Speech recognition method, vehicle-mounted equipment and vehicle-mounted system

Abstract

The invention belongs to the technical field of voice recognition, and particularly relates to a voice recognition method, vehicle-mounted equipment and a vehicle-mounted system. The first coordinate data and the corresponding first network data are recorded in a network quality library, so that prediction reference for the network data during secondary driving is realized; when the initial voice data is obtained, the coordinates of the vehicle which the user finishes speaking the voice are predicted according to the habits of the user and the state of the vehicle; and performing query according to the predicted final third coordinate and the initial second coordinate and a first coordinate stored in a network quality library between the two coordinates, obtaining a network quality score of the predicted road section, and deciding on-line analysis or off-line analysis according to the score. According to the method, an online voice analysis mode or an offline voice analysis mode is selected by means of prediction in advance, so that the problems of loss of audio transmission data, incorrect recognition results or no data return caused by poor network quality are avoided.

Description

Technical Field

[0001] The present invention belongs to the technical field of speech recognition, and in particular relates to a speech recognition method, a vehicle-mounted device, and a vehicle-mounted system. Technical Background

[0002] In-vehicle voice recognition is a very common technology in intelligent driving. It is mainly used to recognize the user's voice, detect the user's intention, and control the corresponding modules to respond to achieve intelligent driving.

[0003] In the existing technology, recognition often adopts two methods, one is online recognition, that is, cloud network recognition, and the other is recognition by the vehicle's internal offline recognition module. When the network status is good, higher accuracy and smarter phenomenon recognition is preferred. When the network status is not good, offline recognition is chosen. Therefore, many car computers are equipped with network status recognition function. However, the driving process of the vehicle is a dynamic process. When the user is speaking, the network quality may be poor, resulting in loss of audio transmission data, incorrect recognition results or no data returned. This problem will especially occur when entering a tunnel at high speed. If the user speaks during this period, the online voice judgment will be interrupted, which will cause the entire voice judgment to fail or there will be high latency. Summary of the invention

[0004] Based on the above problems, the present invention proposes a speech recognition method, a vehicle-mounted device, and a vehicle-mounted system. When a user is detected speaking, network judgment is immediately started to predict the network status from the beginning of the speech to the end of the speech, and to judge in advance whether to use offline or online speech judgment to prevent interruptions in the middle, or the problem of not having enough time to switch between offline and online states.

[0005] The purpose of the present invention is achieved through the following technical solutions:

[0006] In a first aspect, the present invention provides a vehicle speech recognition method, comprising:

[0007] Periodically collecting the first coordinates and the first network data as the first data, calculating the first data score according to the first data by using the network data scoring algorithm and saving the score to the network quality library;

[0008] Obtain the initial voice data and predict the second coordinate of the vehicle when the voice is completed based on the user's voice habits and the current vehicle status;

[0009] According to the second coordinate, the third coordinate of the vehicle at the time when the initial voice data is sent, and the first coordinate between the second coordinate and the third coordinate, query the network quality library and predict the network quality score;

[0010] If the network quality score is lower than the preset score threshold, the user voice data is analyzed using an offline voice analysis method; if the network quality score is higher than the preset score threshold, the user voice data is analyzed using an online voice analysis method.

[0011] By recording the first coordinate data and the corresponding first network data to the network quality library, a prediction reference for the network data can be realized during the second driving; and when the initial voice data is obtained, the coordinates of the vehicle where the user speaks are predicted according to the user's habits and the vehicle status; and according to the predicted final third coordinate and the starting second coordinate, and the first coordinate stored in the network quality library between the two coordinates, a query is made, and the network quality score of the predicted section is obtained, and the online analysis or offline analysis is determined according to the score. This method selects the online voice analysis method or the offline voice analysis method by means of advance prediction, avoiding the problem of audio transmission data loss, incorrect recognition results or no data return due to poor network quality.

[0012] In some implementations, if the third coordinate has not been saved in the network quality library, the current second network data is collected, and the third coordinate is set as the second data, and the second data score is calculated according to the second data by a network data scoring algorithm and saved in the network quality library;

[0013] If the predicted second coordinate of the vehicle has not yet been saved in the network quality library, the second network data and the third coordinate are set as the third data, and the third data score is calculated by the network data scoring algorithm and saved in the network quality library; and when the vehicle travels to the third coordinate, the third network data is re-collected and the third network data and the second coordinate are set as the fourth data, and the fourth data score of the fourth data is calculated by the network data scoring algorithm, and the fourth data score replaces the third data score.

[0014] If the analyzed road section has not yet appeared in the network quality library, it will be scored and stored based on the real-time coordinates and network quality data. If the predicted coordinates are not recorded in the network quality library, the current network data and the predicted coordinates will be used to perform network quality scoring. After driving to the predicted coordinates, the network data will be re-scored and the original predicted data will be replaced.

[0015] In some implementations, periodically collecting the first coordinates and setting the first network data as the first data, calculating the first data score according to the first data by a network data scoring algorithm and saving it to a network quality library; comprising:

[0016] Acquire signal strength, upload rate, download rate, and network delay rate data as first network data;

[0017] Inputting the first coordinate and the first network data into a network data scoring algorithm;

[0018] The network data scoring algorithm queries the scores of signal strength, upload rate, download rate, and network delay rate according to a preset data score comparison table, and calculates the first data score according to the scoring weight.

[0019] The main parameters are judged through signal strength, upload rate, download rate, and network delay rate data. A data score comparison table is preset. The score of the current data is queried according to the data score comparison table, and each data is set with a corresponding weight. The network data score is obtained through the weight and the scores of the queries.

[0020] In some implementations, initial speech data is acquired, and the second coordinate of the vehicle when the speech is completed is predicted based on the user's speech habits and the current vehicle state; including:

[0021] Acquire initial voice data through voice activity detection;

[0022] Based on the initial voice data and the user's voice habits, the complete voice data is calculated; and the voice duration is calculated based on the complete voice data;

[0023] Calculate the current vehicle driving state according to the current vehicle speed, vehicle acceleration state, and vehicle driving direction;

[0024] The second coordinate is calculated based on the predicted voice duration and the current driving status.

[0025] After obtaining the initial voice data through the voice activity detection method, the complete voice data and complete duration are predicted according to the user's voice habits, and the driving speed is obtained according to the current state of the vehicle. The second coordinate is estimated through the driving speed and time, as well as the current coordinate position.

[0026] In some implementations, initial voice data is acquired through voice activity detection; including

[0027] The voice activity detection starts collecting the first initial voice data when the voice data reaches the energy threshold through the energy threshold method;

[0028] According to the first initial voice data, after excluding a preset voice frequency band and performing voice processing, the second initial voice data is obtained, and the second initial voice data is set as the initial voice data.

[0029] Voice activity detection is mainly performed by judging the energy threshold. When the detection reaches the threshold, voice data is collected, and after excluding some preset voice segments, that is, useless voice segments, the second initial voice data, that is, the initial voice data, is obtained.

[0030] In some implementations, the user's voice habits include:

[0031] After the user presets the voice, the online voice analysis method and / or the offline voice analysis method analyzes the first voice habit of the user's voice;

[0032] According to the collected voice data, a second voice habit in the voice is analyzed by an online voice analysis method and / or an offline voice analysis method; and the first voice habit is adjusted according to the second voice habit;

[0033] The first voice habit and the second voice habit both include speech speed, intonation, average duration, and interval of speech.

[0034] The user's voice habit may be a first voice habit obtained by an online or offline voice analysis method by recording a preset voice before driving, and the first voice habit may be further adjusted during driving according to a second voice habit collected in real time.

[0035] In some implementations, querying a network quality library and predicting a network quality score according to the second coordinate, the third coordinate of the vehicle at the time when the initial voice data is sent, and the first coordinate between the second coordinate and the third coordinate includes:

[0036] According to the second coordinate, the network quality database is queried to obtain the second data score; according to the third coordinate, the network quality database is queried to obtain the third data score; according to the first coordinate between the second coordinate and the third coordinate, the fourth data score is queried;

[0037] The fourth data score, the second data score, and the third data score are averaged to obtain a predicted network quality score.

[0038] The quality score of the road section formed by the second coordinate at the time of speaking and the third coordinate after the end of speaking is predicted is queried, and the predicted network quality score of the road section is obtained by taking the average value.

[0039] In some implementations, deriving a predicted network quality score further comprises:

[0040] Subtracting the first mass score, the second mass score, and the third mass score in pairs, and comparing the subtracted result with a preset fluctuation threshold;

[0041] If the result after subtraction is greater than the preset fluctuation threshold, the two quality scores with larger fluctuations are subtracted from the other quality scores. If the result is still greater than the preset fluctuation threshold, the quality score is deleted;

[0042] If the deleted network quality scores include the second network quality score, re-obtaining the second coordinates;

[0043] If the deleted network quality scores include the third network quality score, the third coordinate is obtained again.

[0044] According to the difference between the data scores of the coordinate query, a fluctuation threshold can be set to filter out some locations where network fluctuations may exist to reduce the error of judgment.

[0045] In a second aspect, the present invention provides an in-vehicle device, comprising: a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus;

[0046] The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform the operation of the vehicle voice recognition method as described in any one of the first aspects.

[0047] In a third aspect, the present invention provides a vehicle-mounted system, comprising:

[0048] A data collection and scoring module, used for periodically collecting the first coordinates and setting the first network data as the first data, calculating the first data score according to the first data through a network data scoring algorithm and saving it to a network quality library;

[0049] The data analysis module is used to obtain the initial voice data and predict the second coordinate of the vehicle when the voice is completed based on the user's voice habits and the current vehicle status;

[0050] A network quality scoring module, for querying a network quality library and predicting a network quality score according to the second coordinate, the third coordinate of the vehicle at the time when the initial voice data is sent, and the first coordinate between the second coordinate and the third coordinate;

[0051] The voice judgment decision module is used to judge that if the network quality score is lower than the preset score threshold, an offline voice analysis method is used to analyze the user voice data; if the network quality score is higher than the preset score threshold, an online voice analysis method is used to analyze the user voice data.

[0052] The method of any one of the first aspects is executed through a data collection and scoring module, a data analysis module, a network quality scoring module, and a speech judgment and decision module.

[0053] The beneficial effects of a speech recognition method, vehicle-mounted device, and vehicle-mounted system of the present invention are:

[0054] By recording the first coordinate data and the corresponding first network data to the network quality library, a prediction reference for the network data can be realized during the second driving; and when the initial voice data is obtained, the coordinates of the vehicle where the user speaks are predicted according to the user's habits and the vehicle status; and according to the predicted final third coordinate and the starting second coordinate, and the first coordinate stored in the network quality library between the two coordinates, a query is made, and the network quality score of the predicted section is obtained, and the online analysis or offline analysis is determined according to the score. This method selects the online voice analysis method or the offline voice analysis method by means of advance prediction, avoiding the problem of audio transmission data loss, incorrect recognition results or no data return due to poor network quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] The accompanying drawings are only used to illustrate the embodiments and are not to be considered as limiting the present invention. In addition, the same reference symbols are used to represent the same components throughout the accompanying drawings. In the accompanying drawings:

[0056] Figure 1 A flow chart of a vehicle speech recognition method according to the present invention;

[0057] Figure 2 Another implementation flow chart of step 101 of a vehicle speech recognition method of the present invention;

[0058] Figure 3 Another implementation flow chart of step 102 of a vehicle speech recognition method of the present invention;

[0059] Figure 4 Another implementation flow chart of step 301 of a vehicle speech recognition method of the present invention;

[0060] Figure 5 A flow chart of user voice habit setting of a vehicle voice recognition method of the present invention;

[0061] Figure 6 Another implementation flow chart of step 103 of a vehicle speech recognition method of the present invention;

[0062] Figure 7 A framework diagram of a vehicle device of the present invention;

[0063] Figure 8 The present invention is a framework diagram of a vehicle-mounted system.

[0064] Exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be limited to the embodiments set forth herein.

[0065] Embodiment 1:

[0066] like Figure 1 As shown, this embodiment provides a vehicle voice recognition method, including:

[0067] Step 101: periodically collect first coordinates and first network data as first data, calculate a first data score based on the first data using a network data scoring algorithm, and save the score to a network quality library;

[0068] Specifically, the preset period of 5S sets the first coordinate of the current vehicle and the first network data as the first data. The first coordinate is the global longitude and latitude coordinate, or the local coordinate established according to the local map. The first network data is data collection through other network modules, or it can be obtained by network packet capture and analysis of internal network data. In some embodiments, a special algorithm can also be used for algorithm collection, and the collected network data can be scored by the network data scoring algorithm, and the score can be entered into the network quality library. The beneficial effect of this step is that the coordinates can correspond to a network data score, so that when driving again, the network score can be queried by the coordinates, and the coordinate network prediction has been performed. The network quality library can be saved in the vehicle body memory, or it can be uploaded to the server through the network for storage or backup.

[0069] Furthermore, in order to reduce the complexity of coordinate acquisition, when acquiring the first coordinate, the first coordinate and the surrounding part of the coordinates may be acquired by default, so as to reduce excessive coordinate values ​​in the coordinate library and the resulting excessive data volume.

[0070] Step 102: Acquire initial voice data, and predict the second coordinate of the vehicle when the voice is completed based on the user's voice habits and the current vehicle state;

[0071] Specifically, the initial voice data is obtained. The initial voice data can be a certain character, a certain field, or a certain word. It is not a complete voice at one end, but a field that can determine the intention of the next voice, such as "open", "close", "go", etc. According to the user's voice habits, the length of the next speech that can be said is inferred and the second coordinate is predicted. Furthermore, the spoken words can also be marked at the coordinates to more accurately predict the second-time voice. The beneficial effect of this step is that the moment when the voice is finished can be predicted by voice, and the next second coordinate can be directly inferred by the voice moment. The moving speed can be obtained according to the change of longitude and latitude to obtain the relative speed at longitude and latitude, and the second coordinate position is calculated based on the inferred voice length and the previous direction.

[0072] Step 103: querying a network quality database and predicting a network quality score according to the second coordinate, the third coordinate of the vehicle at the time when the initial voice data is sent, and the first coordinate between the second coordinate and the third coordinate;

[0073] Specifically, according to the second coordinate, the third coordinate of the initial speech, and the first coordinate involved between the second coordinate and the third coordinate, a path is formed through the path coordinates, and according to each point on the path, the network quality library is queried, and the network quality score of the road section is obtained according to the scores of all coordinates. The beneficial effect of this method is that the score of the next road section can be intuitively predicted through the score, and the speech analysis decision can be determined.

[0074] Step 104: If the network quality score is lower than the preset score threshold, an offline speech analysis method is used to analyze the user voice data; if the network quality score is higher than the preset score threshold, an online speech analysis method is used to analyze the user voice data.

[0075] Specifically, the score threshold is used to determine whether to make an offline or online decision, so as to achieve the choice of decision and realize online and offline method analysis. The online speech analysis method includes uploading to the cloud and analyzing through the network, such as the cloud, and the offline speech analysis method includes analysis by the vehicle's built-in speech analysis module.

[0076] It should be noted that step 101 and steps 102 to 104 can be performed simultaneously, and there is no strict order of precedence. There is no relative order of precedence between steps 102 to 104, and they can be executed as long as the conditions in the steps are met.

[0077] The beneficial effects of the method are: by recording the first coordinate data and the corresponding first network data to the network quality library during driving, so as to realize the prediction reference of network data during secondary driving; and when the initial voice data is obtained, the coordinates of the vehicle where the user speaks are predicted according to the user's habits and the vehicle status; and according to the predicted final third coordinate and the starting second coordinate, and the first coordinate stored in the network quality library between the two coordinates, the query is performed, and the network quality score of the predicted section is obtained, and the online analysis or offline analysis is determined according to the score. The method solves the problem of wrong selection of offline analysis or online analysis due to sudden road sections causing network fluctuations, and the problem of long-term voice unresponsiveness due to untimely selection, and avoids the problem of voice analysis errors caused by network fluctuations by means of advance prediction.

[0078] Embodiment 2:

[0079] like Figure 2 to Figure 6 As shown, based on Example 1, this example further optimizes and illustrates the method steps mentioned in Example 1;

[0080] In some embodiments, it also includes:

[0081] If the third coordinate has not been saved in the network quality library, the current second network data is collected, and the third coordinate is set as the second data, and the second data score is calculated according to the second data by the network data scoring algorithm and saved in the network quality library;

[0082] Specifically, if the third coordinate is not found in the network quality library, that is, the coordinate is not saved in the voice when speaking, the second network data is directly obtained, and the second network data can be consistent with the first network data, so that the second data score is consistent with the first data score standard, and then it can be saved to the network quality library. The beneficial effect of this step is to make up for the voice leaks in some sections that are missed in periodic collection and may trigger voice, so as to improve the entire network quality library.

[0083] If the predicted second coordinate of the vehicle has not yet been saved in the network quality library, the second network data and the third coordinate are set as the third data, and the third data score is calculated by the network data scoring algorithm and saved in the network quality library; and when the vehicle travels to the third coordinate, the third network data is re-collected and the third network data and the second coordinate are set as the fourth data, and the fourth data score of the fourth data is calculated by the network data scoring algorithm, and the fourth data score replaces the third data score.

[0084] Specifically, if the predicted second coordinate is not saved, the score is derived from the current network data to the second coordinate, and saved to the network quality library to further fill in the gaps in the network quality library, and further, when driving to the coordinate, the current network data is re-acquired to correct the inferred data. The beneficial effect of this step is to further fill in the gaps in the network quality library to further improve the data.

[0085] The beneficial effect of this method is that if the analyzed road section has not yet appeared in the network quality library, it will be scored and stored based on the real-time coordinates and network quality data. If the predicted coordinates are not recorded in the network quality library, the current network data and the predicted coordinates will be used to perform network quality scoring. After driving to the predicted coordinates, the network data will be re-scored and the original predicted data will be replaced.

[0086] In some embodiments, periodically collecting the first coordinates and setting the first network data as the first data, calculating the first data score according to the first data by a network data scoring algorithm and saving it to a network quality library; including:

[0087] Step 201: Acquire signal strength, upload rate, download rate, and network delay rate data as first network data;

[0088] Specifically, the network signal strength is obtained through TelephonyManager, the upload rate and download speed are counted through TrafficStatsr, and the vehicle location information is obtained through GPS; to form the first network data, and further the network delay duration can be obtained through PING. The beneficial effect of this step is that the network quality can be judged through the parameters of signal strength, upload rate, download rate, and network delay rate data.

[0089] Step 202: inputting the first coordinate and the first network data into a network data scoring algorithm;

[0090] Specifically, a first data score is obtained through a network data scoring algorithm to be saved in a network quality library.

[0091] Step 203: The network data scoring algorithm queries the scores of signal strength, upload rate, download rate, and network delay rate according to the preset data score comparison table, and calculates the first data score according to the scoring weight.

[0092] Specifically, the network data scoring algorithm uses a preset data score comparison table to obtain the scores of signal strength, upload rate, download rate, and network delay rate based on the acquired signal strength, upload rate, download rate, and network delay rate, and according to the weight of the score, different weights can be determined according to the coordinates, that is, some coordinates consider the signal strength more, and the signal strength score weight is greater, and the first data score is calculated based on the scores of all parameters.

[0093] The beneficial effect of this method is that the main parameters are judged through signal strength, upload rate, download rate, and network delay rate data, and a data score comparison table is preset. The score of the current data is queried according to the data score comparison table, and each data is set with a corresponding weight. The network data score is obtained through the weight and the scores of the queries.

[0094] In some embodiments, initial voice data is obtained, and the second coordinate of the vehicle when the voice is completed is predicted according to the user's voice habits and the current vehicle state; including:

[0095] Step 301: Acquire initial voice data through voice activity detection;

[0096] Specifically, through the VAD (Voice Activity Detection) algorithm, the algorithm mainly detects whether there is a human voice signal in the current sound signal. If there is a human voice signal, the initial voice data is obtained.

[0097] Step 302: Calculate the complete voice data based on the initial voice data and the user's voice habits; and calculate the voice duration based on the complete voice data;

[0098] Specifically, the complete voice data is calculated based on the initial voice data and the user's voice habits. For example, when entering a highway section, "open" can be roughly predicted as "open music" or "open "player"". Further, based on the commonly used settings and the user's commonly used voice habits, the complete voice can be predicted as "open the music player and play XX song". Through the prediction method, the voice duration of the complete voice under the commonly used habits is calculated. The beneficial effect of this method is that the complete voice can be obtained by predicting the voice and the duration of the voice can be calculated.

[0099] Step 303: Calculate the current vehicle driving state according to the current vehicle speed, vehicle acceleration state, and vehicle driving direction;

[0100] Specifically, according to the vehicle speed, vehicle acceleration state, and vehicle driving direction, the driving state, such as direction, acceleration, speed, etc., is calculated to facilitate the prediction of coordinates. The beneficial effect of this step is that the predicted coordinates of the vehicle are more accurate through the driving state of the vehicle. Furthermore, the next coordinates can also be determined in combination with the path of the map.

[0101] Step 304: Calculate the second coordinate based on the predicted speech duration and the current driving status.

[0102] Specifically, the speech duration, driving status, and second coordinates are predicted. The speech duration is used as the time, the driving status is used as the direction and speed, and the second coordinates of longitude and latitude are calculated, or the second coordinates in the local coordinate system after local mapping.

[0103] The beneficial effect of this method is that after obtaining the initial voice data through the voice activity detection method, the complete voice data and the complete duration are predicted according to the user's voice habits, and the driving speed is obtained according to the current state of the vehicle. The second coordinate is estimated through the driving speed and time, as well as the current coordinate position.

[0104] In some embodiments, initial voice data is obtained by voice activity detection, including:

[0105] Step 401: voice activity detection uses an energy threshold method to detect that when voice data reaches an energy threshold, start collecting first initial voice data;

[0106] Specifically, voice activity detection can be performed through the energy threshold method, that is, when it is detected that a person's voice reaches a threshold, the first initial data is collected. The first initial data can be the beginning part of the person's voice data, or it can be some specific words obtained through voice analysis after reaching the threshold, such as "open", "close", etc.

[0107] Step 402: According to the first initial voice data, after excluding a preset voice band and performing voice processing, second initial voice data is obtained, and the second initial voice data is set as the initial voice data.

[0108] Specifically, after obtaining the first initial voice data, some preset voice segments are excluded, such as some unambiguous tones, some voices for waking up the system, and other voice segments whose intentions cannot be determined, and the data excluding the preset voice segments are subjected to voice processing such as filtering, and finally the second initial voice data is obtained and set as the initial voice data. Furthermore, if the next voice cannot be determined after excluding the first initial voice data, the next voice in the cache continues to be read. It should be noted that the user's voice data is usually stored in the cache, and the data collection and scoring module not only collects voice for the method to judge, but also saves it in the system's audio cache for voice analysis.

[0109] The beneficial effect of this method is that voice activity detection is mainly performed by judging the energy threshold. When the detection reaches the threshold, voice data is collected, and after excluding some preset voice segments, that is, useless voice segments, the second initial voice data, that is, the initial voice data, is obtained.

[0110] In some implementations, the user's voice habits include:

[0111] Step 501: After the user presets the voice, an online voice analysis method and / or an offline voice analysis method is used to analyze the first voice habit of the user's voice;

[0112] Specifically, the user's voice habit can be a first voice habit obtained through an online voice analysis method and / or an offline voice analysis method after a preset voice; and the two voice habits are adjusted mutually to achieve the first voice habit, that is, the determination of the initial voice habit.

[0113] Step 502: Analyze the second voice habit in the voice according to the collected voice data by an online voice analysis method and / or an offline voice analysis method; and adjust the first voice habit according to the second voice habit;

[0114] Specifically, according to the voice data obtained during driving, the second voice habit is analyzed, and the first voice habit is adjusted according to the second voice habit. The beneficial effect of this embodiment is that the preset voice habit is corrected by obtaining the second voice habit, and then the real-time voice habit is obtained.

[0115] Step 503: The first voice habit and the second voice habit both include speech speed, intonation, average duration, and interval of speech.

[0116] It should be noted that there is no sequential relationship between step 501 to step 503, and the execution can be triggered after the step conditions are met.

[0117] Specifically, speech habits include speech speed, intonation, speech duration, interval, etc., and may further include dialect, language logic habits, etc. The beneficial effect of the method is that speech can be further predicted through speech habits to obtain complete speech.

[0118] The beneficial effect of this method is that the user's voice habit can be a first voice habit obtained by an online or offline voice analysis method by recording a preset voice before driving, and the first voice habit can be further adjusted during driving according to a second voice habit collected in real time.

[0119] In some embodiments, according to the second coordinate, the third coordinate of the vehicle at the time when the initial voice data is sent, and the first coordinate between the second coordinate and the third coordinate, querying the network quality library and predicting the network quality score; including:

[0120] Step 601: query the network quality database according to the second coordinate to obtain the second data score; query the network quality database according to the third coordinate to obtain the third data score; query the fourth data score according to the first coordinate between the second coordinate and the third coordinate;

[0121] Specifically, according to the second coordinate, the third coordinate, and the first coordinate involved between the second coordinate and the third coordinate, the score query of the network quality library is implemented with the coordinates as indexes.

[0122] Step 602: Calculate the average of the fourth data score, the second data score, and the third data score to obtain a predicted network quality score.

[0123] Specifically, the predicted network quality score is obtained by averaging the query scores. Furthermore, weights can be set according to the scores of each data, and the final predicted network quality score is calculated according to the weights. The beneficial effect of this method is that the predicted network quality score of the total road section is calculated by calculating the scores point by point through the path formed by the predicted coordinates and the initial coordinates, so as to judge the network quality of the road section according to the predicted network quality score.

[0124] The beneficial effect of the method is that the quality score query is performed based on the second coordinate at the time of speaking to predict the road section formed by the third coordinate after the end of speaking, and the predicted network quality score of the road section is obtained by taking the average value.

[0125] In some embodiments, deriving a predicted network quality score further comprises:

[0126] Subtracting the first mass score, the second mass score, and the third mass score in pairs, and comparing the subtracted result with a preset fluctuation threshold;

[0127] If the result after subtraction is greater than the preset fluctuation threshold, the two quality scores with larger fluctuations are subtracted from the other quality scores. If the result is still greater than the preset fluctuation threshold, the quality score is deleted;

[0128] If the deleted network quality scores include the second network quality score, re-obtaining the second coordinates;

[0129] If the deleted network quality scores include the third network quality score, the third coordinate is obtained again.

[0130] Specifically, the queried quality scores are subtracted from each other, and the difference is used to obtain the two network quality scores with larger networks, and the quality scores with larger fluctuations are removed to reduce the score of the final data caused by network fluctuations. Among the deleted network quality scores, if the more critical second coordinate is included, the second coordinate is re-obtained, and if the third coordinate is included, the coordinate prediction is re-performed according to the coordinate prediction method. The beneficial effect of this method is that according to the difference between the data scores of the coordinate query, a fluctuation threshold can be set, thereby screening some locations where network fluctuations may exist to reduce the judgment error.

[0131] Embodiment 3:

[0132] like Figure 7 As shown, based on Embodiment 1 and Embodiment 2, this embodiment proposes a vehicle-mounted device, including: a processor, a memory, a communication interface and a communication bus, and the processor, the memory and the communication interface communicate with each other through the communication bus;

[0133] Specifically, the communication test device may include: a processor (processor) 701, a communication interface (Communications Interface) 704, a memory (memory) 702, and a communication bus 705.

[0134] The processor 701, the communication interface 704, and the memory 702 communicate with each other via the communication bus 705. The communication interface 704 is used to communicate with other devices such as a client or other server network elements. The processor 701 is used to execute the program 703, which can specifically execute the relevant steps in the above-mentioned embodiment 1.

[0135] Specifically, the program 703 may include program code, which includes computer executable instructions to execute a speech recognition method as mentioned in Embodiment 1 and Embodiment 2.

[0136] The processor 701 may be a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiment of the present invention. The one or more processors included in the communication test equipment may be processors of the same type, such as one or more CPUs; or may be processors of different types, such as one or more CPUs and one or more ASICs.

[0137] The memory 702 is used to store the program 703. The memory 702 may include a high-speed RAM memory, and may also include a non-volatile memory, such as at least one disk memory.

[0138] The memory is used to store at least one executable instruction, and the executable instruction enables the processor to execute the operations of the vehicle voice recognition method proposed in Example 1 and Example 2.

[0139] Embodiment 4:

[0140] like Figure 8 As shown, a vehicle-mounted system 800 is also included, including:

[0141] The data collection and scoring module 801 is used to periodically collect the first coordinate and the first network data as the first data, calculate the first data score according to the first data through the network data scoring algorithm and save it to the network quality library; the data collection and scoring module 801 executes step 101, the data collection and scoring module 801 periodically collects the first coordinate and the first network data, calculates the first data score according to the first data, and uploads it to the network quality library.

[0142] The data analysis module 802 is used to obtain the initial voice data and predict the second coordinates of the vehicle when the voice is completed based on the user's voice habits and the current vehicle status; step 102 is executed by the data analysis module 802, so that after the data collection and scoring module 801 obtains the initial voice data, the data analysis module 802 predicts the second coordinates of the vehicle when the voice is completed based on the user's voice habits and the current vehicle status.

[0143] The network quality scoring module 803 is used to query the network quality library and predict the network quality score according to the second coordinate, the third coordinate of the vehicle at the time when the initial voice data is sent, and the first coordinate between the second coordinate and the third coordinate; the network quality scoring module 803 is used to execute step 103, and after the second coordinate is obtained by voice analysis by the data analysis module 802, the network quality scoring module 803 queries the network quality library and predicts the network quality score according to the second coordinate, the third coordinate of the vehicle at the time when the initial voice data is sent, and the first coordinate between the second coordinate and the third coordinate.

[0144] The voice judgment decision module 804 is used to judge if the network quality score is lower than the preset score threshold, then use the offline voice analysis method to analyze the user voice data; if the network quality score is higher than the preset score threshold, then use the online voice analysis method to analyze the user voice data. It is used to execute step 104, and judge the network quality score obtained by the network quality scoring module 803 through the voice judgment decision module 804, and judge according to the network quality score, if the network quality score is lower than the preset score threshold, then use the offline voice analysis method to analyze the user voice data; if the network quality score is higher than the preset score threshold, then use the online voice analysis method to analyze the user voice data.

[0145] The algorithms or displays provided herein are not inherently related to any particular computer, virtual system or other device. In addition, the embodiments of the present invention are not directed to any particular programming language.

[0146] In the description provided herein, a large number of specific details are described. However, it is understood that embodiments of the present invention can be practiced without these specific details. Similarly, in order to simplify the present invention and help understand one or more of the various inventive aspects, in the above description of exemplary embodiments of the present invention, the various features of the embodiments of the present invention are sometimes grouped together into a single embodiment, figure, or description thereof. Wherein, the claims that follow the specific embodiment are hereby expressly incorporated into the specific embodiment, wherein each claim itself is a separate embodiment of the present invention.

[0147] Those skilled in the art will appreciate that the modules in the devices in the embodiments may be adaptively changed and arranged in one or more devices different from the embodiments. The modules or units or components in the embodiments may be combined into one module or unit or component, and further may be divided into a plurality of submodules or subunits or subcomponents. Except that at least some of such features and / or processes or units are mutually exclusive.

[0148] It should be noted that the above embodiments illustrate the present invention rather than limit it, and that those skilled in the art may devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference symbol between brackets shall not be construed as a limitation on the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "one" or "an" preceding an element does not exclude the presence of a plurality of such elements. The present invention may be implemented by means of hardware comprising a number of different elements and by means of a suitably programmed computer. In a unit claim enumerating a number of devices, several of these devices may be embodied by the same hardware item. The use of the words first, second, and third, etc. does not indicate any order. These words may be interpreted as names. The steps in the above embodiments, unless otherwise specified, should not be understood as limitations on the order of execution.

Claims

1. A vehicle speech recognition method, It is characterized in that include: Periodically collecting the first coordinates and the first network data as the first data, calculating the first data score according to the first data by using the network data scoring algorithm and saving the score to the network quality library; Obtain the initial voice data and predict the second coordinate of the vehicle when the voice is completed based on the user's voice habits and the current vehicle status; According to the second coordinate, the third coordinate of the vehicle at the time when the initial voice data is sent, and the first coordinate between the second coordinate and the third coordinate, query the network quality library and predict the network quality score; If the network quality score is lower than the preset score threshold, an offline voice analysis method is used to analyze the user voice data; if the network quality score is higher than the preset score threshold, an online voice analysis method is used to analyze the user voice data.

2. The vehicle speech recognition method according to claim 1, It is characterized in that If the third coordinate has not been saved in the network quality library, the current second network data is collected, and the third coordinate is set as the second data, and the second data score is calculated according to the second data by the network data scoring algorithm and saved in the network quality library; If the second coordinate of the predicted vehicle has not been saved in the network quality library, the second network data and the third coordinate are set as the third data, and the third data score is calculated by the network data scoring algorithm and saved in the network quality library; and when the vehicle travels to the third coordinate, the third network data is re-collected and the third network data and the second coordinate are set as the fourth data, and the fourth data score is calculated by the network data scoring algorithm for the fourth data, and the third data score is replaced by the fourth data score.

3. The vehicle speech recognition method according to claim 2, It is characterized in that The periodic collection of the first coordinates and the first network data are set as the first data, and the first data score is calculated according to the first data by the network data scoring algorithm and saved to the network quality library; including: Acquire signal strength, upload rate, download rate, and network delay rate data as the first network data; Inputting the first coordinate and the first network data into the network data scoring algorithm; The network data scoring algorithm queries the scores of the signal strength, the upload rate, the download rate, and the network delay rate according to a preset data score comparison table, and calculates the first data score according to the scoring weight.

4. The vehicle speech recognition method according to claim 2, It is characterized in that Obtain the initial voice data and predict the second coordinate of the vehicle when the voice is completed based on the user's voice habits and the current vehicle status; including: Acquiring the initial voice data through voice activity detection; Calculate the complete voice data based on the initial voice data and the user's voice habits; and calculate the voice duration based on the complete voice data; Calculating the current vehicle driving state according to the current vehicle speed, vehicle acceleration state, and vehicle driving direction; The second coordinate is calculated according to the predicted speech duration and the current driving status.

5. The vehicle speech recognition method according to claim 4, It is characterized in that The initial voice data is obtained through voice activity detection and uploaded to the network data scoring algorithm; including: The voice activity detection starts collecting the first initial voice data when the voice data reaches the energy threshold through the energy threshold method; According to the first initial voice data, after excluding a preset voice band and performing voice processing, second initial voice data is obtained, and the second initial voice data is set as the initial voice data.

6. The vehicle speech recognition method according to claim 5, It is characterized in that The user voice habits include: After the user presets the voice, the online voice analysis method and / or the offline voice analysis method analyzes the first voice habit of the user's voice; According to the collected voice data, a second voice habit in the voice is analyzed by the online voice analysis method and / or the offline voice analysis method; and the first voice habit is adjusted according to the second voice habit; The first voice habit and the second voice habit both include speech speed, intonation, average duration, and interval of speech.

7. The vehicle speech recognition method according to claim 2, It is characterized in that According to the second coordinate, the third coordinate of the vehicle at the time when the initial voice data is sent, and the first coordinate between the second coordinate and the third coordinate, query the network quality library and predict the network quality score; including: querying the network quality database according to the second coordinate to obtain a second data score; querying the network quality database according to the third coordinate to obtain a third data score; querying a fourth data score according to the first coordinate between the second coordinate and the third coordinate; An average of the fourth data score, the second data score, and the third data score is calculated to obtain the predicted network quality score.

8. The vehicle speech recognition method according to claim 7, It is characterized in that The step of obtaining the predicted network quality score also includes: Subtracting the first mass score, the second mass score, and the third mass score in pairs, and comparing the subtraction result with a preset fluctuation threshold; If the result after the subtraction is greater than the preset fluctuation threshold, subtract the two quality scores with larger fluctuations from the other quality scores, and if it is still greater than the preset fluctuation threshold, delete the quality score; If the deleted network quality scores include the second network quality score, re-obtaining the second coordinate; If the deleted network quality score includes the third data score, the third coordinate is re-acquired.

9. A vehicle-mounted device, It is characterized in that It includes a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other through the communication bus; The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform the operation of the vehicle voice recognition method of the first aspect.

10. A vehicle-mounted system, It is characterized in that include: A data collection and scoring module, used for periodically collecting first coordinates and first network data as first data, calculating a first data score according to the first data through a network data scoring algorithm and saving the score to a network quality library; The data analysis module is used to obtain the initial voice data and predict the second coordinate of the vehicle when the voice is completed based on the user's voice habits and the current vehicle status; A network quality scoring module, configured to query a network quality library and predict a network quality score according to the second coordinate, the third coordinate of the vehicle at the time when the initial voice data is sent, and the first coordinate between the second coordinate and the third coordinate; The voice judgment decision module is used to judge that if the network quality score is lower than the preset score threshold, an offline voice analysis method is used to analyze the user voice data; if the network quality score is higher than the preset score threshold, an online voice analysis method is used to analyze the user voice data.

Images