CN113593554A - Voice recognition offline command word awakening application method and system - Google Patents
Voice recognition offline command word awakening application method and system Download PDFInfo
- Publication number
- CN113593554A CN113593554A CN202110827599.6A CN202110827599A CN113593554A CN 113593554 A CN113593554 A CN 113593554A CN 202110827599 A CN202110827599 A CN 202110827599A CN 113593554 A CN113593554 A CN 113593554A
- Authority
- CN
- China
- Prior art keywords
- voice data
- command word
- matching
- offline
- voice
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 44
- 238000012545 processing Methods 0.000 claims description 28
- 230000002618 waking effect Effects 0.000 claims description 27
- 230000015654 memory Effects 0.000 claims description 20
- 230000003139 buffering effect Effects 0.000 claims 1
- 230000000694 effects Effects 0.000 abstract description 5
- 238000013528 artificial neural network Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 238000012163 sequencing technique Methods 0.000 description 4
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- WHXSMMKQMYFTQS-UHFFFAOYSA-N Lithium Chemical compound [Li] WHXSMMKQMYFTQS-UHFFFAOYSA-N 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 229910052744 lithium Inorganic materials 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/22—Procedures used during a speech recognition process, e.g. man-machine dialogue
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/08—Speech classification or search
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/22—Procedures used during a speech recognition process, e.g. man-machine dialogue
- G10L2015/223—Execution procedure of a spoken command
Landscapes
- Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- Memory System Of A Hierarchy Structure (AREA)
Abstract
The application discloses a voice recognition offline command word awakening application method and system, which comprises a preset offline command word set, wherein the offline command word set comprises at least one offline command word; collecting voice data and caching; and matching the cached voice data with the offline command words in the offline command word set, and awakening the voice system if the matching is successful. The quantity of the off-line command words in the off-line command word set can be flexibly set according to requirements, so that different voice systems can be conveniently awakened, the awakening effect is increased, a plurality of voice systems are prevented from being awakened simultaneously as much as possible, and the awakening of the voice systems is enabled to be more indicative.
Description
Technical Field
The invention relates to the field of voice awakening, in particular to a voice recognition offline command word awakening application method.
Background
At present, in daily life, the application field of voice awakening is relatively wide, for example, robots, collection devices, wearable devices, smart homes, existing vehicles and the like, many devices with voice recognition functions are provided with voice loop technologies as a means of man-machine interaction, and different products have different command awakening words. When a user needs to wake up the device, a specific command wake-up word needs to be spoken, and the wake-up process can be regarded as a keyword retrieval task with small resources, wherein the small resources refer to smaller computing resources and smaller space storage resources.
In the related technology, the voice recognition offline command word awakening means that the voice recognition algorithm recognizes awakening command words to be recognized and processed by the algorithm by loading the offline command words; when continuous voice data is input into the awakening system, the voice recognition algorithm can continuously search the offline command words to be recognized from the voice data according to the offline command words contained in the offline command words.
With respect to the related art among the above, the inventors consider that the following drawbacks exist: with the development and application of voice awakening technology, awakening experience with higher recognition rate and awakening requirements for more recognized command words become practical requirements for technology development and application, off-line command words in the existing awakening system are all fixed off-line command words, so that the awakening system can recognize the off-line command words more fixedly, and when a plurality of voice systems exist simultaneously, the voice systems can be awakened, and the awakening of the voice systems is more troublesome.
Disclosure of Invention
In order to solve the problem that off-line command words in the existing awakening system are all fixed off-line command words, and when a plurality of voice systems exist simultaneously, the plurality of voice systems can be awakened, so that the awakening of the voice systems is troublesome, the application provides an awakening application method and system for voice recognition off-line command words.
The application method for awakening the voice recognition offline command word adopts the following technical scheme:
a voice recognition offline command word awakening application method is characterized by presetting an offline command word set, wherein the offline command word set comprises at least one offline command word;
collecting voice data and caching;
and matching the cached voice data with the offline command words in the offline command word set, and awakening the voice system if the matching is successful.
By adopting the technical scheme, the voice system is in a dormant state in an initial state, an offline command word set is preset, at least one offline command word is included in the command word set, voice data in the environment is collected, corresponding voice data is cached, the cached voice data is matched with the offline command words in the offline command word set, the voice system can be awakened after successful matching, the number of the offline command words in the offline command word set can be flexibly set according to requirements, so that different voice systems can be awakened conveniently, the awakening effect is increased, the phenomenon that a plurality of voice systems are awakened at the same time is avoided as much as possible, and the awakening of the voice system is more indicative.
Optionally, the preset offline command word set includes:
storing an offline command word set in a storage area of a memory;
the off-line command word set is preset for the corresponding voice system or added by the user.
By adopting the technical scheme, when the number of the voice systems of the same model in the same environment is single, the voice systems can be awakened by the offline awakening words preset in the system, when the number of the voice systems of the same model in the same environment is multiple, different voice systems of the same model can be awakened by different offline awakening words, and the voice systems provided with the same offline awakening words can also be awakened by the same offline awakening words at the same time. The method greatly increases the way of waking up the voice system, and makes the way of waking up the voice system more humanized.
Optionally, the acquiring and caching voice data includes:
collecting the voice data;
carrying out noise reduction processing on the collected voice data;
and caching the voice data after the noise reduction processing.
By adopting the technical scheme, the voice data is collected, the noise reduction processing is carried out on the collected voice data, and the voice data after the noise reduction processing is cached, so that the background noise of the cached voice data is reduced, and the recognition degree is better.
Optionally, the matching of the cached voice data with the plurality of offline command words is performed, and if the matching is successful, the voice system is awakened, including;
matching the off-line command words in the off-line command word set with the voice data item by item, and awakening the voice system if the matching is successful;
or the like, or, alternatively,
and simultaneously matching the off-line command words in the off-line command word set with the voice data, and awakening the voice system if the matching is successful.
By adopting the technical scheme, the cached voice data is matched with the off-line command words, the matching modes comprise two modes, namely off-line command words and voice data in the off-line command word set are matched item by item, if the matching is successful, the voice system is awakened, or the off-line command words and the voice data in the off-line command word set are simultaneously matched, if the matching is successful, the voice system is awakened, when the memory for awakening the system to run is less, the mode of item by item matching can be adopted, the occupation of the memory is reduced, and when the memory for awakening the system to run is more, the mode of simultaneous matching can be adopted, and the matching speed is increased.
Optionally, the matching, item by item, the offline command words in the offline command word set with the voice data, and if matching is successful, waking up the voice system, including:
loading a first offline command word in the offline command word set;
presetting a matching threshold;
matching the first offline command word with the voice data to generate a matching value;
if the matching value of the first offline command word is larger than or equal to the matching threshold value, awakening a voice system according to the voice data;
and if the matching value is smaller than the matching threshold value, loading a second off-line command word in the off-line command word set, and if the matching values of all off-line command words are smaller than the matching threshold value, emptying the cached voice data and caching the next voice data.
By adopting the technical scheme, when the voice data are matched item by item, the first off-line command word is loaded, the matching threshold value is preset, and the first off-line command word is matched with the voice data to generate the matching value;
if the matching value of the first off-line command word is larger than or equal to the matching threshold value, the voice system is awakened according to the voice data, so that the voice system is awakened;
and if the matching value is smaller than the matching threshold, loading a second off-line command word in the off-line command word set, and if the matching values of all off-line command words are smaller than the matching threshold, emptying the cached voice data and caching the next voice data to match the next voice data so that the off-line command words can be matched with the corresponding voice data.
Optionally, after waking up the voice system according to the voice data, the method further includes:
judging whether an offline awakening command word for awakening the voice system is the first offline command word;
if yes, not updating the sequence of the off-line command words in the off-line command word set;
if not, the off-line awakening command word is set to be in a first order, the order of the first off-line command word is adjusted to be in a second order, and other off-line command words are moved backwards in the orders.
By adopting the technical scheme, after the voice system is annularly processed, whether the offline wake-up command word for waking up the voice system is the first offline command word is judged;
if yes, not updating the sequence of the offline command words in the offline command word set;
if not, the offline wakeup command word is set to be in the first order, the order of the first offline command word is adjusted to be in the second order, and other offline command words are moved backwards in order, so that the judgment time in the next wakeup process is reduced, and the identification efficiency and the wakeup efficiency are improved.
Optionally, the simultaneously matching the offline command words in the offline command word set with the voice data, and if the matching is successful, waking up the voice system includes:
simultaneously loading all the offline command words in the offline command word set;
simultaneously matching all the offline command words with the voice data to generate a plurality of different matching values;
comparing the different matching values with corresponding matching thresholds respectively;
if the voice data are less than the corresponding matching threshold value, the matching is unsuccessful, the cached voice data are emptied, and the next voice data are cached;
and if one matching value is larger than or equal to the corresponding matching threshold value, awakening the voice system according to the voice data.
By adopting the technical scheme, the off-line command words in the off-line command word set are matched with the voice data at the same time, and if the matching is successful, the voice system is awakened, and the method comprises the following steps: simultaneously loading all the offline command words in the offline command word set; simultaneously matching all the offline command words with the voice data to generate a plurality of different matching values; comparing the different matching values with corresponding matching thresholds respectively; if the voice data are less than the corresponding matching threshold value, the matching is unsuccessful, the cached voice data are emptied, and the next voice data are cached; if one matching value is larger than or equal to the corresponding matching threshold value, the voice system is awakened according to the voice data, so that the awakening process of the voice system is faster, and the awakening efficiency of the voice system is improved.
A speech recognition offline command word wakeup application system, comprising: the device comprises a power supply module, a voice data temporary storage module, an offline command word storage module and a processing module;
the processing module is used for comparing the voice data with the offline command words;
the voice data caching module is used for caching the voice data;
the offline command word storage module is used for storing an offline command word set, and the offline command word set comprises at least one offline command word;
and the power supply module is used for supplying power to the processing module, the voice data temporary storage module and the offline command word storage module.
By adopting the technical scheme, the power supply module supplies power to the processing module, the voice data temporary storage module and the offline command word storage module, the voice data caching module collects and reduces noise of voice data and caches the voice data with the noise reduction completed, the offline command word storage module stores an offline command word set and can add, delete and sort the offline command word sets in the offline command word set, and the processing module calls the voice data in the voice data temporary storage module and the offline command word set in the offline command word storage module to compare the voice data with the offline command words, so that whether the voice system can be awakened or not is determined.
In summary, the present application includes at least one of the following beneficial technical effects:
1. when the voice system is in a dormant state in an initial state, an offline command word set is preset, at least one offline command word is included in the command word set, voice data in the environment are collected, corresponding voice data are cached, the cached voice data are matched with the offline command words in the offline command word set, the voice system can be awakened after the matching is successful, the number of the offline command words in the offline command word set can be flexibly set according to requirements, so that different voice systems can be awakened conveniently, and the awakening effect is improved;
2. the cached voice data is matched with a plurality of off-line command words, the matching modes include two modes, the off-line command words and the voice data in the off-line command word set are matched item by item, if the matching is successful, the voice system is awakened, or the off-line command words and the voice data in the off-line command word set are matched at the same time, if the matching is successful, the voice system is awakened, when the memory for awakening the operation of the system is less, the mode of matching item by item can be adopted, the occupation of the memory is reduced, and when the memory for awakening the operation of the system is more, the mode of matching at the same time can be adopted, and the matching speed is increased;
3. simultaneously loading all the offline command words in the offline command word set; simultaneously matching all the offline command words with the voice data to generate a plurality of different matching values; comparing the different matching values with corresponding matching thresholds respectively; if the voice data are less than the corresponding matching threshold value, the matching is unsuccessful, the cached voice data are emptied, and the next voice data are cached; if one matching value is larger than the corresponding matching threshold value, the voice system is awakened according to the voice data, so that the awakening process of the voice system is faster, and the awakening efficiency of the voice system is improved.
Drawings
FIG. 1 is a flowchart illustrating a method for waking up an application by using offline command words for speech recognition according to an embodiment of the present application;
FIG. 2 is a schematic flow chart illustrating the steps of collecting and caching voice data according to an embodiment of the present invention;
FIG. 3 is a flow chart illustrating how to wake up a voice system if matching is successful by matching offline command words and voice data in an offline command word set item by item according to an embodiment;
FIG. 4 is a flowchart illustrating an embodiment of a wake-up procedure for a voice system according to voice data;
fig. 5 is a flowchart illustrating an embodiment of a process of matching offline command words in an offline command word set with voice data and waking up a voice system if matching is successful.
Detailed Description
The present application is described in further detail below with reference to figures 1-5.
The embodiment of the application discloses a voice recognition offline command word awakening application method.
Referring to fig. 1, the method includes:
s1000, presetting an offline command word set, wherein the offline command word set comprises at least one offline command word.
The off-line command word set comprises at least one off-line command word, the number of the off-line command words is determined according to the number of the voice systems, public off-line command words and proprietary off-line command words can be set in the off-line command word set, for example, when a plurality of voice systems need to be awakened simultaneously in the same environment, the same public off-line command words can be set in the off-line command word sets, different voice systems can be awakened simultaneously through the public off-line command words, and the proprietary off-line command words can be freely selected to awaken the dedicated voice systems of the user.
S2000, voice data are collected and cached.
The collected voice data is voice data in a natural environment, the collected voice data comprises different languages, different dialects and other voice data in the same language, various different voice data can be collected and cached, the time for caching the voice data can be set freely relatively, the voice data used for awakening is usually less in vocabulary, so that the voice duration is shorter, for example, the voice data can be set to be shorter 1 second and more moderate two seconds and can also be set to be longer 3 seconds, and the voice data is a specific statement which can be clearly spoken by a person in a short time.
S3000, matching the cached voice data with the offline command words in the offline command word set, and awakening the voice system if the matching is successful.
When the voice recognition system is in the initial state, the voice system is in the dormant state, the voice system in the dormant state does not receive external voice, the voice data is matched with the offline command words in the offline command word set, the voice data is recognized through a voice recognition algorithm and is matched with the offline command word set after the recognition is finished, and if the matching is successful, the voice system is awakened, so that the voice system can receive the external voice and perform voice recognition.
The implementation principle of the embodiment is as follows:
when the voice system is in a dormant state in an initial state, an offline command word set is preset, at least one offline command word is included in the command word set, voice data in the environment are collected, corresponding voice data are cached, the cached voice data are matched with the offline command words in the offline command word set, the voice system can be awakened after the matching is successful, the number of the offline command words included in the offline command word set can be flexibly set according to requirements, therefore, the awakening of different voice systems is facilitated, and the awakening effect is improved.
In step S1000 of the embodiment shown in fig. 1, in order to make the offline command words in the offline command word set reasonably settable.
And storing an offline command word set in a storage area of the memory, wherein the offline command word set comprises a plurality of offline command words, and the offline command word set is operated in the memory.
The off-line command word set is preset by a voice system or added by a user, the off-line command word set is the voice system, namely off-line command words are preset by a memory corresponding to the voice system, the off-line command words preset by different memories corresponding to different voice systems are the same off-line command words, the off-line command words preset in the memory can be deleted, and the preset off-line command words can be restored;
when the offline command words in the offline command word set are added by the user, personalized offline command words can be added by the user according to different memories or different voice systems, so that the personalized offline command words can be annularly arranged according to the individual voice systems, the influence on other voice systems is reduced, for example, the offline command words preset in the system can be 'hello my equipment', the personalized offline command words can be 'classmate open personal computers' or 'give to the power on' and the like, and the personalized offline command words are set according to personal interest and preferences.
In step S2000 of the embodiment shown in fig. 1, in order to collect the voice in the environment, voice data that can be matched with the offline command word in the offline command word set is generated;
referring to fig. 2, the voice data collection and caching method includes the following steps:
s2100, voice data are collected.
The collected voice data includes a language spoken by a user in the environment, the types of the utterance include different languages, and different types of accents are included in different languages, for example, different regional dialects are included in chinese, american english includes american slang, and the like.
S2200, noise reduction processing is carried out on the collected voice data.
And carrying out noise reduction processing on the collected voice data.
The method comprises the steps of performing background environment noise reduction and voice recognition on collected voice; the noise reduction of the collected voice can be carried out by adopting a deep neural network, and can also be carried out by adopting a traditional signal processing mode;
the deep neural network recognition model is trained with enough amount before the deep neural network is denoised, the data volume of voice data which is usually used for off-line awakening is small, the sufficient training is convenient to denoise the voice, the deep learning method is adopted to denoise different voices in different scenes, and the collected voice data can be subjected to voice enhancement, voice separation, echo cancellation and the like in the recognition process, so that different voice data can be recognized in different scenes.
And S2300, caching the voice data subjected to noise reduction processing.
The voice data are cached in the flash memory, the caching time is short, the voice data are cleared and new voice data are cached again when the caching time is exceeded, different voice data can be cached, the time length of the cached voice data is generally the time length corresponding to the collected voice data, in order to reduce the time for recognizing the voice data, the cached voice data are generally 1-3 seconds, and 2 seconds are better time.
The implementation principle of collecting and caching the voice data in the embodiment of the application is as follows: the method comprises the steps of collecting voice data, wherein a collection object comprises a user speaking language in an environment, the speaking type comprises different languages, different types of accents are also contained in different languages, background environment noise reduction and voice recognition are carried out on the collected voice data after collection is finished, a deep neural network is adopted to carry out noise reduction and recognition on the voice data, the voice data are cached after recognition is finished, the caching time is short, and the caching time corresponds to the time length of the voice data.
In step S3000 shown in fig. 1, matching the noise-reduced and cached voice data with the offline command word, waking up the voice system if the matching is successful, and not waking up the voice system if the matching is unsuccessful, and the voice system still maintains the initial dormant state.
Referring to fig. 3, the cached voice data is matched with a plurality of offline command words, and if the matching is successful, the voice system is awakened.
Matching the off-line command words in the off-line command word set with the voice data item by item, and awakening the voice system if the matching is successful; in the matching mode, the voice data is respectively matched with each item of voice data, so that the matching of the voice data is single-thread matching, and if the matching is successful, the voice system is awakened.
Matching the off-line command words in the off-line command word set with the voice data item by item, and awakening the voice system if the matching is successful, comprising the following steps:
s3101, loading a first offline command word in the offline command word set.
The method comprises the steps that a plurality of off-line command words in an off-line command word set are randomly sequenced, the first off-line command word matched with voice data is the first off-line command word, and the first off-line command word is loaded to be firstly matched with the voice data.
S3102, presetting a matching threshold;
the matching threshold is the same point between the offline command word and the voice data, and the more the same points between the features of the voice data and the features of the offline command word are, that is, the higher the similarity between the voice data and the offline command word is, for example, when the similarity between the voice data and the offline command word is set to reach or exceed 90%, the default voice data is the voice data for waking up the voice system, the voice data can wake up the voice system, the matching threshold is 90%, and the voice data is not the voice data for waking up the voice system if the matching threshold is lower than.
S3103, matching the first off-line command word with the voice data to generate a matching value.
The method comprises the steps of randomly sequencing a plurality of offline command words in an offline command word set, wherein the first offline command word matched with voice data is a first offline command word, sequencing the offline command words in the offline command word set according to a random sequence, loading the offline command words in sequence from the first offline command word to enable the offline command words in the offline command word set to be respectively matched with the voice data, wherein the process of matching the offline command words with the voice data is a comparison process of the offline command words and the voice data, the higher the comparison similarity is, the higher the matching value of the offline command words is, and when the matching value is larger than or equal to a matching threshold value, the voice system is awakened.
S3104, if the matching value of the first offline command word is greater than or equal to the matching threshold, the voice system is awakened according to the voice data, and the cached voice data is emptied.
When the matching value between the first offline command word and the voice data is smaller than the matching threshold value, the plurality of offline command words behind the first offline command word are sequentially matched with the voice data according to the sorted sequence, and if the matching value of one offline command word is larger than or equal to the matching threshold value, the voice system is awakened.
S3105, if the matching value is smaller than the matching threshold, loading a second offline command word in the offline command word set, and if the matching values of all the offline command words are smaller than the matching threshold, emptying the cached voice data and caching the next voice data.
If the matching values of all the offline command words and the cached offline command words are smaller than the matching threshold, the voice data is not offline data for awakening the voice system, namely the voice data is cleared, the next voice data is cached, and the next voice data is used for being respectively matched with the offline command words in the offline command word set again.
The implementation principle of the embodiment is as follows: the method comprises the steps of randomly sequencing a plurality of offline command words in an offline command word set, wherein the first offline command word matched with voice data is the first offline command word, sequencing the offline command words in the offline command word set according to a random sequence, and sequentially loading from the first offline command word to enable the offline command words in the offline command word set to be respectively matched with the voice data, wherein the matching value of one offline command word and the voice data exceeds a matching threshold value, so that the voice system can be awakened, the occupation of a memory is reduced in the process of matching with the voice data one by one, the composite awakening requirements of various scenes are met, system resources can be efficiently utilized, the voice awakening is beneficial to the normal operation of intelligent equipment with limited system resources as a man-machine interaction entrance, and the generation of unsmooth cards is reduced.
Referring to fig. 4, after the voice system is woken up according to the voice data, the method further includes the following steps:
s3111, determining whether the offline wake-up command word for waking up the voice system is the first offline command word.
After the voice system is awakened for the first time, whether an offline command word for awakening the voice system is a first offline command word is judged, so that the offline command words can be conveniently sequenced in the follow-up process.
S3112, if yes, the order of the offline command word in the offline command word set is not updated.
If the offline command word for waking up the voice system is the first offline command word, the order of the offline command words in the offline command word set is not updated, and the first offline command word is usually the offline command word used by the user, so that the matching process of the offline command words and the voice data is reduced, and the matching speed is increased.
S3113, if not, setting the offline wakeup command word to a first order, adjusting the order of the first offline command word to a second order, and moving other offline command words backwards in order.
When the off-line command word for waking up the off-line voice system is not the first off-line command word, the original first off-line command word is modulated to the second order, the off-line command word for waking up the voice system is modulated to the first order, other off-line command words are all shifted backwards in order, usually, the off-line command word in the front order is the conventional off-line command word of the user, so that the sequence of the off-line command words is optimized, and the waking process of the voice system is further accelerated.
The implementation principle of the embodiment is as follows: by optimizing the sequence of the offline command words in the offline command word set for waking up the voice system, the conventional offline command words are positioned in the first order and the preposed order, the matching times of the offline command words in the offline command word set and the voice data are reduced, and the waking performance is improved.
Or the like, or, alternatively,
referring to fig. 5, the method for matching the off-line command word in the off-line command word set with the voice data at the same time, and waking up the voice system if the matching is successful, includes the following steps:
s3210, loading all the offline command words in the offline command word set at the same time.
When the use of the memory is sufficient, all the off-line command words in the off-line command word set can be loaded simultaneously, and the loading speed of the off-line command words is increased.
S3220 matches all the offline command words with the voice data at the same time, and generates a plurality of different matching values.
All the off-line command words are matched with the voice data at the same time, a plurality of different matching values are generated at the same time, the matching speed is increased, and the matching time is shortened.
S3230, comparing each of the plurality of different matching values with a corresponding matching threshold.
The comparison process of the different matching values and the corresponding matching threshold values is respectively carried out, so that the comparison time is shortened, and the comparison efficiency is improved.
S3240, if the matching is less than the corresponding matching threshold value, the matching is unsuccessful, the cached voice data is emptied, and the next voice data is cached;
if the matching values are all smaller than the matching threshold value, the matching is unsuccessful, the cached voice data are not the voice data used for awakening the voice system, the voice data are cleared, and the next voice data are cached.
S3250, if a matching value is greater than or equal to the corresponding matching threshold, the voice system is awakened according to the voice data, and the voice data is cleared after the awakening is completed.
If the matching value of one voice data is larger than or equal to the matching threshold value, the voice system can be awakened, and the awakening efficiency of the voice system is higher.
The implementation principle of the embodiment is as follows: when the use of the memory is sufficient, all the off-line command words are matched with the voice data to generate a plurality of matching values, and the voice system can be awakened as long as one matching value is larger than or equal to the matching threshold value, so that the awakening efficiency is increased, and the awakening time of the voice is reduced.
Example two:
a speech recognition offline command word wakeup application system, comprising: the device comprises a power supply module, a voice data temporary storage module, an offline command word storage module and a processing module;
the processing module is used for comparing the voice data with the offline command words;
the processing module is a processing chip, a comparison program runs in the processing chip and compares the called voice data with the off-line command words, and an ADC data acquisition submodule is arranged in the processing chip and is used for respectively acquiring and calling the voice data and the off-line command words.
The voice data caching module is used for caching the voice data;
the voice data cache module comprises a microphone array, a flash memory and a Bluetooth chip, the flash caches voice data, a voice noise reduction and voice recognition algorithm is arranged in the microphone array, the microphone array collects the voice data and performs noise reduction processing on the voice data, when a plurality of command words need to be searched and recognized at the same time, the microphone array can select a time point as an initial time and starts to cache the voice data for 2 seconds in the flash, so that the characteristics of the voice data cached in the flash are more prominent, the voice data are convenient to match with offline command words, the microphone array is in a normally open state, the voice data can be collected for a long time, and the Bluetooth chip is used for transmitting the voice data, so that the voice data are convenient to call.
The offline command word storage module is used for storing an offline command word set, and the offline command word set comprises at least one offline command word;
the off-line command word storage module is a RAM random access memory, namely a system memory, and the off-line command words are a smaller database stored in the RAM in a set and can be used for calling the off-line command words in the off-line command word set or changing the sequence of different off-line command words.
The power supply module is used for supplying power to the processing module, the voice data temporary storage module and the offline command word storage module, the power supply module can supply power in a wired mode, a power grid can be connected to supply power, the power supply module can also supply power to a battery, and the battery supplies power to a lithium battery which can be charged and discharged.
The implementation principle of the voice recognition offline command word awakening application system in the embodiment of the application is as follows: the power supply module supplies power to the processing module, the voice data temporary storage module and the offline command word storage module, the voice data caching module collects and denoises voice data and caches the denoised voice data, the offline command word storage module stores an offline command word set and can add, delete and sort the offline command word set in the offline command word set, and the processing module calls the voice data in the voice data temporary storage module and the offline command word set in the offline command word storage module to compare the voice data with the offline command words so as to determine whether the voice system can be awakened or not.
The above embodiments are preferred embodiments of the present application, and the protection scope of the present application is not limited by the above embodiments, so: all equivalent changes made according to the structure, shape and principle of the present application shall be covered by the protection scope of the present application.
Claims (8)
1. A voice recognition offline command word awakening application method is characterized by comprising the following steps:
presetting an offline command word set, wherein the offline command word set comprises at least one offline command word;
collecting voice data and caching;
and matching the cached voice data with the offline command words in the offline command word set, and awakening the voice system if the matching is successful.
2. The method of claim 1, wherein the predetermined set of offline command words comprises:
storing an offline command word set in a storage area of a memory;
the off-line command word set is preset for the corresponding voice system or added by the user.
3. The method for waking up application by voice recognition offline command word according to claim 1 or 2, wherein the collecting and buffering voice data comprises:
collecting the voice data;
carrying out noise reduction processing on the collected voice data;
and caching the voice data after the noise reduction processing.
4. The method as claimed in claim 1, wherein the step of matching the cached voice data with a plurality of offline command words and waking up the voice system if matching is successful comprises:
matching the off-line command words in the off-line command word set with the voice data item by item, and awakening the voice system if the matching is successful;
or the like, or, alternatively,
and simultaneously matching the off-line command words in the off-line command word set with the voice data, and awakening the voice system if the matching is successful.
5. The method as claimed in claim 4, wherein the step of matching the offline command words in the offline command word set with the voice data item by item, and if matching is successful, waking up the voice system comprises:
loading a first offline command word in the offline command word set;
presetting a matching threshold;
matching the first offline command word with the voice data to generate a matching value;
if the matching value of the first offline command word is larger than or equal to the matching threshold value, awakening a voice system according to the voice data;
and if the matching value is smaller than the matching threshold value, loading a second off-line command word in the off-line command word set, and if the matching values of all off-line command words are smaller than the matching threshold value, emptying the cached voice data and caching the next voice data.
6. The method as claimed in claim 5, further comprising, after waking up the voice system according to the voice data:
judging whether an offline awakening command word for awakening the voice system is the first offline command word;
if yes, not updating the sequence of the off-line command words in the off-line command word set;
if not, the off-line awakening command word is set to be in a first order, the order of the first off-line command word is adjusted to be in a second order, and other off-line command words are moved backwards in the orders.
7. The method as claimed in claim 4, wherein the step of matching the off-line command words in the off-line command word set with the voice data at the same time and the step of waking up the voice system if the matching is successful comprises:
simultaneously loading all the offline command words in the offline command word set;
simultaneously matching all the offline command words with the voice data to generate a plurality of different matching values;
comparing the different matching values with corresponding matching thresholds respectively;
if the voice data are less than the corresponding matching threshold value, the matching is unsuccessful, the cached voice data are emptied, and the next voice data are cached;
and if one matching value is larger than or equal to the corresponding matching threshold value, awakening the voice system according to the voice data.
8. An application system for waking up voice recognition offline command words, which adopts the method for waking up voice recognition offline command words as claimed in any one of claims 1 to 7, and comprises: the device comprises a power supply module, a voice data temporary storage module, an offline command word storage module and a processing module;
the processing module is used for comparing the voice data with the offline command words;
the voice data caching module is used for caching the voice data;
the offline command word storage module is used for storing an offline command word set, and the offline command word set comprises at least one offline command word;
and the power supply module is used for supplying power to the processing module, the voice data temporary storage module and the offline command word storage module.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110827599.6A CN113593554A (en) | 2021-07-21 | 2021-07-21 | Voice recognition offline command word awakening application method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110827599.6A CN113593554A (en) | 2021-07-21 | 2021-07-21 | Voice recognition offline command word awakening application method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113593554A true CN113593554A (en) | 2021-11-02 |
Family
ID=78248909
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110827599.6A Pending CN113593554A (en) | 2021-07-21 | 2021-07-21 | Voice recognition offline command word awakening application method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113593554A (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105575395A (en) * | 2014-10-14 | 2016-05-11 | 中兴通讯股份有限公司 | Voice wake-up method and apparatus, terminal, and processing method thereof |
| CN105791931A (en) * | 2016-02-26 | 2016-07-20 | 深圳Tcl数字技术有限公司 | Smart television and voice control method of the smart television |
| CN107591156A (en) * | 2017-10-10 | 2018-01-16 | 杭州嘉楠耘智信息科技股份有限公司 | Audio recognition method and device |
| CN108172228A (en) * | 2018-01-25 | 2018-06-15 | 深圳阿凡达智控有限公司 | Voice command words replacing options, device, voice-operated device and computer storage media |
| CN108335695A (en) * | 2017-06-27 | 2018-07-27 | 腾讯科技（深圳）有限公司 | Sound control method, device, computer equipment and storage medium |
| CN110503962A (en) * | 2019-08-12 | 2019-11-26 | 惠州市音贝科技有限公司 | Speech recognition and setting method, device, computer equipment and storage medium |
| CN111429915A (en) * | 2020-03-31 | 2020-07-17 | 国家电网有限公司华东分部 | Scheduling system and scheduling method based on voice recognition |
-
2021
- 2021-07-21 CN CN202110827599.6A patent/CN113593554A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105575395A (en) * | 2014-10-14 | 2016-05-11 | 中兴通讯股份有限公司 | Voice wake-up method and apparatus, terminal, and processing method thereof |
| CN105791931A (en) * | 2016-02-26 | 2016-07-20 | 深圳Tcl数字技术有限公司 | Smart television and voice control method of the smart television |
| CN108335695A (en) * | 2017-06-27 | 2018-07-27 | 腾讯科技（深圳）有限公司 | Sound control method, device, computer equipment and storage medium |
| CN107591156A (en) * | 2017-10-10 | 2018-01-16 | 杭州嘉楠耘智信息科技股份有限公司 | Audio recognition method and device |
| CN108172228A (en) * | 2018-01-25 | 2018-06-15 | 深圳阿凡达智控有限公司 | Voice command words replacing options, device, voice-operated device and computer storage media |
| CN110503962A (en) * | 2019-08-12 | 2019-11-26 | 惠州市音贝科技有限公司 | Speech recognition and setting method, device, computer equipment and storage medium |
| CN111429915A (en) * | 2020-03-31 | 2020-07-17 | 国家电网有限公司华东分部 | Scheduling system and scheduling method based on voice recognition |
Non-Patent Citations (3)
| Title |
|---|
| 余飞: "大数据下模式匹配算法研究", 《九江学院学报（自然科学版）》, no. 4, 31 December 2018 (2018-12-31), pages 72 - 75 * |
| 衣治安 等: "入侵检测系统中模式匹配算法探讨", 《辽宁工程技术大学学报》, vol. 23, no. 6, 31 December 2004 (2004-12-31), pages 816 - 819 * |
| 赵丽: "面向入侵检测的高效模式匹配算法研究", 计算机与数字工程, vol. 45, no. 8, 31 December 2017 (2017-12-31), pages 1592 - 1596 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108182937B (en) | Keyword recognition method, device, equipment and storage medium | |
| US20230409102A1 (en) | Low-power keyword spotting system | |
| CN111880856B (en) | Voice wakeup method and device, electronic equipment and storage medium | |
| CN105632486B (en) | Voice awakening method and device of intelligent hardware | |
| CN110570873B (en) | Voiceprint wake-up method and device, computer equipment and storage medium | |
| CN107481718A (en) | Audio recognition method, device, storage medium and electronic equipment | |
| CN103151039A (en) | Speaker age identification method based on SVM (Support Vector Machine) | |
| CN112825248A (en) | Voice processing method, model training method, interface display method and equipment | |
| CN111508493B (en) | Voice wake-up method and device, electronic equipment and storage medium | |
| CN110223687B (en) | Instruction execution method and device, storage medium and electronic equipment | |
| CN111429901B (en) | IoT chip-oriented multi-stage voice intelligent awakening method and system | |
| CN112102850A (en) | Processing method, device and medium for emotion recognition and electronic equipment | |
| CN110364147B (en) | Awakening training word acquisition system and method | |
| CN108595406B (en) | User state reminding method and device, electronic equipment and storage medium | |
| KR20210153165A (en) | An artificial intelligence device that provides a voice recognition function, an operation method of the artificial intelligence device | |
| WO2002073600A1 (en) | Method and processor system for processing of an audio signal | |
| CN112825250A (en) | Voice wake-up method, apparatus, storage medium and program product | |
| CN113593554A (en) | Voice recognition offline command word awakening application method and system | |
| CN111048068B (en) | Voice wake-up method, device and system and electronic equipment | |
| CN112289311B (en) | Voice wakeup method and device, electronic equipment and storage medium | |
| CN115101063B (en) | Low-computation-power voice recognition method, device, equipment and medium | |
| CN115206297B (en) | Variable-length voice emotion recognition method based on space-time multiple fusion network | |
| CN116386633A (en) | Intelligent terminal equipment control method and system suitable for noise condition | |
| Wang et al. | A fusion model for robust voice activity detection | |
| CN114171009A (en) | Voice recognition method, device, equipment and storage medium for target equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |