JPH04254896A - Speech recognition correction device - Google Patents

Abstract

PURPOSE:To obtain a voice recognition correction device making the best use of its excellent characters without losing the simplicity and quickness that is excellent characters of speech input. CONSTITUTION:A voice recognition correction device recognizes a first input voice, and stores the sound analysis result and the voice recognition result. When it becomes necessary to change the voice recognition result, an operator inputs the changing part as a second voice. In a spotting part 14, matching is carried out by using the sound analysis result and the speech recognition result of the first input and the second input, and the spotting of the changing part is carried out for changing it.

Description

[Detailed description of the invention]

0001

[Industrial field of application] Related to speech recognition devices.

0002

2. Description of the Related Art A conventional speech recognition and correction device will be explained with reference to FIG.

[0003] In a conventional speech recognition and correction device, an acoustic analysis section 1 analyzes input speech and outputs characteristics. Based on the output from the acoustic analysis section 1, the speech recognition section 2
Recognizes the input voice. The result recognized by the voice recognition unit 2 is output so that the operator can confirm it. The result of feature extraction obtained by analyzing the input speech is stored in the storage unit 3. If there is no need to change the voice recognition result, the process moves to the next voice input operation. However, if it is desired to change the voice recognition result, the user operates the cursor indicator 4 to move the cursor to the beginning of the part to be changed and the end of the part to mark it, thereby specifying the part to be changed. In the conventional speech recognition correction device, the operator operates the cursor indicator 4 to indicate the part that needs to be changed due to misrecognition or the like. In order to change the part indicated by the operation of the cursor indicator 4, the operator can use the key input device in the changing unit 5 to change the part, or if several candidates for speech recognition results are displayed. changes the speech recognition result by selecting a candidate from among them.

0004

[Problem to be solved by the invention] Voice input allows data input without performing key input operations, and there is no need to know the key layout positions of the key input device, key operation methods, etc., and anyone can do it. This is an easy-to-use input method. However, unlike the input method using keys, the voice input method tends to have a slightly lower probability that the input data will be accurately recognized even if the operator inputs the data accurately. Therefore,
When we looked at the number of phonemes that make up one word for the words that make up the sentences in the Journal of the Acoustical Society of Japan, we found that the average number of phonemes per word was about 9 phonemes. Here, the phoneme recognition rate is 9
Even with a speech recognition device that has a rate of 5% or more, only 1 per word is used.
Errors in character recognition are to be expected. Also, the phoneme recognition rate is 98
% or more, it is considered that it is not so rare that a single character recognition error occurs in one word. Also, when we looked at the number of phonemes that make up one sentence for sentences in the Journal of the Acoustical Society of Japan, we found that 1
The average number of phonemes per sentence was approximately 47 phonemes. Here, even if a speech recognition device with a phoneme recognition rate of 95% or more is used,
Misrecognition of two to three phonemes in one sentence is unavoidable. Further, even when considering a speech recognition device with a phoneme recognition rate of 98% or more, it is quite possible that a recognition error of one phoneme will occur in one sentence. Considering the above facts,
It can be seen that it is not uncommon for it to be necessary to change the recognition results of input speech. When considering speech recognition devices, the role of speech recognition correction devices is considered to be extremely important. Furthermore, the speech recognition correction device is used not only when correcting errors in speech recognition, but also when it is desired to change part of input data.

[0005] When it becomes necessary to make changes to the speech recognition results, it is important to first correctly specify the changes. However, in the speech recognition and correction apparatus using the prior art described above, the operator must himself operate the cursor designator 4 shown in FIG. 9 to specify the section in which a recognition error or change has occurred. It is very burdensome for the operator to have to frequently operate the cursor using such a key input device. Work that requires frequent movement of the line of sight is
This not only greatly fatigues the operator's optic nerves and reduces work efficiency, but also may cause a significant decrease in visual acuity. Also,
While using the excellent input method of voice input, the conventional technology requires manually moving the cursor to specify the changed part, but the voice recognition device's characteristic "
Anyone can easily operate it. ” and “data entry speed is faster than key entry or other methods.” In other words,
Even if data input operations can be performed quickly, it is extremely difficult for anyone to easily and quickly change the voice recognition and correction apparatus using the conventional technology. As described above, in the speech recognition and correction apparatus using the conventional technology, the change operation takes a very long time, and the work efficiency is extremely low. The conventional technology has the problems described above.

[0006]

[Means for Solving the Problems] A speech recognition and correction device of the present invention includes an acoustic analysis section that outputs characteristics of an input first speech, and a speech recognition section that converts the output of the acoustic analysis section into a code string. an acoustic analysis storage section that stores the output of the acoustic analysis section; a speech recognition storage section that stores the output of the speech recognition section; and an acoustic analysis storage section that stores the input second speech and the speech recognition storage section. a spotting unit that detects a portion corresponding to the second voice from data of the first voice in comparison with data in the first voice; and a changing unit that changes a portion of the code string that corresponds to the corresponding portion. It is characterized by consisting of.

[0007]

EXAMPLES The present invention will be explained in detail below based on examples. (Embodiment 1) FIG. 1 is a block diagram of the principle of a speech recognition and correction device according to the present invention, and FIG. 2 is a block diagram of a speech recognition and correction device that generates speech by dividing it into words according to an embodiment of the present invention. The processing path indicated by the dashed line in FIG. 2 is that of the second voice. The first voice that is divided into words and uttered is sent to the voice input unit 21 in FIG. 2, which is composed of a microphone, a high-frequency emphasis filter, and an AD converter, which are the components of the acoustic analysis unit 11 in FIG.
It is sampled as an 8KHz, 12-bit digital signal. Furthermore, the same figure 1 acoustic analysis section 11
In the feature extraction circuit 22 in FIG. 2, which is a component of the digital signal, the frequency of the audio signal converted into a digital signal is converted frame by frame, with a 16 ms interval as one frame, and the feature parameters in the frequency domain are extracted. is expressed as a feature parameter sequence. The first audio feature parameter string extracted by the acoustic analysis unit 11 in FIG. 1 is stored in the feature parameter string storage circuit 27 in FIG. The feature parameter storage circuit 27 in FIG. 2 constitutes the acoustic analysis storage section 16 in FIG. Furthermore, the first speech feature parameter string is matched with the phoneme memory dictionary 24 in FIG. 2 in the DP matching circuit 23 in FIG. 2 which constitutes the speech recognition unit 12 in FIG. The phoneme lattice recognized and determined by the speech recognition unit 12 in FIG. 1 is stored in the phoneme lattice storage circuit 32 in FIG. 2, and displayed on the display unit 26 in FIG. The phoneme lattice storage circuit 32 in FIG. 2 constitutes the speech recognition storage section 13 in FIG. If there is an error in the voice recognition result displayed on the display unit 26 in FIG. 2, or if there is a part that should be changed, the operator notifies the user of the need to change the recognition result by touching a correction key or the like. The speech recognition correction device prepares for the second speech input by the transmission from the correction key control circuit 28 in FIG. The broken line in FIG. 2 shows the processing of the second input voice. The phonemes that are the input second voice are as shown in Figure 2, similar to the first voice.
It is converted into a feature parameter string via the voice input section 21 and the feature extraction circuit 22 shown in FIG. First, the feature parameter string of the second voice is subjected to DP matching in the DP matching circuit 29 of FIG. 2 with the feature parameter string of the first input voice stored in the feature parameter string storage circuit 27 of FIG. By inputting the incorrectly recognized phoneme or the phoneme that needs to be changed as the second voice, the changed part is reliably spotted. Spotting the changed parts obtained by this process 1
shall be. Furthermore, in order to ensure spotting of the changed part, the second voice is also used in the DP matching circuit 23 in FIG. 2 in the same way as the first voice.
P matching and voice recognition. The result of voice recognition of this second voice and the voice recognition result of the first voice stored in the phoneme lattice storage circuit 32 of FIG. 2 are subjected to DP matching in the DP matching circuit 33 of FIG. This is a process of spotting changed parts using the voice recognition results. The changed portion obtained by this process is referred to as spotting 2. Next, in the spotting result comparison circuit 34 shown in FIG. 2, the changed portion was spotted using the first voice feature parameter string and the second voice feature parameter string stored in the feature parameter string storage circuit 27 shown in FIG. As a result, spotting 1 is compared with spotting 2, which is the result of spotting the changed part using the voice recognition results of the first voice and the voice recognition results of the second voice stored in the phoneme lattice storage circuit 32 in FIG. do. As a result of comparison in the spotting result comparison circuit 34 of FIG. 2, if spotting 1 and spotting 2 are the same, it is assumed that the changed portion has been definitely spotted, and the process moves on to the next process. If spotting 1 and spotting 2 do not match, the one with the larger DP value when spotting 1 and spotting 2 are matched by DP matching is determined to be the changed part. If the changed part is spotted by mistake, the operator can check the changed result on the display section 26 in FIG.
You can find out by seeing what is displayed on the screen. In such a case, the correction key controlled by the correction key control circuit 28 in FIG. 2 is immediately operated, and the spotting result comparing circuit 34 in FIG. The other spotting result that was not selected as a candidate for the changed part is used as the changed part, and the changing operation for the changed part is proceeded again. The DP matching circuit 29 in FIG. 2, the DP matching circuit 33 in FIG. 2, the spotting result comparison circuit 34 in FIG.
4. Regarding the changed part spotted by the above processing, the second candidate phoneme stored in the phoneme lattice storage circuit 32 in FIG. 2 is rearranged as the first candidate phoneme by the phoneme lattice replacement circuit 30 in FIG. FIG. 2 is controlled by the display control circuit 25.
It is displayed on the display section 26. When the changing unit is changed to the desired phoneme by this operation, the user operates the enter key controlled by the enter key control circuit 31 in FIG. 2 to confirm the speech recognition result. If the second candidate phoneme in the change part is not the desired phoneme to be changed, the confirm key control circuit 30 in FIG. 2 recognizes that the confirm key is not operated, and the phoneme lattice replacement circuit 30 in FIG. The candidate with
to be displayed. When the operator finds the desired phoneme, he or she operates the confirmation key to confirm the speech recognition result. If the desired phoneme does not exist among the displayed second, third, and lower candidates, the voice is input again and voice recognition is performed using the same route as when the voice was first input. Figure 2 Phoneme lattice exchange circuit 30, Figure 2 Determining key control circuit 3
1 constitutes a modification unit 15 in FIG.

Based on the embodiment of the present invention, FIG.
This will be further explained using FIG. 4.

FIG. 3 shows an embodiment of the speech recognition and correction apparatus of the present invention, which constitutes a speech recognition and correction apparatus in which each word is uttered separately. FIG. 3 is a diagram showing audio power that is part of the feature extraction results stored in the acoustic analysis storage unit 16. FIG.

[0010] In a speech recognition and correction device that recognizes speech uttered by dividing it into words, an operator uses an "electronic computer"
Assume that the word ``1'' is input as the first voice. When this input is received, the phoneme lattice stored in the phoneme lattice storage circuit 32 in FIG. 2 is as shown in FIG. ”
It is. The first candidate for the phoneme lattice structure of “n” is “n”
, the second candidate is "mu", and the third candidate is "u". Regarding the phoneme lattice structure of "shi", the first candidate is "chi" and the second candidate is "shi". In the phoneme lattice structure of "ke", the first candidate is "ke", and there is no second or third candidate. Regarding the phoneme lattice structure of "i", the first candidate is "i", the second candidate is "hi",
The third candidate is "shi". For the “sa” phoneme lattice structure, the first candidate is “sa”, the second candidate is “ha”, and the third candidate is “a”
It is. The first candidate for the phoneme lattice structure of “n” is “n”
, the second candidate is "mu". In the phoneme lattice structure of "ki", the first candidate is "ki", and there are no second or third candidates. Also shown in FIG. 3 together with the phoneme lattice structure is the voice power, which is part of the voice characteristics analyzed by the acoustic analysis unit 11 in FIG. From the above results, when the first candidates of each phoneme recognition result are connected, the recognition result is "Denchi Keisanki". In this case, since the input voice is "denshikeisanki", the third character "shi" is incorrectly recognized as "chi". The operator checks the voice recognition results displayed on the display section 26 in FIG. 2, and confirms that the voice recognition results need to be changed.
Notify using the correction key. Figure 2 Correction key control circuit 28
recognizes that there is an error in the speech recognition result and immediately requests input of the changed phoneme as the second speech.

FIG. 4 shows an embodiment of the speech recognition and correction device of the present invention. FIG. 1 shows a phoneme spotting unit 14 that constitutes a speech recognition and correction device that recognizes speech uttered by dividing it into words.
It is a figure explaining the process. FIG. 4 shows and explains the voice power and phoneme lattice, which is one of the first voice feature parameter strings, and the voice power and phoneme lattice, which is one of the second voice feature parameter strings.

[0012] The operator inputs the erroneously recognized phoneme itself, in this case ``chi'', as the second voice. This is to spot the changed part without specifying it with a cursor or the like. The input "chi" is converted into a feature parameter through the speech input section 21 in FIG. 2 and the feature extraction circuit 22 in FIG. The second voice converted into a feature parameter string is subjected to DP matching in the DP matching circuit 29 of FIG. 2 with the feature parameters of the first voice stored in the feature parameter string storage circuit 27 of FIG. By performing DP matching between the feature parameter string of the first input voice and the feature parameter string of the second input voice, it is found that the second voice is the "chi" part of the first voice, and the changed part is Spotted as a candidate. This result is called spotting 1. As shown here, it is not that difficult to spot the phoneme "chi" once its features have been extracted and voice recognized. Figure 2D shows the second input audio.
In the P matching circuit 23, the phoneme memory dictionary 24 shown in FIG.
and DP matching and voice recognition. Here, the second voice is recognized as "chi". Speech recognition result of second voice “
The DP matching circuit 33 in FIG. 2 performs DP matching on the phoneme lattice storage circuit 32 in FIG. This DP
As a result of the matching, the ``chi'' in ``Denchi Keisanki'' is spotted. This is called spotting 2. In the spotting result comparison circuit 34 shown in FIG. 2, spotting 1 and spotting 2 are compared as candidates for the changed portion. Here, since spotting 1 is "chi" and spotting 2 is also "chi", the "chi" of "Denchi Keisanki" is spotted as the changed part. In this way, the misrecognized part "chi" is detected as a phoneme that needs to be changed. In response to this result, the phoneme lattice exchange circuit 30 in FIG. 2 uses the phoneme lattice storage circuit 32 in FIG.
"chi" was spotted as a changed part by operating
The second candidate "shi" in the speech recognition result is replaced with the first candidate and displayed on the display section 26 in FIG. 2 under the control of the display section control circuit 25 in FIG. Fortunately, the correct phoneme "shi" exists as the second candidate, so the operator uses the confirmation key to confirm and correct the speech recognition result. By the above operations, the correction of the voice recognition result is completed and the "Electronic computer" can be obtained.

[0013] The processing process of (Embodiment 1), which is an embodiment of the present invention, will be explained in detail with reference to FIGS. 8, 2, 3, and 4. FIG. 8 is a flowchart showing a processing example of a speech recognition and correction device that recognizes speech uttered by dividing it into words, which is an embodiment of the present invention.

First, a first voice is input by the operator. The voice input section 2 in Figure 2 is involved in inputting voice data.
It is 1. The input voice is immediately processed by the feature extraction circuit 2 in Figure 2.
In step 2, analysis and feature extraction are performed. The extracted features are stored in the feature parameter string storage circuit 27 in FIG. Further, the first speech feature parameter string is subjected to DP matching with the phoneme dictionary 24 in FIG. 2 in the DP matching circuit 23 in FIG. 2, and speech recognition is performed as a code string. This result is stored in the phoneme lattice storage circuit 32 in FIG. The result of voice recognition of the first voice is stored in the phoneme lattice storage circuit 32 in FIG. 2, and is also displayed on the display section 26 in FIG. 2 so that the operator can confirm it. The display control circuit 25 shown in FIG. 2 and the display section 26 shown in FIG. 2 are involved in this display. An example of the displayed speech recognition results is as shown in FIG. If there is an error or a need to change the voice recognition result displayed on the display unit 26 in FIG. 2, the operator uses the correction key to notify the user of the need for change. The correction key used here is controlled by the correction key control circuit 28 shown in FIG. If it is necessary to make changes,
Changes need to be detected immediately. In order to spot the changed part from among the voice recognition results, the changed part itself is inputted as a second voice by voice. The changed part input as the second voice is immediately extracted as a feature and becomes a feature parameter string.
At step 9, DP matching is performed with the feature parameter string of the first voice stored in the feature parameter string storage circuit 27 of FIG. Spotting 1 is spotted as a candidate for the changed part by DP matching. moreover,
In order to ensure spotting of the changed part, the result of voice recognition of the second voice and the phoneme lattice storage circuit 3 in FIG.
DP matching circuit 33 performs DP matching on the voice recognition result of the first voice stored in DP matching circuit 33 in FIG. As a result, spotting 2 is spotted as a candidate for the changed portion. The situation is shown in Figure 4. Then, in the spotting result comparison circuit 34 shown in FIG. 2, spotting 1 and spotting 2 are compared. If the results match, it is assumed that the changed portion has been accurately spotted, and the process immediately proceeds to change the changed portion. However, spotting 1, which spots the changed part using the feature parameter sequence that is the result of extracting the features of the first voice and the second voice, and the change part using the voice recognition results of the first voice and the second voice. If spotting 2, which was used to spot the area, is different, the spotting result comparison circuit 34 shown in FIG. Proceed with the process as a changed part. However, if there is an error in spotting the changed part and a phoneme other than the desired change is changed, the operator confirms the result displayed on the display and notifies the user by using the correction key. Then, the other spotting result, which was not changed because the DP value was small, becomes the changed part and the change processing is continued. DP matching circuit 29 in FIG. 2, DP matching circuit 33 in FIG. 2, spotting result comparison circuit 34 in FIG.
constitutes the spotting section 14 in FIG. Regarding the spotted change portion, the second candidate phoneme stored in the phoneme lattice storage circuit 32 in FIG. 2 is rearranged as the first candidate phoneme by the phoneme lattice replacement circuit 30 in FIG. 2 Display control circuit 2
5, it is displayed on the display section 26 in FIG. When the change unit is changed to the desired phoneme by this process, the voice recognition result is confirmed by operating the confirm key controlled by the confirm key control circuit 31 in FIG. Second part of the change
If the candidate phoneme is not the phoneme you wish to change, the confirmation key control circuit 31 in FIG. 2 recognizes that the confirmation key is not operated, and the phoneme lattice replacement circuit 30 in FIG. 2 selects a candidate with a lower priority. Replace the order as the first candidate and display it in the same way as before. When the operator finds the desired phoneme, he or she operates the confirmation key to confirm the speech recognition result. If the desired phoneme does not exist among the displayed second, third, and lower candidates, the voice is input again and voice recognition is performed using the same route as when the voice was first input. The circuits involved in this change operation are the phoneme lattice exchange circuit 30 in FIG. 2 and the confirmation key control circuit 31 in FIG.
Phoneme lattice storage circuit 32, FIG. 2 display unit control circuit 25,
This is the display section 26 in FIG. (Embodiment 2) FIG. 5 is a block diagram of a continuous speech recognition and correction device that recognizes speech uttered continuously without dividing each word, which is an embodiment of the present invention. The processing path indicated by the dashed line is that of the second voice.

The voice input as the first voice is shown in FIG.
The sound is sampled as an 8 KHz, 12 bits digital signal by the audio input section 41 in FIG. Furthermore, in the feature extraction circuit 42 in FIG. 5, which is also a component of the acoustic analysis unit 11 in FIG. is extracted and expressed as a string of characteristic parameters of the uttered word. The feature parameter string of the input voice extracted by the feature extraction circuit 42 in FIG. 5 is stored in the feature parameter string storage circuit 47 in FIG. The characteristic parameter string storage circuit 47 in FIG. 5 constitutes the acoustic analysis storage section 16 in FIG. 1st
The speech feature parameter string is subjected to continuous DP matching with the word storage dictionary 44 in FIG. 5 in the continuous DP matching circuit 43 in FIG. The word lattice storage circuit 52 shown in FIG. 5, which stores the speech recognition result of the first speech as a code string, constitutes the speech recognition storage section 13 shown in FIG. The speech recognition result of the first speech is shown in FIG. 5. Word lattice storage circuit 5
2, the voice is synthesized by the voice synthesis circuit 45 in FIG. 5, and outputted from the speaker under the control of the voice output control circuit 46 in FIG. The operator listens to the voice recognition result of the first voice output by voice, and if there is an error in the voice recognition result or there is a part that needs to be changed, the operator can touch the correction key etc. Notify the need to change the recognition result by action. The broken line in FIG. 5 shows the processing of the second input voice. Similar to the first voice, the word input as the second voice is converted into a feature parameter string through the voice input section 41 in FIG. 5 and the feature extraction circuit 42 in FIG. The second voice feature parameter string input as the part that needs to be changed is combined in the continuous DP matching circuit 49 in FIG. 5 with the first voice feature parameter string stored in the feature parameter string storage circuit 47 in FIG. Continuous DP matching. By inputting the incorrectly recognized part or the part to be changed as the second voice, it is compared with the feature parameter string of the first voice, and by searching for a matching part, the voice recognition result can be changed. It's spotting the parts. This result is set as spotting 1 as a candidate for the changed portion. Additionally, to ensure spotting of changes,
Similarly to the first voice, the second voice is also connected to the word memory dictionary 44 in FIG.
Match and voice recognition. The result of voice recognition of this second voice and the voice recognition result of the first voice stored in the word lattice storage circuit 52 in FIG.
A matching circuit 53 performs continuous DP matching. This is a process of spotting changed parts using the voice recognition results. The changed portion obtained by this process is referred to as spotting 2. Next, the spotting result comparison circuit 54 in FIG. 5 uses the first voice feature parameter string and the second voice feature parameter string previously stored in the feature parameter string storage circuit 47 in FIG. As a result of spotting, spotting 1,
The results of spotting the changed portion using the voice recognition results of the first voice and the voice recognition results of the second voice stored in the phoneme lattice storage circuit 52 in FIG. 5 are compared with spotting 2. FIG. 5 Spotting result comparison circuit 54
As a result of the comparison, if spotting 1 and spotting 2 are the same, it is assumed that the changed portion has been definitely spotted, and the process moves on to the next step. If spotting 1 and spotting 2 do not match, the DP when spotting 1 and spotting 2 are detected
The one with the larger value is determined to be the changed part. If a changed part is accidentally spotted, the operator can hear the changed result played from the speaker and know. In such a case, the correction key controlled by the correction key control circuit 48 in FIG. 5 is immediately operated, and the spotting result comparing circuit 54 in FIG. The other spotting result that was not selected as a candidate for the changed part can be used as the changed part, and the changing operation for the changed part can be proceeded again. Figure 5 Continuous DP matching circuit 49, Figure 5
Continuous DP matching circuit 53, Figure 5 spotting 54
constitutes the spotting section 14 in FIG. Regarding the spotted changed portion, the second candidate word stored in the word lattice storage circuit 52 in FIG. 5 is rearranged as the first candidate word by the word lattice replacement circuit 50 in FIG. 5 Speech synthesis circuit 45
The synthesized sound is synthesized at the step 5 and outputted from the speaker under the control of the sound output control circuit 46 shown in FIG. If the change section is changed to the desired word by this process, then
A confirmation key controlled by the confirmation key control circuit 51 is operated to confirm the voice recognition result. If the second candidate word of the change part is not the word desired to be changed, the confirmation key control circuit 50 in FIG. 5 recognizes that the confirmation key is not operated,
The word lattice replacement circuit 50 in FIG. 5 replaces the candidate with a lower priority as the first candidate,
Output the audio as before. When the operator finds the desired word, he or she operates the confirmation key to confirm the speech recognition result. If the desired word does not exist among the displayed second, third, and lower candidates, the user inputs the voice again and performs voice recognition using the same route as when the voice was input for the first time. The word lattice exchange circuit 50 in FIG. 5 and the confirmation key control circuit 51 in FIG. 5 constitute the changing unit 15 in FIG.

The present invention will be further explained based on (Embodiment 2) of the present invention using FIGS. 6 and 7.

FIG. 6 shows an embodiment of the speech recognition and correction device of the present invention, which constitutes a recognition and correction device that recognizes speech generated without dividing each word. 1 is a diagram showing the audio power of audio as part of data in an acoustic analysis storage unit.

Assume that the operator inputs the sentence "Today's weather is sunny." as the first voice. When this input is received, the word lattice structure in the word lattice storage circuit 52 of FIG. 5 is shown in FIG. 6, as shown in FIG. 6, the word lattice structure of "Today" is the first candidate "Today", and there is no second candidate or third candidate. The word lattice structure of “no” is the first candidate “no”,
The second candidate is "mo". The word lattice structure of "weather" has the first candidate "weather" and the second candidate "angel."Also,"
The word lattice structure of ``wa'' is the first candidate ``wa'' and the second candidate ``wa''.
It's "A". Further, the word lattice structure for "hare" is the first candidate "needle", the second candidate "hare", and the third candidate "hashi". Similarly, the word lattice structure for "desu" is the first candidate "desu" and the second candidate "fat". The voice power, which is a type of extracted feature, of the first input voice "Today's weather is sunny." is as shown in FIG. In this case, "sunny" is mistakenly recognized as "needle". The speech recognition result is synthesized by the speech synthesis circuit 45 shown in FIG. 5, and outputted from the speaker under the control of the speech output control circuit 46 shown in FIG. The operator listens to the voice recognition result of ``Today's weather is needles.'', learns that the voice recognition result needs to be changed, and uses the correction key to notify the operator. The correction key control circuit 48 in FIG. 5 recognizes that the voice recognition result needs to be changed, and immediately prepares a system to request input of the changed part as a second voice.

FIG. 7 shows an embodiment of the speech recognition and correction apparatus of the present invention. Based on (Embodiment 2), FIG.
FIG. 3 is a diagram illustrating processing of the phoneme spotting unit 14. FIG. FIG. 7 shows and explains the speech power and word lattice, which is one of the first speech feature parameter strings, and the speech power and speech recognition result, which is one of the second speech feature parameter strings. .

[0020] The operator inputs the erroneously recognized word itself, ``needle'', as the second voice. this is,
This is for spotting changed parts. The input second voice "needle" is converted into a feature parameter string through the voice input section 41 in FIG. 5 and the feature extraction circuit 42 in FIG. Continuous DP matching is performed with the speech feature parameter string in the continuous DP matching circuit 49 of FIG. The result is
As shown in FIG. 7, the "needle" was spotted as a candidate for the changed part. Spotting the “needle” 1
shall be. As shown here, it is not that difficult to spot the word "needle" once features have been extracted and voice recognized. The second input voice is subjected to continuous DP matching with the phoneme memory dictionary 44 in FIG. 5 in the continuous DP matching circuit 43 in FIG. 5 for speech recognition. Assume here that the second voice is recognized as "weather." The speech recognition result of the second speech "weather" is subjected to continuous DP matching with the word lattice storage circuit 52 of FIG. 5 in the continuous DP matching circuit 53 of FIG. As a result of this continuous DP matching, "weather" in "Today's weather is needles" is spotted. This is called spotting 2. In the spotting result comparison circuit 54 in FIG. 5, spotting 1 and spotting 2 are compared as candidates for the changed portion. here,
Since spotting 1 is "needle" and spotting 2 is "weather" and they do not match, the DP matching values when spotting 1 and spotting 2 are spotted by continuous DP matching are compared. Then, it can be seen that the DP value of spotting 1 is larger than the DP value of spotting 2. Here, spotting 1 becomes a candidate for the change part, and the "needle" in "Today's weather is needles." is spotted. In this way, the erroneously recognized part "needle" is detected as a word that needs to be changed. Based on this result, Figure 5
The word lattice replacement circuit 50 operates the word lattice storage circuit 52 shown in FIG.
is replaced with the first candidate, and the result is voice synthesized in the voice synthesis circuit 45 in FIG. 5, and the voice is output from the speaker under the control of the voice output control section 46 in FIG. Fortunately, the correct word ``Hare'' exists in the second candidate, and the operator hears the voice output ``Hare'' from the speaker, confirms it using the Confirm key, and corrects the voice recognition result. By the above operations, the correction of the speech recognition result is completed and "Today's weather is sunny." can be obtained. Suppose that in the spotting result comparison circuit 54 in FIG.
(needle) and Spotting 2 (weather), if Spotting 2 has a larger DP value, then Spotting 28 (Weather) is selected as the changed part. The change processing after the changed part spotting is performed for "weather". Referring to this case, an example of a case where spotting of a changed part is incorrect will be explained. The word lattice replacement circuit 50 in FIG. 5 rearranges the ranking of "angel" as the second candidate for "weather" stored in the word lattice storage circuit 52 in FIG. 5, and outputs it as the first candidate to the speech synthesis circuit 45 in FIG. However, the voice is synthesized in the voice synthesis circuit 45 in FIG. 5, and when the user hears "angel" being played from the speaker under the control of the voice output control circuit 46 in FIG. 5, he knows that spotting of the changed part has failed. Therefore, the correction key controlled by the correction key control circuit 48 in FIG. 5 is used to notify that there is an error in spotting the changed portion. Accordingly, in the comparison of the DP values in the spotting result comparison circuit 54 in FIG. 5, "needle", which could not be a candidate for the changed part due to its small DP value, is sent to the word lattice replacement circuit 50 in FIG. 5 as the changed part. In the word lattice exchange circuit 50 in FIG. 5, the word lattice storage circuit 52 in FIG.
The second candidate for "Needle" stored in "Hare" is the first.
The ranking will be changed as a candidate. Figure 5 Speech synthesis circuit 4
5, the speech output of "Hare" is synthesized, and the speech output control circuit 4 in FIG.
6 outputs a voice saying "Sunny". The operator hears the voice "Sunny", recognizes that the desired change part has been changed to the desired recognition result, and uses the confirmation key controlled by the confirmation key control circuit 51 in FIG. 5 to confirm the result. Correct. Through the above operations, even if the changed part is spotted incorrectly, the desired speech recognition result "Today's weather is sunny." can be obtained.

In (Example 1) and (Example 2), the audio input section consisted of a microphone, a high-frequency emphasis filter, and an AD converter, and was sampled as an 8 KHz, 12-bit digital signal. , other configurations may be used as long as the input audio can be sampled quickly. In addition, the feature extraction circuit converts the frequency of the audio signal converted into a digital signal for each frame, with a 16 ms interval as one frame, extracts feature parameters in the frequency domain, and generates a string of feature parameters of the uttered word. Although this method is used, other methods may be used as long as they can accurately extract the features. Furthermore, as a means for notifying the operator of the voice recognition results, in (Embodiment 1) a method of displaying the voice recognition results on the display unit was used. In addition, in (Example 2), a method was used in which the voice recognition result was generated by voice synthesis and output as a synthesized voice to notify the operator, but other methods may also be used.
Any method may be used as long as it can quickly notify the operator of the voice recognition results. In addition, in order to confirm the word or phoneme that you wish to change, we used the confirm key here, but you can use any other method, such as confirming by voice, as long as it allows you to confirm the word or phoneme accurately. It doesn't matter.

[0022]

Effects of the Invention As described above, the speech recognition correction device of the present invention eliminates the need to move the cursor to specify the changed part when changing the input speech recognition result, and allows the changed part to be changed by voice. By re-entering the information, you can specify and change the changed part very quickly. Therefore, even when voice recognition errors may occur frequently due to deterioration of the usage environment of the voice recognition device due to noise, etc., or due to the physical condition of the operator inputting input to the voice recognition device, etc., voice recognition correction can be performed. Corrections can be made using operations similar to voice input operations, without requiring any special operations or knowledge.
The burden on the operator has been reduced and work efficiency has been significantly improved.

[Brief explanation of the drawing]

FIG. 1 is a principle block diagram of a speech recognition and correction device according to the present invention.

FIG. 2 is a block diagram of one embodiment of the invention.

FIG. 3 shows the voice power, which is a type of feature of the first voice stored in the acoustic analysis storage unit, and the phoneme lattice of the first voice stored in the voice recognition storage unit, according to an embodiment of the present invention; FIG. 3 is a diagram showing structure 3;

FIG. 4 is a diagram illustrating a changing unit spotting unit and a changing unit according to an embodiment of the present invention.

FIG. 5 is a block diagram of one embodiment of the present invention.

FIG. 6 shows the voice power, which is a type of feature of the first voice stored in the acoustic analysis storage unit, and the word lattice of the first voice stored in the voice recognition storage unit according to an embodiment of the present invention; It is a figure showing a structure.

FIG. 7 is a diagram illustrating a changing unit spotting unit and a changing unit according to an embodiment of the present invention.

FIG. 8 is a diagram illustrating processing according to an embodiment of the present invention.

FIG. 9 is a block diagram of a conventional speech recognition correction device.

[Explanation of symbols]

1. Acoustic analysis department 2 Speech recognition section 3. Storage section 4 Cursor instruction section 5　Change section 11 Acoustic analysis department 12 Speech recognition section 13 Voice recognition memory unit 14 Audio spotting section 15 Change section 16 Acoustic analysis storage unit 21 Audio input section 22 Feature extraction circuit 23 DP matching circuit 24 Phoneme memory dictionary 25 Display control circuit 26 Display section 27 Feature parameter string storage circuit 28 Correction key control circuit 29 DP matching circuit 30 Phoneme lattice replacement circuit 31 Confirmation key control circuit 32 Phoneme lattice control circuit 33 DP matching circuit 34 Spotting result comparison circuit 41 Audio input section 42 Feature extraction circuit 43 Continuous DP matching circuit 44 Word memory dictionary 45 Speech synthesis circuit 46 Audio output control circuit 47 Feature parameter string storage circuit 48 Correction key control circuit 49 Continuous DP matching circuit 50 Word lattice replacement circuit 51 Confirmation key control circuit 52 Word lattice memory circuit 53 Continuous DP matching circuit 54 Spotting result comparison circuit

Claims (1)

[Claims] 1. An acoustic analysis unit that outputs characteristics of an input first voice, a speech recognition unit that converts the output of the acoustic analysis unit into a code string, and an acoustic analysis unit that stores the output of the acoustic analysis unit. a storage unit, a voice recognition storage unit that stores the output of the voice recognition unit; and a voice recognition storage unit that stores the output of the voice recognition unit, and compares the input second voice with the data in the acoustic analysis storage unit and the voice recognition storage unit,
A speech recognition and correction device comprising: a spotting section that detects a portion corresponding to the second speech from data of the speech; and a changing section that changes a portion of the code string that corresponds to the corresponding section.