end of the interval is reached) an unrecognized command or an incorrect key is pressed, the system provides synthesized instructions and returns to find an answer. If after 5 times there is no recognized response or after three completed periods of time, the system is disconnected with a message (step 309). If a DTMF key is pressed, the synthesizer provides the "wrong key" message. After each period of time period, each wrong key and after the third and fourth unrecognized voice commands, the synthesizer can set "say right (okay) or cancel (cancel)" or for more detailed instructions, "say CORRECT ( OKAY) to continue adding this name, say cancel to cancel adding this name ". If "OK" is recognized, even if it is a short name, the system advances to the Update Add Input of Figure 8.

To determine recognition, the system uses the garbage model with the penalties listed below. start_garbage _pssd). _garbage_pssd, 0.6 - > _rhot, sl_rhot. _garbage_pssd, 0.6 - > _backv, sl_backv. _garbage_pssd, 0.6 - > _frontv, sl_frontv. _garbage_pssd, 0.6 - > _fric, sl_fric. _garbage_pssd, 0.6 - > _nasal, sl_nasal. _garbage__pssd, 0.6 - > _stop, sl_stop. _garbage_pssd, 0.6 - > _sib, sl_sib. _garbage_pssd, 0.6 - > _lowv, sl_lo v. sl_rhot, 6e-06 - > SSS6. s2_rhot, 0.06 - > 66SS. s3_rhot, 0.6 - > S6SS. sljbackv, 6e-06 - > ßß6ß. s2_backv, 0.06 - > 66S6. s3_backv, 0.18 - > 66ßß. s4_backv, 0.3 - > 6SSS. s5_backv, 0.6 - > 6666. sl_frontv, 6e-06 - > ßß66. s2_frontv, 0.06 - > ßß66. s3_frontv, 0.18 - > 66Sß. s4_frontv, 0.3 - > "". s5 frontv, 0.6 - > ßßßs. sl_fric, 6e-06 - > 666S. s2_fric, 0.06 - > 6S. "S3_fric, 0.18 -> S666, s4_fric, 0.3 -> 66ss, s5_fric, 0.6 -> 66SS, sl_nasal, 6e-06 -> 666ß, s2_nasal, 0.06 - > sßßs s3_nasal, 0.6 - > 6ßßí. Sl_stop, 6e-06 -> 6ßss. S2_stop, 0.06 - > sßS6. S3_stop, 0.6 - > ß6SS. Sl_sib, 6e-06 - > sß66. S2_sib, 0.06 - > ßß66. S3_sib, 0.18 -> SS6S, S4_sib, 0.3 -> sßßs, s5_sib, 0.6 -> sßßß, sl_lo v, 6e-06 -> ssss, s2_lowv, 0.06 -> gß6ß, s3_lo v, 0.18 - > 6 ßs. s4_lo v, 0.3 -> S66S. s5_lowv, 0.6 -> 6SSß If the received template has a high rating for any of the garbage models listed, it receives a high non-recognizable speech rating and is rejected as unrecognized speech.

With reference to Figure 8, the template is downloaded (step 401) and the system determines if there has been an update. During update, the garbage model is used to explain speech that is not in the enrollment template. For example, if during registration the user says "uh, Roger Rabbit", then the garbage model explains the "uh" ("one breath") and only the "Roger Rabbit" portion of the update mention is averaged in the template again name. If the update counter (ULPC) is zero (step 402), this means that we have not made an update and the system asks the user through the synthesizer, that "please say the name again". The update counter is incremented (step 403) and when there is a response, an update of the template is performed (step 404) using that response. The template is verified to determine if a good update occurred. If a good update occurs, the user is asked to provide the telephone number for that name (step 405). This can be typed or mentioned, using voice recognition with independent recognition models of the interlocutor. If the update fails, the system advances to the stages Retry Add Input of Figure 9. If the update fails, the update and enrollment mentions are changed and registration and update are attempted in that order. Often, a user is not ready to mention the first time and so an insertion such as "uh" (a "puff") can probably occur before the name is mentioned but when we ask to be spoken again, they are ready to say it . The first template has the "uh" address and when an update is made, the update may fail because there is no "uh". When mentions are changed, the second cleanest mention is used to register and update with the first mention, so that the "uh" breath at the beginning of the mention is explained by the garbage model and the "uh" is not included in the template. If this change of the first and second mentions fails, a third mention is requested for termination of attempts (step 406) is requested and the response and the second mention are used for the update. If a third mention is requested to register, then that name is checked first to see if it is too similar to another name on the list. If not used, processing proceeds to feed A in Figure 5. If the enrollment fails because the mention was too short, the system will notify the subscriber and will signal again for another mention. If enrollment is successful, but the mention (frame length) is too short (it is less than the minimum length threshold), the subscriber will be given a warning that poor recognition results may be obtained because the registration name It's too short. The subscriber is notified to say "CORRECT" (OKAY) OR "CANCEL" (CANCEL). In short, if an update fails, then the mentions are changed, to see if the second mention (or third if required) make a better mention of registration than the first. The following order of registration and updates is attempted but only a maximum of three mentions of the user are requested. Inscriptions Update 1 2 2 1 2 3 3 2 3 1 1 3 OTHER MQDALIDApgS Although the present invention and its advantages have been described in detail, it will be understood that various changes, substitutions and alterations can be practiced, without departing from the spirit and scope of the invention. invention as defined by the appended claims. It is noted that in relation to this date, the best method known to the applicant to carry out the aforementioned invention, is that which is clear from the present description of the invention.

Having described the invention as above, property is claimed as contained in the following:

Claims (37)

1. CLAIMS 1. - A method for registering speech recognition models in a speech recognition database, characterized in that it comprises: providing a penalized "garbage" model, to explain strange speech, -receiving a new mention of speech recognition to register for a user; generate a template of the recipient's mention for registration, - request that the user repeat the mention to enroll again, - receive a second acknowledgment received; compare the second mention to the generated template and the garbage models X to determine if there is correspondence; and add the new template to a speed dial list if there is correspondence with in-vocabulary speech.
2. 2. A method according to claim 1, characterized in that the comparison step includes the step of comparing the second mention with the garbage model penalized to reject at the second mention, any mention corresponding to the predetermined degree of the model of Speech penalized, in terms of out-of-vocabulary speech, compared to the registered template.
3. 3. - A method according to claim 1, characterized in that it includes the step of changing the template and the second acknowledgment received, if the comparison fails to correspond and repeating the comparison stage.
4. 4. - A method according to claim 3, characterized in that it includes the step of requesting and receiving a third mention if after the change stage, it fails to obtain a comparison and the third response and the second mention are compared and if there is a correspondence by providing a template of the second mention in the database.
5. 5. - A method according to claim 3, characterized in that it includes the step of requesting successive mentions, if after changing it fails to obtain a comparison of the previous mentions, and the successive response with a previous mention is compared and if there is a correspondence by providing a template of successive mention to the database.
6. 6. - A method for inscribing speed dialing names in a telephone system characterized in that it comprises: providing a garbage model penalized to explain strange speech, - receiving a new name mention for speed dialing, to register a user, -generate a name mention template for speed marking received for enrollment, - request that the user repeat the name referencing for speed marking again to enroll, - receive a second rename mention for speed marking received; compare the second mention of name for marking with velocity to the generated template and the garbage models penalized to determine if there is a correspondence; and add the new template to a speed dial list if there is correspondence.
7. 7. - A method according to claim 6, characterized in that it includes the step of requesting and adding a telephone number to be associated with the new speed dial template.
8. 8. - A method according to claim 7, characterized in that it includes the step of changing the template and the second mention of received speed dialing name if the comparison fails to correspond and repeating the comparison step.
9. 9. - A method according to claim 8, characterized in that it includes the step of requesting and receiving a third mention of name for marking with speed, if after the change stage fails to obtain a comparison and the third response and the second Mention is compared and if there is a match by providing the second measurement to the speed dialing list.
10. 10. The method of compliance with the claim 8, characterized in that it includes the step of requesting and receiving successive new name mentions for marking with speed, and if after changing it fails to obtain a comparison with previous mentions, and the successive answers and a previous mention are compared, and if there is a correspondence when providing a template and a successive mention to the database.
11. The method according to claim 6, characterized in that the comparison step includes the step of comparing a mention to the garbage model penalized to reject at the second mention, any mention that corresponds within the predetermined degree of the garbage garbage model.
12. The method according to claim 11, characterized in that it includes the step of exchanging the template and the second received mention of name for marking with speed, if the comparison fails to correspond and repeating the comparison step.
13. The method according to claim 12, characterized in that it includes the step of requesting and receiving a third mention of a new speed-dialing name, if after the change stage it fails to obtain a comparison and the third mention and the second mentions are compared and if there is a correspondence by providing the second mention, to the speed dialing list.
14. The method according to claim 10, characterized in that it includes the step of requesting and receiving successive mentions of new speed-dialing names, and if after changing it fails to obtain a comparison in previous mentions, and the successive responses and A previous mention is compared and if there is a correspondence provide a template of a subsequent mention to the database.
15. 15. A method for registering addresses in a speech recognition database, characterized in that it comprises the steps of: providing dependent templates of the name address partner, - providing a garbage model penalized to explain strange speech; and determine if the address to be registered corresponds with either a dependent template of the interlocutor previously provided or the model of garbage penalized and if there is correspondence with the dependent template of the interlocutor previously provided, then reject the registration.
16. 16. A method for inscribing name for speed dialing in a telephone system, characterized in that it comprises: providing dependent templates of a partner, name for dialing with associated speed and telephone number, - providing a garbage model penalized for unrecognized speech; and determine if the name to be registered corresponds with either a previously provided interlocutor dependent template or the garbage model that is penalized and if it corresponds to a previously provided interlocutor dependent template, then reject the registration.
17. 17. The method in accordance with the claim 16, characterized in that it includes the step of: determining if what is mentioned to be registered is less than a minimum length threshold.
18. 18. The method of compliance with the claim 17, characterized in that if the mention is less than the minimum length threshold, determine the approval of the user before adding the template of the mention to the speed dialing list.
19. 19. A method for registering addresses in a database with speech recognition, characterized in that it comprises the steps of: providing template of dependent addresses of interlocutor; provide a model of criminalized garbage for unrecognized speech; receive the address to register; and determine if the address to enroll is too short.
20. 20. A method for registering names for dialing with speed in a telephone system, characterized in that it comprises: providing name-dependent templates for speed dialing and associated telephone numbers; provide a model of garbage penalized for unrecognized speech, - receive a new markup with speed to register a user, - generate a template of the name registration for marking with speed received, - determine if the name to be registered it is too similar to a template dependent on the interlocutor already provided, and if it is too similar, reject the registration or if it corresponds to the garbage model, then allow registration; determine whether the mention to enroll is less than a minimum length threshold and if the mention is less than the minimum length threshold, determine the approval of the user before adding the mention template to the market list with speed; request the user to repeat the new mention of the name to mark with speed to register; receive a second mention of the name for marking with received speed, - compare the second new mention of the name with speed marking to the generated template and the garbage models penalized to determine if there is correspondence; and add a new name template for speed dialing to the dialing list with speed, if there is correspondence.
21. 21. The method according to the claim 20, characterized in that it includes the step of requesting and adding a telephone number to associate with the new name template for speed dialing.
22. 22. The method of compliance with the claim 21, characterized in that it includes the step of changing the template and the second mention of name for marking with received velocity if the comparison fails to correspond and repeating the comparison step.
23. 23. The method according to claim 22, characterized in that it includes the step of requesting and receiving a third new mention of name for marking with speed if after the change stage fails to obtain a comparison and the third response and the second mention is compared and if there is correspondence by providing the second mention to the list for speed dialing.
24. The method according to claim 20, characterized in that it includes the step of comparing the second mention to the penalized garbage model, to reject at the second mention any mention corresponding to the predetermined degree of the garbage garbage model.
25. 25. A telephone apparatus for entering market names with speed, characterized in that it comprises: a memory storage of partner dependent templates, speed dialing names and associated telephone numbers; a garbage model memory storage penalized for unrecognized speech; and a comparator for the name to be entered in the dependent template of the stored interlocutor or the garbage model penalized to reject the registration if it is too similar.
26. 26. The telephone apparatus according to claim 25, characterized in that it includes means for determining whether the mention to be entered is less than a minimum length threshold or if the mention is less than the minimum length threshold that determines the approval of the user before adding the template of the mention to the list of marked with speed.
27. 27. A telephone apparatus for registering speed dialing names, characterized in that it comprises: a storage device that stores a criminalized garbage model for unrecognized speech, - a receiver for a new speed dial name reference to register a user; a generator coupled to the receiver to generate a template of the mention of the name for marking with speed received for registration, - means for requesting the user to repeat the new marking with speed again, to register; the receiver in response to obtaining the new name mention for marking with received velocity, compares the second mention of the name for speed marking to the template generated and the garbage models penalized to determine if there is correspondence, - and means to add the new name template for speed dialing to a dialing list with speed, if there is correspondence.
28. The telephone apparatus according to claim 27, characterized in that it includes means for requesting and adding a telephone number to be associated with the new name template for speed dialing.
29. 29. The telephone apparatus according to claim 27, characterized in that it includes means for changing the template at the second mention of the name for marking with received velocity, if the comparison fails to correspond and again compare.
30. 30. The telephone apparatus according to claim 29, characterized in that it includes means for requesting and receiving a third mention of the new speed-dialing name, if after the exchange stage it fails to obtain a comparison and the third response and the Second mention is made and if there is a correspondence by providing the second mention to the speed dialing list.
31. The telephone apparatus according to claim 27, characterized in that the receiver includes means for comparing the second mention to the penalized garbage model, to reject at the second mention any mention corresponding within a predetermined degree to the garbage model penalized .
32. The telephone apparatus according to claim 31, characterized in that the receiver includes means for changing the template and the second mention of the dialing name with received speed, if the comparison fails correspondingly when repeating the comparison stage.
33. 33. The telephone apparatus according to claim 32, characterized in that the receiver includes means for requesting and receiving a third new mention of name for speed dialing if after the change stage it fails to obtain a comparison and the third mention and the second mentions are compared and if there is a correspondence proportional to the second mention to the speed dialing list.
34. 34. An apparatus for inscribing speech templates in a database for speech recognition, characterized in that it comprises: a storage device that stores a garbage model penalized for unrecognized speech; a receiver for a new mention of speech address, to register in the database from a user, - a generator coupled to the receiver, to generate a template of the speech address mention received for registration; means to request that the user repeat the new mention for speech address again to enroll; the receiver in response to receiving the second new mention of received speech address, compare the second mention to the template generated and the garbage models penalized, to determine if there is correspondence; and means to add the new template to the database if there is correspondence of the new name template for speed dialing.
35. 35. The telephone apparatus according to claim 34, characterized in that the receiver includes means for changing the template and the second received indication if the comparison fails to correspond and again compare.
36. 36. The telephone apparatus according to claim 35, characterized in that the receiver includes means for connecting and receiving a third mention, if after the change stage it fails to obtain a comparison if the third response and the second mention are compared and if there is a correspondence proportional to the second mention to the database.
37. 37. The telephone apparatus according to claim 34, characterized in that the receiver includes means for comparing the second mention to the model of garbage penalized, to reject at the first mention any mention that corresponds within a predetermined radius to the trash model. penalized.