Aircraft simulation training voice code transmission method, device, equipment and medium

Abstract

The invention provides an aircraft simulation training voice coding transmission method, device and equipment and a medium, and relates to the technical field of voice, and the method comprises the steps: collecting voice data of a pilot, and generating a target voice recognition text corresponding to the voice data through a voice recognition model; converting the target speech recognition text into first coded data based on a pre-constructed aircraft proprietary language corpus; and transmitting the first coded data to terminals associated with other pilots, so as to convert the first coded data into a target voice recognition text through the terminals associated with other pilots based on an aircraft proprietary language corpus, and broadcasting the target voice recognition text corresponding to the first coded data, and voice transmission between two or more pilots is realized. According to the invention, the accuracy of speech recognition is remarkably improved, the communication bandwidth consumed in the transmission process can be greatly reduced, and the transmission efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of speech technology, and in particular to a speech coding transmission method, device, equipment and medium for aircraft simulation training. Background Art

[0002] Traditional simulators train language interaction through transmission of separate voice streams, mainly through radio and satellite, which can transmit real voice. However, traditional simulators have the following problems: they receive large interference and severe noise, and do not process the voice (or accent) of a single person, so there is a certain degree of unreliability in the entire communication process; the volume of voice files is large, resulting in the transmission process consuming more communication bandwidth and slow transmission efficiency. Summary of the invention

[0003] In view of this, the purpose of the present invention is to provide an aircraft simulation training voice coding transmission method, device, equipment and medium, which significantly improves the accuracy of voice recognition, and can also greatly reduce the communication bandwidth consumed in the transmission process and improve the transmission efficiency.

[0004] In a first aspect, the present invention provides a method for speech coding and transmission in aircraft simulation training, the method being applied to a terminal associated with a pilot, and the terminals associated with different pilots being connected in communication, the method comprising: Collect the pilot's voice data and use the voice recognition model to generate the target voice recognition text corresponding to the voice data; wherein the voice recognition model is trained based on the pilot's personal voice library, and the voice data and its corresponding target voice recognition text are also used to optimize the voice recognition model; Based on a pre-built aircraft-specific corpus, converting the target speech recognition text into first coded data; The first coded data is transmitted to terminals associated with other pilots, so that the first coded data is converted into target speech recognition text based on an aircraft-specific terminology corpus through the terminals associated with other pilots, and the target speech recognition text corresponding to the first coded data is broadcast, so as to realize voice transmission between two or more pilots.

[0005] In one embodiment, the aircraft-specific terminology corpus includes a plurality of aircraft-specific terminology and corresponding encoding data; based on the pre-built aircraft-specific terminology corpus, converting the target speech recognition text into the first encoding data includes: Based on a pre-built aircraft-specific terminology corpus, the target speech recognition text is split to obtain a plurality of aircraft-specific terms contained in the target speech recognition text; For any aircraft-specific term, the following operations are performed: if the aircraft-specific term is included in the aircraft-specific term, based on the correspondence between the aircraft-specific term and the coded data, the coded data corresponding to the aircraft-specific term is determined; if the aircraft-specific term is not included in the aircraft-specific term, the aircraft-specific term is encoded to obtain the corresponding coded data; Based on the encoding data corresponding to each aircraft-specific term included in the target speech recognition text, first encoding data corresponding to the target speech recognition text is generated.

[0006] In one embodiment, the method further comprises: receiving second coded data transmitted by a terminal associated with another pilot; decoding the second coded data into a plurality of aircraft-specific terms based on a mapping relationship between the aircraft-specific terms and the coded data; A target speech recognition text corresponding to the second coding data is generated according to the multiple aircraft-specific terms obtained by decoding, and the target speech recognition text corresponding to the second coding data is broadcasted in standard pronunciation.

[0007] In one embodiment, the method further comprises: Counting the first usage frequency of each aircraft-specific term included in the aircraft-specific term corpus, encoding each aircraft-specific term according to the encoding strategy corresponding to the first usage frequency, and obtaining encoding data corresponding to each aircraft-specific term; Among them, when the first usage frequency is higher than the first preset threshold, the encoding strategy is to encode each aircraft-specific term with one byte; when the first usage frequency is lower than the first preset threshold, the encoding strategy is to encode each aircraft-specific term with two bytes.

[0008] In one embodiment, the method further comprises: The second usage frequency of the aircraft-specific terms not included in the aircraft-specific terminology corpus is counted. When the second usage frequency is higher than a second preset threshold, each aircraft-specific term is encoded according to the encoding strategy corresponding to the second usage frequency, and the obtained encoding data is added to the extension code of the aircraft-specific terminology corpus.

[0009] In one implementation, the target speech recognition text is determined from a plurality of speech recognition texts output by the speech recognition model, and the method further includes: A weight is marked for each speech recognition text output by the speech recognition model; wherein the weight of the target speech recognition text is higher than the weight of other speech recognition texts except the target speech recognition text; The speech data, each speech recognition text output by the speech recognition model for the speech data, and its weight are used as sample data; Assigning an initial sampling probability to the sample data, and saving the sample data and its sampling probability to a personal voice database; wherein the sampling probability of the sample data is dynamically adjustable, and the sampling probability of the sample data is related to the number of times the sample data is sampled; According to a preset sampling ratio and based on the sampling probability of each sample data, the sample data stored in the personal voice database are sampled to obtain a target sample data set, and the sampling probability of each sample data stored in the personal voice database is adjusted using the target sample data set; The speech recognition model is optimized using the target sample data set.

[0010] In one embodiment, generating a target speech recognition text corresponding to speech data using a speech recognition model includes: Input the speech data into the speech recognition model; Extracting time series features and local features corresponding to the speech data through the speech recognition model, performing feature screening on the time series features and the local features respectively, fusing the time series features after feature screening and the local features after feature screening, and outputting multiple speech recognition texts corresponding to the speech data based on the fused features; A target speech recognition text is determined from the speech recognition texts.

[0011] In one embodiment, transmitting the first coded data to a terminal associated with another pilot includes: The first coded data is encrypted at least once to transmit the encrypted first coded data to a terminal associated with another pilot.

[0012] In a second aspect, the present invention further provides an aircraft simulation training voice coding transmission device, the device is applied to a terminal associated with a pilot, and the terminals associated with different pilots are connected for communication, and the device includes: A speech recognition module is used to collect the pilot's speech data and generate a target speech recognition text corresponding to the speech data using a speech recognition model; wherein the speech recognition model is trained based on the pilot's personal speech library, and the speech data and its corresponding target speech recognition text are also used to optimize the speech recognition model; A text encoding module, for converting the target speech recognition text into first encoded data based on a pre-built aircraft-specific terminology corpus; The voice transmission module is used to transmit the first coded data to the terminal associated with other pilots, so as to convert the first coded data into a target voice recognition text based on the aircraft-specific terminology corpus through the terminal associated with other pilots, and broadcast the target voice recognition text corresponding to the first coded data, so as to realize voice transmission between two or more pilots.

[0013] In a third aspect, the present invention further provides an electronic device, comprising a processor and a memory, wherein the memory stores computer executable instructions that can be executed by the processor, and the processor executes the computer executable instructions to implement any one of the methods provided in the first aspect.

[0014] In a fourth aspect, the present invention further provides a computer-readable storage medium, which stores computer-executable instructions. When the computer-executable instructions are called and executed by a processor, the computer-executable instructions prompt the processor to implement any one of the methods provided in the first aspect.

[0015] The present invention provides a method, device, equipment and medium for voice coding transmission of aircraft simulation training. First, voice data of a pilot is collected, and a target voice recognition text corresponding to the voice data is generated by using a voice recognition model. The voice recognition model is obtained by training based on the pilot's personal voice library. The voice data and the corresponding target voice recognition text are also used to optimize the voice recognition model. Then, based on a pre-built aircraft-specific terminology corpus, the target voice recognition text is converted into first coding data. Finally, the first coding data is transmitted to terminals associated with other pilots, so that the first coding data is converted into the target voice recognition text based on the aircraft-specific terminology corpus through the terminals associated with other pilots, and the target voice recognition text corresponding to the first coding data is broadcast, so as to realize voice transmission between two or more pilots. The above method uses the pilot's personal voice library to train the speech recognition model for aircraft-specific terms, and obtains a speech recognition model dedicated to the pilot, thereby significantly improving the accuracy of speech recognition of aircraft-specific terms; in addition, the target speech recognition text is converted into first coded data through the aircraft-specific terminology corpus, and then transmitted to other pilot-associated terminals in the form of coded data, and the terminal broadcasts the target speech recognition text corresponding to the first coded data, which can not only greatly reduce the communication bandwidth consumed in the transmission process and improve the transmission efficiency, greatly broaden the transmission means, but also achieve the effects of reducing noise, shielding the pilot's personal accent, and standardizing voice broadcasting.

[0016] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description, claims and drawings.

[0017] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0019] Figure 1 A flow chart of a method for speech coding and transmission of aircraft simulation training provided by an embodiment of the present invention; Figure 2 A technical framework diagram of a speech coding and transmission method for aircraft simulation training provided by an embodiment of the present invention; Figure 3 A schematic diagram of a coding rule provided by an embodiment of the present invention; Figure 4 A practical application flow chart of a method for speech coding and transmission in aircraft simulation training provided by an embodiment of the present invention; Figure 5 A schematic diagram of the structure of a speech coding and transmission device for aircraft simulation training provided by an embodiment of the present invention; Figure 6 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described in combination with the embodiments below. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] At present, the existing technology has problems such as poor reliability of the communication process, large consumption of communication bandwidth in the transmission process, and slow transmission efficiency. Based on this, the implementation of the present invention provides an aircraft simulation training voice coding transmission method, device, equipment and medium, which significantly improves the accuracy of voice recognition, and can also greatly reduce the communication bandwidth consumed in the transmission process and improve the transmission efficiency.

[0022] To facilitate understanding of this embodiment, firstly, a method for speech coding and transmission of aircraft simulation training disclosed in an embodiment of the present invention is described in detail. The method is applied to terminals associated with pilots, and the terminals associated with different pilots are connected to each other in communication. Figure 1 The flowchart of a method for speech coding and transmission of aircraft simulation training is shown in FIG. 1 , and the method mainly includes the following steps S102 to S106: Step S102: collect the pilot's voice data and use the voice recognition model to generate a target voice recognition text corresponding to the voice data.

[0023] The speech recognition model is trained based on the pilot's personal speech library, and the speech data and its corresponding target speech recognition text are also used to optimize the speech recognition model. In one example, the pilot's speech data can be collected using an onboard headset. After receiving the speech data, the terminal inputs it into the speech recognition model. The speech recognition model can output one or more speech recognition texts, and the target speech recognition text is obtained after the pilot confirms or selects it.

[0024] Step S104 : based on a pre-built aircraft-specific terminology corpus, convert the target speech recognition text into first coded data.

[0025] The aircraft-specific terminology corpus includes multiple aircraft-specific terms and their corresponding encoding data, and aircraft-specific terms with different usage probabilities are obtained by encoding using different encoding strategies. In one example, the aircraft-specific terminology corpus can be used to split the target speech recognition text to obtain multiple aircraft-specific terms contained in the target speech recognition text. For any aircraft-specific terminology, if the aircraft-specific terminology corpus contains the terminology, its corresponding encoding data can be directly determined, otherwise it can be encoded to obtain its corresponding encoding data, and then the first encoding data is obtained according to the encoding data corresponding to each aircraft-specific term contained in the target speech recognition text.

[0026] Step S106, transmitting the first coded data to terminals associated with other pilots, so that the first coded data is converted into target speech recognition text based on the aircraft-specific terminology corpus through the terminals associated with other pilots, and the target speech recognition text corresponding to the first coded data is broadcast, so as to realize voice transmission between two or more pilots.

[0027] In one example, a terminal associated with a pilot encrypts first coded data and then transmits it to terminals associated with other pilots; the terminals associated with other pilots decrypt the received first coded data, and then use an aircraft-specific terminology corpus to convert the coded data contained in the first coded data into corresponding aircraft-specific terminology, and then generate target speech recognition text based on the aircraft-specific terminology, and finally broadcast the target speech recognition text in standard pronunciation.

[0028] The aircraft simulation training speech coding transmission method provided in the embodiment of the present invention uses a pilot's personal speech library to train a speech recognition model for aircraft-specific terms, thereby obtaining a speech recognition model dedicated to the pilot, thereby significantly improving the accuracy of speech recognition of aircraft-specific terms; in addition, the target speech recognition text is converted into first coding data through the aircraft-specific terminology corpus, and then transmitted to other pilot-associated terminals in the form of coding data, and the terminal broadcasts the target speech recognition text corresponding to the first coding data, which can not only greatly reduce the communication bandwidth consumed in the transmission process and improve the transmission efficiency, greatly broaden the transmission means, but also achieve the effects of reducing noise, shielding the pilot's personal accent, and standardizing speech broadcasting.

[0029] For ease of understanding, the embodiments of the present invention provide Figure 2 The technical framework diagram of a method for transmitting speech coding for aircraft simulation training shown in FIG. 1 includes: S202, collecting aircraft-specific terms to form a corpus.

[0030] S204 , using Huffman coding to encode the aircraft-specific terms based on the first usage frequency of the aircraft-specific terms (aircraft-specific terms not included in the catalog are uniformly encoded using UTF-8 extension code) to form a coding rule library.

[0031] In one embodiment, the first usage frequency of each aircraft-specific term included in the aircraft-specific term corpus is counted, and each aircraft-specific term is encoded according to the encoding strategy corresponding to the first usage frequency to obtain the encoding data corresponding to each aircraft-specific term; wherein, when the first usage frequency is higher than the first preset threshold, the encoding strategy is to encode each aircraft-specific term with one byte; when the first usage frequency is lower than the first preset threshold, the encoding strategy is to encode each aircraft-specific term with two bytes. Optionally, commonly used sentences can also be encoded as aircraft-specific terms.

[0032] Optionally, the second usage frequency of aircraft-specific terms that are not included in the aircraft-specific terminology corpus is counted, and when the second usage frequency is higher than the second preset threshold, each aircraft-specific term is encoded according to the encoding strategy corresponding to the second usage frequency, and the obtained encoding data is added to the extension code of the aircraft-specific terminology corpus. The second preset threshold is lower than the first preset threshold, and for some aircraft-specific terms with a lower first usage frequency (that is, the second usage frequency is lower than the second preset threshold), it is possible to choose not to encode them, and reserve extension codes for them, so that when the second usage frequency is higher than the second preset threshold, after encoding them according to the corresponding encoding strategy, the code corresponding to the aircraft-specific term is included in the extension code.

[0033] Specifically, according to the frequency of use of aircraft-specific terms, Huffman coding and fixed-length strategy are used to encode the corpus. Figure 3 A schematic diagram of a coding rule is shown, using the first two digits to distinguish between unusual terms, high-frequency terms and basic terms. The 00 code supports UTF-8 coding of less than 63 sentences, the 01 code is encoded with a maximum frequency of 63 words, and 10 and 11 are other commonly used flight terms that use 2 bytes to encode and support a total of 32,766 words. In a specific implementation, the 01 code corresponds to high-frequency terms (i.e., aircraft-specific terms with a first usage frequency higher than a first preset threshold), the 10 code and the 11 code correspond to basic terms (i.e., aircraft-specific terms with a first usage frequency lower than a first preset threshold), the 00 code corresponds to a sentence, and when the second usage frequency of the aircraft-specific terms contained therein is higher than the second preset threshold, the coding data corresponding to the corresponding aircraft-specific terms is encoded into the extension code.

[0034] S206, collecting the pilot's voice input according to the aircraft-specific terms to form a personal voice database.

[0035] The voice of each trainee is collected to form a personal voice library, which is initially based on 1,000 voice texts, and is subsequently enriched based on the collected and recognized voices. Since the 1,000 basic voice texts have covered about 20% of the common phrases in flight simulation training, they are sufficient for accurately identifying each flight trainer. Voice collection requires each trainee to collect according to the corpus template, such as "01 01 I am 02", "Escort me", and "Climb to 3,500 meters". Therefore, the collected data is targeted, so its recognition rate is extremely high.

[0036] S208, using the personal voice database to train the speech recognition model and optimize the speech recognition model parameters.

[0037] Model architecture: For speech recognition with accents, more complex model architectures may be needed, such as deep neural networks (DNNs), convolutional neural networks (CNNs) combined with recurrent neural networks (RNNs), especially long short-term memory networks (LSTMs) or gated recurrent units (GRUs), to capture long-term dependencies and accent characteristics in speech.

[0038] Model training: Use the personal voice library to train the voice recognition model for each pilot, and optimize the relevant parameter data of the voice recognition model until the voice recognition model reaches the set indicators. The voice recognition model is relatively mature. The embodiment of the present invention can use a general voice recognition algorithm, but it is limited in the specific application field, and the voice and corresponding text are optimized to improve the efficiency of dynamic voice recognition. Optionally, the voice recognition model can also be trained in a multi-task learning mode, such as using accent recognition as an auxiliary task and voice recognition as the main task, to improve the accuracy of the voice recognition model in recognizing voice data with an accent.

[0039] S210, randomly perform a speech test on the speech recognition model, and update the speech recognition model after the test passes.

[0040] S212, collect the pilot's voice data, generate the corresponding target speech room recognition text through the speech recognition model, and continuously optimize the speech recognition model.

[0041] The process of speech recognition is as follows: first, the speech data is input into the speech recognition model; then, the time series features and local features corresponding to the speech data are extracted through the speech recognition model, and the time series features and local features are screened respectively, and the time series features after feature screening and the local features after feature screening are fused to output multiple speech recognition texts corresponding to the speech data based on the fused features; finally, the target speech recognition text is determined from the speech recognition text.

[0042] Among them, time series features include Mel frequency cepstral coefficients and their variant features, gene frequency and energy features, spectral flatness and spectral centroid features, and local features include zero crossing rate and short-time energy features, formant and form resistance features, and voice activity detection features. Mel frequency cepstral coefficients (MFCC) and their variant features are standard speech features. For accented speech, the number and spacing of Mel filters may need to be adjusted to better capture the spectral features unique to the accent; gene frequency and energy features vary depending on the accent, so the accented speech data can be better recognized by extracting fundamental frequency and energy features; spectral flatness and spectral centroid features can capture the spectral shape of speech data, which may be helpful for distinguishing different accents; zero crossing rate and short-time energy features are useful for all speech, but when processing accented speech data, the threshold or parameters may need to be adjusted to adapt to different speech characteristics; formant and form resistance features are important for identifying the speech structure of a specific accent; voice activity detection feature (VAD): When processing accented speech data, accurate VAD can help remove silence and non-speech parts, thereby improving the quality of the feature.

[0043] The process of continuously optimizing the speech recognition model includes the following (1) to (5): (1) Label the weight of each speech recognition text output by the speech recognition model. The weight of the target speech recognition text is higher than the weight of other speech recognition texts except the target speech recognition text.

[0044] (2) The speech data, each speech recognition text output by the speech recognition model for the speech data, and its weight are used as sample data.

[0045] (3) Assign an initial sampling probability to the sample data and save the sample data and its sampling probability to the personal voice database.

[0046] In one embodiment, the sampling probability of sample data is dynamically adjustable, and the sampling probability of sample data is related to the number of times the sample data is sampled. For example, if a certain sample data is sampled a higher number of times, the sampling probability of the sample data will decrease, otherwise the sampling probability of the sample data will increase.

[0047] In another embodiment, the sampling probability of sample data is dynamically adjustable, and the sampling probability of sample data is related to whether it was sampled in the last sampling process. For example, if a certain sample data was sampled as target sample data last time, the sampling probability of the sample data will be lowered according to a preset ratio; conversely, if a certain sample data was not sampled as target sample data last time, the sampling probability of the sample data will be increased according to a preset ratio.

[0048] In another embodiment, the sampling probability of the sample data may be related to the number of times the sample data is sampled and whether the sample data is sampled in the previous sampling process.

[0049] (4) According to the preset sampling ratio and based on the sampling probability of each sample data, the sample data stored in the personal voice database is sampled to obtain a target sample data set, and the sampling probability of each sample data stored in the personal voice database is adjusted using the target sample data set. The adjustment process can be referred to in (3) above, and will not be described in detail in the embodiment of the present invention.

[0050] (5) Use the target sample data set to optimize the speech recognition model.

[0051] S214, based on a pre-built aircraft-specific terminology corpus, convert the target speech recognition text into first encoded data.

[0052] The encoding process of the target speech recognition text data includes the following (a) to (c): (a) Based on the aircraft-specific terminology corpus, the target speech recognition text is split to obtain multiple aircraft-specific terms contained in the target speech recognition text; (b) For any aircraft-specific term, the following operations are performed: if the aircraft-specific term is included in the aircraft-specific term corpus, the coded data corresponding to the aircraft-specific term is determined based on the correspondence between the aircraft-specific term and the coded data; if the aircraft-specific term is not included in the aircraft-specific term corpus, the aircraft-specific term is encoded to obtain its corresponding coded data.

[0053] In the specific implementation, if the corpus contains the split aircraft-specific terms, the encoding data corresponding to the aircraft-specific terms can be directly called; if the corpus does not contain the split aircraft-specific terms, the aircraft-specific terms are encoded in UTF-8 encoding to obtain the corresponding encoding data.

[0054] (c) generating first coded data corresponding to the target speech recognition text based on the coded data corresponding to each aircraft-specific term included in the target speech recognition text. In one example, the corresponding coded data may be concatenated in the order of the aircraft-specific terms to obtain the first coded data.

[0055] S216, transmitting the first coded data to terminals associated with other pilots, so that the first coded data is converted into target speech recognition text based on the aircraft-specific terminology corpus through the terminals associated with other pilots, and the target speech recognition text corresponding to the first coded data is broadcast, so as to realize voice transmission between two or more pilots.

[0056] Optionally, when transmitting the first coded data, the first coded data may be encrypted at least once, and then the encrypted first coded data may be transmitted to terminals associated with other pilots. The embodiment of the present invention exemplarily provides a specific implementation method for encrypting the first coded data at least once, including: determining a hash value corresponding to the first coded data, first encrypting the data using a symmetric encryption algorithm (such as AES), then encrypting the key of the symmetric encryption algorithm using an asymmetric encryption algorithm (such as RSA), obtaining the encrypted first coded data, and transmitting the hash value and the encrypted first coded data together to terminals associated with other pilots.

[0057] Furthermore, the terminal associated with the pilot can also receive second coded data transmitted by terminals associated with other pilots, first decode the second coded data; then, based on the mapping relationship between aircraft-specific terms and the coded data, decode the second coded data into multiple aircraft-specific terms; finally, generate target speech recognition text corresponding to the second coded data based on the multiple aircraft-specific terms obtained by decoding, and broadcast the target speech recognition text corresponding to the second coded data in standard pronunciation. In one example, the decoded aircraft-specific terms can be directly spliced ​​to obtain the target speech recognition text.

[0058] The embodiment of the present invention converts speech into encoded encrypted text for transmission, thereby greatly reducing the communication bandwidth, and performs speech training and proprietary terminology training for specific pilots, thereby greatly improving the accuracy of speech recognition.

[0059] The embodiment of the present invention provides a specific implementation of a method for speech coding and transmission of aircraft simulation training. The embodiment of the present invention takes pilot A and pilot B as examples and provides a method as follows: Figure 4 The figure shows a practical application flow chart of a method for speech coding and transmission in aircraft simulation training.

[0060] Step 1: The onboard headset carried by the aircraft where pilot A is located collects the voice data of pilot A, and generates the target voice recognition text corresponding to the voice data through the voice recognition model using the terminal associated with pilot A. The generated text needs to be selected, confirmed and sent by the pilot.

[0061] Step 2: The terminal associated with pilot A saves pilot A's speech data and its corresponding target speech recognition text into pilot A's personal speech library for continuous optimization of the speech recognition model.

[0062] Step 3: The terminal associated with pilot A splits the target speech recognition text through the corpus and encodes it according to the encoding rule to obtain the first encoded data. Optionally, encryption processing can be further performed after encoding.

[0063] Step 4: The terminal associated with pilot A transmits the first coded data to the terminal associated with pilot B remotely via a communication link such as a network or radio.

[0064] Step 5: The terminal associated with pilot B receives the first coded data, decrypts it, and decodes it into the target speech recognition text according to the coding rules and corpus.

[0065] In step 6, the terminal associated with pilot B broadcasts the target speech recognition text through standard pronunciation, and pilot B receives the voice message sent by pilot A. It should be noted that the same process is required for pilot B to reply.

[0066] The aircraft simulation training voice coding and transmission method provided by the embodiment of the present invention has at least the following characteristics: 1. Since the voice is communicated and transmitted in a mixed coding method of text and common words, the transmission bandwidth of the present invention is very low, so it can also be transmitted through Beidou short messages, which greatly broadens the means of transmission; 2. Similarly, since text is used for transmission, noise is reduced, personal accents are shielded, and voice broadcasting is standardized; 3. Since the embodiment of the present invention is mainly aimed at flight training terms, it has a high accuracy rate for voice recognition.

[0067] Based on the above-mentioned embodiment, the embodiment of the present invention provides an aircraft simulation training voice coding transmission device, which is applied to the terminal associated with the pilot, and the terminals associated with different pilots are connected to each other in communication, see Figure 5 The structure diagram of a speech coding transmission device for aircraft simulation training is shown, and the device mainly includes the following parts: A speech recognition module 502 is used to collect the pilot's speech data and generate a target speech recognition text corresponding to the speech data using a speech recognition model; wherein the speech recognition model is trained based on the pilot's personal speech library, and the speech data and the corresponding target speech recognition text are also used to optimize the speech recognition model; A text encoding module 504, configured to convert the target speech recognition text into first encoded data based on a pre-built aircraft-specific terminology corpus; The voice transmission module 506 is used to transmit the first coded data to the terminal associated with the other pilots, so as to convert the first coded data into a target voice recognition text based on the aircraft-specific terminology corpus through the terminal associated with the other pilots, and broadcast the target voice recognition text corresponding to the first coded data, so as to realize voice transmission between two or more pilots.

[0068] The aircraft simulation training speech coding transmission device provided in the embodiment of the present invention uses the pilot's personal speech library to train the speech recognition model for aircraft-specific terms, and obtains a speech recognition model dedicated to the pilot, thereby significantly improving the accuracy of speech recognition for aircraft-specific terms; in addition, the target speech recognition text is converted into first coding data through the aircraft-specific terminology corpus, and then transmitted to other pilot-associated terminals in the form of coding data, and the terminal broadcasts the target speech recognition text corresponding to the first coding data, which can not only greatly reduce the communication bandwidth consumed in the transmission process and improve the transmission efficiency, greatly broaden the transmission means, but also achieve the effects of reducing noise, shielding the pilot's personal accent, and standardizing speech broadcasting.

[0069] In one embodiment, the aircraft-specific terminology corpus includes a plurality of aircraft-specific terms and their corresponding encoding data; the text encoding module 504 is specifically used to: Based on a pre-built aircraft-specific terminology corpus, the target speech recognition text is split to obtain a plurality of aircraft-specific terms contained in the target speech recognition text; For any aircraft-specific term, the following operations are performed: if the aircraft-specific term is included in the aircraft-specific term, based on the correspondence between the aircraft-specific term and the coded data, the coded data corresponding to the aircraft-specific term is determined; if the aircraft-specific term is not included in the aircraft-specific term, the aircraft-specific term is encoded to obtain the corresponding coded data; Based on the encoding data corresponding to each aircraft-specific term included in the target speech recognition text, first encoding data corresponding to the target speech recognition text is generated.

[0070] In one embodiment, a decoding module is further included, which is used to: receiving second coded data transmitted by a terminal associated with another pilot; decoding the second coded data into a plurality of aircraft-specific terms based on a mapping relationship between the aircraft-specific terms and the coded data; A target speech recognition text corresponding to the second coding data is generated according to the multiple aircraft-specific terms obtained by decoding, and the target speech recognition text corresponding to the second coding data is broadcasted in standard pronunciation.

[0071] In one embodiment, a precoding module is further included, which is used to: Counting the first usage frequency of each aircraft-specific term included in the aircraft-specific term corpus, encoding each aircraft-specific term according to the encoding strategy corresponding to the first usage frequency, and obtaining encoding data corresponding to each aircraft-specific term; Among them, when the first usage frequency is higher than the first preset threshold, the encoding strategy is to encode each aircraft-specific term with one byte; when the first usage frequency is lower than the first preset threshold, the encoding strategy is to encode each aircraft-specific term with two bytes.

[0072] In one implementation, the precoding module is further configured to: The second usage frequency of the aircraft-specific terms not included in the aircraft-specific terminology corpus is counted. When the second usage frequency is higher than a second preset threshold, each aircraft-specific term is encoded according to the encoding strategy corresponding to the second usage frequency, and the obtained encoding data is added to the extension code of the aircraft-specific terminology corpus.

[0073] In one embodiment, the target speech recognition text is determined from a plurality of speech recognition texts output by the speech recognition model, and further includes a model optimization module for: A weight is marked for each speech recognition text output by the speech recognition model; wherein the weight of the target speech recognition text is higher than the weight of other speech recognition texts except the target speech recognition text; The speech data, each speech recognition text output by the speech recognition model for the speech data, and its weight are used as sample data; Assigning an initial sampling probability to the sample data, and saving the sample data and its sampling probability to a personal voice database; wherein the sampling probability of the sample data is dynamically adjustable, and the sampling probability of the sample data is related to the number of times the sample data is sampled; According to a preset sampling ratio and based on the sampling probability of each sample data, the sample data stored in the personal voice database are sampled to obtain a target sample data set, and the sampling probability of each sample data stored in the personal voice database is adjusted using the target sample data set; The speech recognition model is optimized using the target sample data set.

[0074] In one implementation, the speech recognition module 502 is specifically used to: Input the speech data into the speech recognition model; Extracting time series features and local features corresponding to the speech data through the speech recognition model, performing feature screening on the time series features and the local features respectively, fusing the time series features after feature screening and the local features after feature screening, and outputting multiple speech recognition texts corresponding to the speech data based on the fused features; A target speech recognition text is determined from the speech recognition texts.

[0075] In one implementation, the voice transmission module 506 is specifically configured to: The first coded data is encrypted at least once to transmit the encrypted first coded data to a terminal associated with another pilot.

[0076] The device provided in the embodiment of the present invention has the same implementation principle and technical effects as those of the aforementioned method embodiment. For the sake of brief description, for matters not mentioned in the device embodiment, reference may be made to the corresponding contents in the aforementioned method embodiment.

[0077] An embodiment of the present invention provides an electronic device. Specifically, the electronic device includes a processor and a storage device. The storage device stores a computer program, and when the computer program is executed by the processor, it executes the method described in any one of the above-mentioned implementation methods.

[0078] Figure 6 A structural diagram of an electronic device provided in an embodiment of the present invention, the electronic device 100 includes: a processor 60, a memory 61, a bus 62 and a communication interface 63, wherein the processor 60, the communication interface 63 and the memory 61 are connected via the bus 62; the processor 60 is used to execute an executable module stored in the memory 61, such as a computer program.

[0079] The memory 61 may include a high-speed random access memory (RAM), and may also include a non-volatile memory, such as at least one disk memory. The communication connection between the system network element and at least one other network element is realized through at least one communication interface 63 (which may be wired or wireless), and the Internet, wide area network, local area network, metropolitan area network, etc. may be used.

[0080] The bus 62 may be an ISA bus, a PCI bus, or an EISA bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 6 Only one bidirectional arrow is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0081] Among them, the memory 61 is used to store programs, and the processor 60 executes the program after receiving the execution instruction. The method executed by the device for flow process definition disclosed in any embodiment of the above-mentioned embodiment of the present invention can be applied to the processor 60 or implemented by the processor 60.

[0082] The processor 60 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the hardware integrated logic circuit or software instructions in the processor 60. The above processor 60 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present invention can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in the embodiments of the present invention can be directly embodied as a hardware decoding processor to execute, or can be executed by a combination of hardware and software modules in the decoding processor. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory 61, and the processor 60 reads the information in the memory 61 and completes the steps of the above method in combination with its hardware.

[0083] The computer program product of the readable storage medium provided in the embodiment of the present invention includes a computer-readable storage medium storing program code, and the instructions included in the program code can be used to execute the methods described in the previous method embodiments. The specific implementation can be referred to the previous method embodiments, which will not be repeated here.

[0084] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0085] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above-described embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-described embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A method for transmitting speech coding in aircraft simulation training, characterized in that: The method is applied to terminals associated with pilots, and the terminals associated with different pilots are connected to each other in communication, and the method includes: Collecting the pilot's voice data, and generating a target voice recognition text corresponding to the voice data using a voice recognition model; wherein the voice recognition model is trained based on the pilot's personal voice library, and the voice data and the corresponding target voice recognition text are also used to optimize the voice recognition model; Based on a pre-built aircraft-specific corpus, converting the target speech recognition text into first encoded data; The first coded data is transmitted to the terminal associated with the other pilots, so that the first coded data is converted into the target speech recognition text based on the aircraft-specific terminology corpus through the terminal associated with the other pilots, and the target speech recognition text corresponding to the first coded data is broadcasted, so as to realize voice transmission between two or more pilots.

2. The aircraft simulation training speech coding transmission method according to claim 1, characterized in that: The aircraft-specific terminology corpus includes a plurality of aircraft-specific terms and their corresponding encoding data; Based on a pre-built aircraft-specific terminology corpus, converting the target speech recognition text into first coded data includes: Based on a pre-built aircraft-specific terminology corpus, the target speech recognition text is split to obtain a plurality of aircraft-specific terms contained in the target speech recognition text; For any of the aircraft-specific terms, the following operations are performed: if the aircraft-specific terminology corpus contains the aircraft-specific terminology, based on the correspondence between the aircraft-specific terminology and the coded data, the coded data corresponding to the aircraft-specific terminology is determined; if the aircraft-specific terminology corpus does not contain the aircraft-specific terminology, the aircraft-specific terminology is encoded to obtain the corresponding coded data; Based on the encoding data corresponding to each of the aircraft-specific terms included in the target speech recognition text, first encoding data corresponding to the target speech recognition text is generated.

3. The aircraft simulation training speech coding transmission method according to claim 2 is characterized in that: The method further comprises: receiving second coded data transmitted by the terminal associated with the other pilot; decoding the second coded data into a plurality of aircraft-specific terms based on a mapping relationship between the aircraft-specific terms and the coded data; The target speech recognition text corresponding to the second coding data is generated according to the multiple aircraft-specific terms obtained by decoding, and the target speech recognition text corresponding to the second coding data is broadcasted in standard pronunciation.

4. The aircraft simulation training speech coding transmission method according to claim 1, characterized in that: The method further comprises: Counting the first usage frequency of each aircraft-specific term included in the aircraft-specific term corpus, encoding each aircraft-specific term according to the encoding strategy corresponding to the first usage frequency, and obtaining encoding data corresponding to each aircraft-specific term; Among them, when the first usage frequency is higher than the first preset threshold, the encoding strategy is to encode each of the aircraft-specific terms with one byte; when the first usage frequency is lower than the first preset threshold, the encoding strategy is to encode each of the aircraft-specific terms with two bytes.

5. The aircraft simulation training speech coding transmission method according to claim 4 is characterized in that: The method further comprises: The second usage frequency of the aircraft-specific terms not included in the aircraft-specific terminology corpus is counted, and when the second usage frequency is higher than a second preset threshold, each of the aircraft-specific terms is encoded according to the encoding strategy corresponding to the second usage frequency, and the obtained encoding data is added to the extension code of the aircraft-specific terminology corpus.

6. The aircraft simulation training speech coding transmission method according to claim 1, characterized in that: The target speech recognition text is determined from a plurality of speech recognition texts output by the speech recognition model, and the method further comprises: Marking a weight for each of the speech recognition texts output by the speech recognition model; wherein the weight of the target speech recognition text is higher than the weights of other speech recognition texts except the target speech recognition text; Taking the speech data, each speech recognition text output by the speech recognition model for the speech data and its weight as sample data; Assigning an initial sampling probability to the sample data, and saving the sample data and the sampling probability to the personal voice database; wherein the sampling probability of the sample data is dynamically adjustable, and the sampling probability of the sample data is related to the number of times the sample data is sampled; According to a preset sampling ratio and based on the sampling probability of each sample data, the sample data stored in the personal voice database are sampled to obtain a target sample data set, and the sampling probability of each sample data stored in the personal voice database is adjusted by using the target sample data set; The speech recognition model is optimized using the target sample data set.

7. The aircraft simulation training speech coding transmission method according to claim 1, characterized in that: Generating a target speech recognition text corresponding to the speech data using a speech recognition model includes: Inputting the speech data into a speech recognition model; Extracting time series features and local features corresponding to the voice data through the voice recognition model, performing feature screening on the time series features and the local features respectively, fusing the time series features after feature screening and the local features after feature screening, and outputting a plurality of voice recognition texts corresponding to the voice data based on the fused features; A target speech recognition text is determined from the speech recognition texts.

8. An aircraft simulation training voice coding transmission device, characterized in that: The device is applied to terminals associated with pilots, and the terminals associated with different pilots are connected to each other in communication, and the device includes: A speech recognition module, used for collecting the pilot's speech data and generating a target speech recognition text corresponding to the speech data using a speech recognition model; wherein the speech recognition model is trained based on the pilot's personal speech library, and the speech data and the target speech recognition text corresponding to the speech data are also used for optimizing the speech recognition model; A text encoding module, configured to convert the target speech recognition text into first encoded data based on a pre-built aircraft-specific terminology corpus; A voice transmission module is used to transmit the first coded data to the terminal associated with the other pilots, so as to convert the first coded data into the target voice recognition text based on the aircraft-specific terminology corpus through the terminal associated with the other pilots, and broadcast the target voice recognition text corresponding to the first coded data, so as to realize voice transmission between two or more pilots.

9. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the memory stores computer executable instructions that can be executed by the processor, and the processor executes the computer executable instructions to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are called and executed by a processor, the computer-executable instructions prompt the processor to implement the method according to any one of claims 1 to 7.

Images