A listening and interpreting broadcast and standard broadcast and listening and interpreting broadcast and standard broadcast method for maritime very high frequency communication

Abstract

The application discloses a listening and interpreting broadcast and standard broadcast and listening and interpreting broadcast and standard broadcast method for maritime very high frequency communication, and relates to the technical field of radio communication. In the listening and interpreting broadcast link, a radio frequency control unit samples the audio level of a radio communication terminal, determines the start and stop of an incoming call and reports the event, a host unit collects audio, cuts and sends to an acceleration calculation unit, the acceleration calculation unit sequentially executes noise reduction, speech recognition, term correction and faithful direct translation and returns to the host unit. In the mother tongue transcription sending link, the host unit obtains the mother tongue text and sends to the acceleration calculation unit, the acceleration calculation unit executes standard translation, the host unit performs broadcast standardization processing on the incoming call translation text, the acceleration calculation unit returns after synthesizing the speech, and the host unit controls the radio frequency control unit to make the radio communication terminal enter the transmitting state and play the synthesized audio. The application can realize full offline, automatic, high-precision, high-safety maritime very high frequency multilingual real-time communication.

Description

Technical Field

[0001] This application relates to the field of radio communication technology, and in particular to a method for interpreting and broadcasting standardized messages and messages for maritime VHF communication. Background Technology

[0002] In multinational maritime operations, VHF radio voice communication is the core communication method for close-range interaction between ships and between ships and shore, and is widely used in critical scenarios such as ship avoidance, waterway communication, operational coordination, and emergency rescue. Due to significant differences in the native languages ​​of seafarers from different countries, cross-language communication barriers have long existed, which can easily lead to misunderstandings and misinterpretations of instructions, seriously affecting maritime communication efficiency and navigation safety.

[0003] Currently, there is a lack of translation assistance solutions in the industry for cross-language VHF communication at sea. Existing alternative technologies all have obvious shortcomings in scenario adaptation and cannot meet the real-time, stable, professional, and highly secure communication needs at sea. The specific problems are as follows: First, the purely manual translation model is highly dependent on human translation and lacks sufficient support capabilities. Traditional maritime cross-language communication relies heavily on onboard translators or crew members learning foreign languages ​​on their own. This model is costly in terms of manpower and has poor staffing flexibility. Furthermore, human translation suffers from inherent speech delays, making it unsuitable for the communication characteristics of VHF real-time intercom. Moreover, during nighttime duty or in sudden emergency maritime scenarios, human resources are limited and response times are slow, making it impossible to achieve uninterrupted and highly reliable communication and translation support.

[0004] Secondly, cloud-based online voice translation solutions have extremely low availability at sea. Mainstream commercial speech recognition, machine translation, and speech synthesis services all rely on stable internet cloud data interaction. However, remote sections of ocean and coastal waters generally suffer from weak network signals and lack of internet coverage, creating network communication blind spots. Cloud-based online translation solutions cannot operate independently offline; they become completely ineffective once disconnected from the network, making them unsuitable for the communication scenarios required for a ship's entire voyage.

[0005] Third, existing modular splicing solutions suffer from poor system integration and insufficient communication accuracy and professionalism. Currently, most localized voice translation modules or devices are simply physical splices of general speech recognition, ordinary machine translation, and speech synthesis modules. Each functional module operates independently without dedicated linkage control logic, and they are not customized for maritime VHF communication scenarios. On the one hand, the lack of dedicated channel monitoring and timing scheduling mechanisms prevents automatic identification of channel call status, requiring manual monitoring and control of transmission and reception, which easily leads to problems such as missing words, mis-sent commands, and transmission / reception errors. On the other hand, general translation models are not adapted to the maritime professional system, lacking standardized correction capabilities for maritime terminology, channel coordinates, vessel numbers, and operational digital information, resulting in extremely high misjudgment rates in the recognition and translation of maritime terminology and professional instructions.

[0006] In summary, existing technologies generally suffer from a series of problems, such as reliance on manual labor and networks, fragmented modules, insufficient professionalism, lack of time-series control, and lack of emergency response capabilities, making it impossible to achieve fully offline, automated, high-precision, and highly secure maritime VHF multilingual real-time communication. Summary of the Invention

[0007] The purpose of this application is to provide a method for interpreting and broadcasting standardized messages and messages for maritime VHF communication, so as to achieve fully offline, automated, high-precision, and highly secure real-time multilingual maritime VHF communication.

[0008] To achieve the above objectives, this application provides the following solution.

[0009] In a first aspect, this application provides a method for interpreting and broadcasting and standardizing information transmission for maritime VHF communications. The method is applied to an interpreting and broadcasting and standardizing information transmission system including a radio frequency control unit, a host unit, and an accelerated computing unit. The method includes: executing an interpreting and broadcasting link and a native language transcription and transmission link respectively. The audio-visualization and broadcasting link includes: The radio frequency control unit continuously samples the audio level of the radio communication terminal of the maritime VHF communication at a preset listening frequency. When the audio level is continuously higher than the preset level threshold and the duration reaches the preset confirmation time, it determines that the incoming call has started and reports the incoming call start event to the host unit. When the audio level is continuously lower than the preset level threshold and the duration reaches the preset silence time, it determines that the incoming call has ended and reports the incoming call end event to the host unit. After receiving a call start event, the host unit sends a call acquisition path establishment command to the radio frequency control unit. The radio frequency control unit connects the acquisition path from the radio communication terminal to the audio acquisition port of the host unit according to the call acquisition path establishment command. The host unit acquires the call audio received by the radio communication terminal through the acquisition path. After receiving a call end event, the host unit stops acquiring the call audio, performs voice activity detection on the acquired call audio and segments it by sentence to obtain multiple raw audio segments, and sends each raw audio segment to the acceleration computing unit. The accelerated computing unit performs noise reduction processing on each original audio segment to obtain the corresponding enhanced audio segment. Then, it sequentially performs speech recognition, forced substitution correction, and faithful translation on each enhanced audio segment to obtain the corresponding translated speech. The unit then returns each original audio segment, the corresponding enhanced audio segment, and the translated speech to the host unit. The host unit plays each original audio segment and the corresponding enhanced audio segment on the local player. The native language transcription transmission link includes: The host unit acquires the native language text of its personnel and sends the native language text to the accelerated computing unit; The accelerated computing unit performs a standardized translation on the native language text to obtain a standardized translation, and then sends the standardized translation to the host unit; The host unit performs normalization processing on the numbers in the standard translation according to the target language rules to obtain normalized text, and then sends the normalized text to the accelerated computing unit. The accelerated computing unit performs speech synthesis on the read-out standardized text to obtain synthesized audio, and returns the synthesized audio to the host unit. The host unit sends a PTT enable command and a transmit path establishment command to the radio frequency control unit. The radio frequency control unit enables the radio communication terminal to enter the transmit state according to the PTT enable command, and connects the transmit path from the host unit's audio playback port to the transmit port of the radio communication terminal according to the transmit path establishment command. After the radio communication terminal is in the transmit state and the transmit path has been established, the host unit plays the synthesized audio, and after the playback ends, sends a PTT release command and a transmit path shutdown command to the radio frequency control unit. The radio frequency control unit enables the radio communication terminal to exit the transmit state according to the PTT release command, and disconnects the transmit path from the host unit's audio playback port to the transmit port of the radio communication terminal according to the transmit path shutdown command.

[0010] In one embodiment, the preset listening frequency is 100Hz, the preset level threshold is a 12-bit ADC count of 100 to 200, the preset confirmation time is 80ms, and the preset silence time is 2.5s.

[0011] In one embodiment, during the sentence-by-sentence segmentation process, the host unit has a speech activity detection trigger threshold of 0.5, a minimum silence duration of 300ms, a speech pre- and post-speech filling duration of 200ms, and a sampling rate of 16kHz.

[0012] In one embodiment, the noise reduction process employs the DeepFilterNetv3 model.

[0013] In one embodiment, the speech recognition employs the faster-whisper-large-v3-turbo model.

[0014] In one embodiment, the forced replacement correction includes: based on a glossary, replacing terms in the text obtained from speech recognition with standard spellings according to the principle of longest match priority.

[0015] In one embodiment, the host unit acquires the native language text of its own personnel, including: When the radio frequency control unit detects that the voice input button has been pressed, it reports a voice input start event to the host unit. After receiving the voice input start event, the host unit sends a native language acquisition channel establishment command to the radio frequency control unit. The radio frequency control unit connects the audio acquisition port of the host unit according to the native language acquisition channel establishment command. The host unit acquires the native language audio of its personnel through the channel. When the radio frequency control unit detects that the voice input button has been released, it reports a voice input end event to the host unit. After receiving the voice input end event, the host unit stops acquiring native language audio and sends the acquired native language audio to the accelerated computing unit. The accelerated computing unit sequentially performs noise reduction processing, speech recognition, and forced substitution correction on the native language audio to obtain native language text, and sends the native language text to the host unit. Alternatively, the host unit may receive native language text directly input by its own personnel.

[0016] In one embodiment, the standardization process for broadcasting includes: using NATO / ICAO reading for English, using radio standard reading for Chinese, converting numbers in the standardized translation of Japanese to kana reading, and using the corresponding conventional reading for Korean.

[0017] Secondly, this application provides a method for interpreting and broadcasting information for maritime VHF communications. The method is applied to an interpreting and broadcasting and specification broadcasting system including a radio frequency control unit, a host unit, and an acceleration computing unit. The method includes: executing an interpreting and broadcasting link. The audio-visualization and broadcasting link includes: The radio frequency control unit continuously samples the audio level of the radio communication terminal of the maritime VHF communication at a preset listening frequency. When the audio level is continuously higher than the preset level threshold and the duration reaches the preset confirmation time, it determines that the incoming call has started and reports the incoming call start event to the host unit. When the audio level is continuously lower than the preset level threshold and the duration reaches the preset silence time, it determines that the incoming call has ended and reports the incoming call end event to the host unit. After receiving a call start event, the host unit sends a call acquisition path establishment command to the radio frequency control unit. The radio frequency control unit connects the acquisition path from the radio communication terminal to the audio acquisition port of the host unit according to the call acquisition path establishment command. The host unit acquires the call audio received by the radio communication terminal through the acquisition path. After receiving a call end event, the host unit stops acquiring the call audio, performs voice activity detection on the acquired call audio and segments it by sentence to obtain multiple raw audio segments, and sends each raw audio segment to the acceleration computing unit. The accelerated computing unit performs noise reduction processing on each original audio segment to obtain the corresponding enhanced audio segment. Then, it sequentially performs speech recognition, forced substitution correction, and faithful translation on each enhanced audio segment to obtain the corresponding translated speech. The unit then returns each original audio segment, the corresponding enhanced audio segment, and the translated speech to the host unit. The host unit plays each original audio segment and the corresponding enhanced audio segment on the local player.

[0018] Thirdly, this application provides a standard broadcasting method for maritime VHF communication, which is applied to a listening and interpreting broadcasting and standard broadcasting system including a radio frequency control unit, a host unit and an acceleration computing unit. The standard broadcasting method includes: executing a native language transcription transmission link. The native language transcription transmission link includes: The host unit acquires the native language text of its personnel and sends the native language text to the accelerated computing unit; The accelerated computing unit performs a standardized translation on the native language text to obtain a standardized translation, and then sends the standardized translation to the host unit; The host unit performs normalization processing on the numbers in the standard translation according to the target language rules to obtain normalized text, and then sends the normalized text to the accelerated computing unit. The accelerated computing unit performs speech synthesis on the read-out standardized text to obtain synthesized audio, and returns the synthesized audio to the host unit. The host unit sends a PTT enable command and a transmit path establishment command to the radio frequency control unit. The radio frequency control unit enables the maritime VHF radio communication terminal to enter the transmit state according to the PTT enable command, and connects the transmit path from the host unit's audio playback port to the transmit port of the radio communication terminal according to the transmit path establishment command. After the radio communication terminal is in the transmit state and the transmit path has been established, the host unit plays the synthesized audio, and after the playback ends, sends a PTT release command and a transmit path shutdown command to the radio frequency control unit. The radio frequency control unit enables the radio communication terminal to exit the transmit state according to the PTT release command, and disconnects the transmit path from the host unit's audio playback port to the transmit port of the radio communication terminal according to the transmit path shutdown command.

[0019] According to the specific embodiments provided in this application, the following technical effects are disclosed: This application discloses a method for interpreting and broadcasting standard audio in maritime VHF communication. By continuously sampling audio levels and combining this with time accumulation for discrimination, the system automatically determines the start and end of incoming calls and reports events, enabling subsequent processing links to be triggered without manual monitoring. The host unit and accelerated computing unit locally collaborate to complete acquisition, segmentation, noise reduction, recognition, terminology correction, translation, and synthesis, forming a complete interpreting and broadcasting and native language transcription transmission link without relying on an external network. The host unit controls the RF control unit to connect the transmission channel and play synthesized audio by issuing PTT enable commands and transmission path establishment commands, ensuring strict synchronization between transmission timing and audio broadcasting. This avoids word loss, out-of-order issues, or channel occupancy conflicts that may occur due to manual switching. Simultaneously, a mutual exclusion state machine ensures that only one of the two links is active at any given time, thus achieving automatic closed-loop transmission and reception of full-duplex maritime VHF communication without human intervention. This results in fully offline, automated, high-precision, and highly secure real-time multilingual maritime VHF communication. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 A schematic flowchart of a method for interpreting and broadcasting messages for maritime VHF communications provided in an embodiment of this application; Figure 2 A schematic diagram of a standard broadcasting method for maritime VHF communication provided in an embodiment of this application; Figure 3 This is a schematic diagram of the structure for listening, translating, broadcasting, and standardized broadcasting. Detailed Implementation

[0022] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0023] The purpose of this application is to provide a method for interpreting and broadcasting standardized messages and messages for maritime VHF communication, aiming to achieve fully offline, automated, high-precision, and highly secure real-time multilingual maritime VHF communication.

[0024] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0025] In one exemplary embodiment, such as Figure 1 and Figure 2 As shown, a method for interpreting and broadcasting standardized messages in maritime VHF communications is provided. This method is applied to applications such as... Figure 3 The system shown includes a radio frequency control unit, a host unit, and an accelerated computing unit. The method of listening, translating, and transmitting standard text includes: executing the listening, translating, and transmitting link and the native language transcription transmission link respectively.

[0026] The listening, translation, and broadcasting process includes the following steps.

[0027] Step 11: The radio frequency control unit continuously samples the audio level of the radio communication terminal of the maritime VHF communication at a preset listening frequency. When the audio level is continuously higher than the preset level threshold and the duration reaches the preset confirmation time, it determines that the incoming call has started and reports the incoming call start event to the host unit. When the audio level is continuously lower than the preset level threshold and the duration reaches the preset silence time, it determines that the incoming call has ended and reports the incoming call end event to the host unit.

[0028] As an optional implementation, in step 11, the preset listening frequency is 100Hz, the preset level threshold is 100 to 200 for a 12-bit ADC count, the preset confirmation time is 80ms, and the preset silence time is 2.5s.

[0029] Specifically, the radio frequency control unit continuously samples the audio level of the speaker of the radio communication terminal after voltage division and DC blocking at a frequency of 100Hz. Its discrimination logic is as follows: when the sampled audio level is continuously higher than the level threshold and lasts for a preset confirmation duration (80ms), it determines that the incoming call has started and sends an incoming call start event (SIGNAL_ON) to the host unit; when the audio level is continuously lower than the threshold and lasts for a preset silence duration (2.5s), it determines that the incoming call has ended and sends an incoming call end event (SIGNAL_OFF). This step replaces the energy threshold of traditional voice-operated exchange (VOX) with time-cumulative discrimination, filtering out transient noises such as clicking sounds and carrier noise, and triggering the subsequent high-load recognition and translation process only when there is a confirmed incoming call.

[0030] In fact, the preset level threshold, preset confirmation time, and preset silence time of about 2.5 seconds can all be calibrated according to the aircraft model and sea state.

[0031] Step 12: After receiving the incoming call start event, the host unit sends an incoming call acquisition path establishment command to the radio frequency control unit. The radio frequency control unit connects the acquisition path from the radio communication terminal to the audio acquisition port of the host unit according to the incoming call acquisition path establishment command. The host unit acquires the incoming call audio received by the radio communication terminal through the acquisition path. After receiving the incoming call end event, the host unit stops acquiring the incoming call audio, performs voice activity detection on the acquired incoming call audio and segments it by sentence to obtain multiple raw audio segments, and sends each raw audio segment to the accelerated computing unit.

[0032] As an optional implementation, in step 12, during the sentence segmentation process, the host unit sets the trigger threshold for voice activity detection to 0.5, the minimum silence duration to 300ms, the pre- and post-speech filling duration to 200ms, and the sampling rate to 16kHz.

[0033] Specifically, after the Web UI on the host unit receives the SIGNAL_ON signal forwarded by the serial port proxy service via WebSocket (or can be manually monitored continuously on a specific channel, always triggering the LISTEN_ON event), the RF control unit connects the "incoming call → host acquisition port" path. The Web UI then acquires the incoming call audio from the sound card via the browser's WebAudio API, forming a streaming audio stream that is sent to the main process. The Web UI performs voice activity detection (VAD, trigger threshold 0.5, minimum silence duration 300ms, speech pre- and post-pause 200ms, sampling rate 16kHz) on the streaming audio and segments it by sentence to obtain segmented 16kHz mono PCM audio segments (i.e., original audio segments), which serve as the basic audio units for subsequent processing. The purpose of segmentation is to decompose continuous incoming calls into semantically complete short segments, facilitating segment-by-segment low-latency processing. In addition, it can remove invalid segments where the channel is connected but there is no voice input.

[0034] Step 13: The accelerated computing unit performs noise reduction processing on each original audio segment to obtain the corresponding enhanced audio segment. Then, it performs speech recognition, forced substitution correction, and faithful translation on each enhanced audio segment to obtain the corresponding translated speech. The original audio segments, the corresponding enhanced audio segments, and the translated speech are returned to the host unit. The host unit plays each original audio segment and the corresponding enhanced audio segment on the local player.

[0035] As an optional implementation, in step 13, the noise reduction process uses the DeepFilterNetv3 model.

[0036] As an optional implementation, in step 13, the speech recognition uses the faster-whisper-large-v3-turbo model.

[0037] As an optional implementation, in step 13, the forced replacement correction includes: based on the glossary, replacing the terms in the text obtained from speech recognition with the standard spelling according to the principle of longest match priority.

[0038] Specifically, the Web UI uploads each original audio segment to the noise reduction service of the accelerated computing unit via HTTP, performs real-time noise reduction (using the DeepFilterNetv3 model, with background noise attenuation intensity adjustable between approximately 10-100dB), and returns the enhanced audio segment (i.e., the enhanced audio segment). In maritime channels with high background noise and wind and wave noise, noise reduction can filter out invalid background noise, significantly improving the accuracy of subsequent speech recognition; simultaneously, the original audio segment before noise reduction is retained for playback comparison. The original audio segment and the enhanced audio segment after noise reduction are played back by the host unit sound card controlled by the Web UI through the local speaker.

[0039] The web UI sends the enhanced audio segment to the Speech-to-Text (STT) service (HTTP interface, currently implemented as Faster-WhisperLarge-v3-Turbo, CTranslate2int8 quantization GPU inference) to obtain language discrimination and recognized text results. Then, using a terminology list, the recognized text is forcibly replaced and corrected based on the longest match, correcting error-prone channel numbers, station names, and maritime abbreviations to standard spellings. This step truncates the most error-prone proper nouns from the recognition stage before translation, preventing errors from being amplified by the translation service.

[0040] The Web UI submits the corrected text to the translation service (HTTP interface, currently implemented using Tencent's HY-MT translation model), with prompts requiring "accurate translation, no additional explanations, and no stylistic bias" to ensure the sender accurately understands the recipient's original meaning. If the identified language is already the sender's language, translation is skipped. Faithful literal translation from the incoming call direction and standardized translation using International Maritime Organization Standard Marine Communication Phrases (IMO SMCP) from the outgoing call direction together constitute the asymmetric translation strategy of this method.

[0041] The native language transcription transmission chain follows these steps.

[0042] Step 21: The host unit obtains the native language text of its personnel and sends the native language text to the accelerated computing unit.

[0043] As an optional implementation, in step 21, the host unit acquires the native language text of its personnel, including: When the radio frequency control unit detects that the voice input button has been pressed, it reports a voice input start event to the host unit. After receiving the voice input start event, the host unit sends a native language acquisition channel establishment command to the radio frequency control unit. The radio frequency control unit connects the audio acquisition port of the host unit according to the native language acquisition channel establishment command. The host unit acquires the native language audio of its personnel through the channel. When the radio frequency control unit detects that the voice input button has been released, it reports a voice input end event to the host unit. After receiving the voice input end event, the host unit stops acquiring native language audio and sends the acquired native language audio to the accelerated computing unit. The accelerated computing unit sequentially performs noise reduction processing, speech recognition, and forced substitution correction on the native language audio to obtain native language text, and sends the native language text to the host unit.

[0044] Alternatively, the host unit may receive native language text directly input by its own personnel.

[0045] Step 22: The accelerated computing unit performs a normalized translation on the native language text to obtain a normalized translation, and sends the normalized translation to the host unit.

[0046] Specifically, the Web UI submits the native language text to the translation service. Prompt words constrain the translation to adopt the concise style of the IMO SMCP, and appropriate stylistic features are applied to different target languages ​​(e.g., using concise IMO SMCP sentence structures for English). If a glossary is enabled, a terminology comparison block is added to the prompt words, or terminology pre-substitution is performed on the native language text. This step ensures that the sent messages conform to maritime communication conventions and that proper nouns are consistent, representing the essential difference in translation strategies between the sending and receiving directions.

[0047] Step 23: The host unit performs reading normalization processing on the numbers in the standard translation according to the target language rules to obtain the reading normalized text, and sends the reading normalized text to the accelerated computing unit.

[0048] As an optional implementation, step 23, the standardization process for reading aloud includes: using NATO / ICAO reading for English, using radio standard reading for Chinese, converting numbers in the standardized translation of Japanese to kana reading, and using the corresponding conventional reading for Korean.

[0049] Specifically, after obtaining the normalized text for playback, the normalized text is displayed on the Web UI, and after user confirmation, it is synthesized and sent to the acceleration computing unit.

[0050] Step 24: The accelerated computing unit performs speech synthesis on the read-out standardized text to obtain synthesized audio, and returns the synthesized audio to the host unit.

[0051] Specifically, the Web UI submits the standardized text to the Text-to-Speech (TTS) service. Using the provider's pre-registered fixed voice (based on pre-recorded native language speech and corresponding text), the system synthesizes speech in the target language and returns it to the Web UI of the host unit. The voice used in the speech synthesis includes, but is not limited to, voices recorded by the provider. The purpose is to allow the provider to input any voice, but selectively synthesize audio from specific individuals and with specific tones during broadcast, maintaining consistency in voice and delivery across multiple languages ​​for easy listening and recognition. In scenarios requiring the sequential transmission of multilingual versions of messages (e.g., the same content transmitted in English, Chinese, and Japanese), using the provider's pre-registered fixed voice ensures consistency in the voiceprints of the three synthesized audio segments, making it easy for the receiver to identify them as multilingual broadcasts from the same source.

[0052] Step 25: The host unit sends a PTT (Push-to-Talk) enable command and a transmit path establishment command to the RF control unit; the RF control unit enables the radio communication terminal to enter the transmit state according to the PTT enable command, and connects the transmit path from the host unit's audio playback port to the radio communication terminal's transmit port according to the transmit path establishment command; after the radio communication terminal is in the transmit state and the transmit path has been established, the host unit plays synthesized audio, and after the playback ends, sends a PTT release command and a transmit path shutdown command to the RF control unit; the RF control unit enables the radio communication terminal to exit the transmit state according to the PTT release command, and disconnects the transmit path from the host unit's audio playback port to the radio communication terminal's transmit port according to the transmit path shutdown command.

[0053] Specifically, the synthesized audio can be previewed and confirmed for transmission via the Web UI, or it can be synthesized by default and sent directly. Subsequently, the Web UI sends a PTT enable command (PTT_ON) to the serial port proxy service, which forwards it to the RF control unit. The RF control unit then sequentially performs the following steps: connecting the "host playback port → radio transmission port" path, enabling PTT to put the radio communication terminal into transmission mode, and playing the synthesized audio via the Web UI. After the audio playback ends, the Web UI sends a PTT release command (PTT_OFF), and the RF control unit sequentially performs the steps of releasing PTT and disconnecting the transmission path to ensure that no words are lost at the beginning or end of the transmission and that no no-load mistransmissions occur.

[0054] Specifically, the two links of listening and translation broadcasting and native language forwarding share the following language processing strategy. The relevant rules are all based on an editable glossary, which users can maintain online.

[0055] (1) Mandatory pre-replacement by glossary. Before translation, the recognized text is subjected to mandatory replacement with longest match priority according to the glossary, to forcibly correct the high-error-rate content in the recognition stage, such as channel numbers (e.g., "Channel Sixty" → "CH16"), maritime terms (e.g., "Mayday" → "Mayday", "State" → "Say again"), and regional accented expressions (e.g., "Do you coffee?" → "Do you copy?") to standard writing, so as to prevent misrecognition from being further amplified by translation. The common glossary is stored in the high-frequency confusing term glossary, which allows users to add, delete and modify online.

[0056] (2) Asymmetric translation strategy. The two links share the same translation service, and behavior switching is realized through different translation strategy parameters: the incoming call direction is passed with faithful literal translation parameters, and the prompt restricts the translator to "accurately convey the original meaning, without additional explanation, and without applying a specific style"; the outgoing call direction is passed with the Standard Marine Communication Phrases (SMCP) normalization parameters, and the prompt restricts the translation to adopt the concise sentence pattern of IMO SMCP, and apply the corresponding style to different target languages. In addition, in the outgoing call direction, according to official diplomatic and legal documents, the standardized constraints on the exclusive translations involving specific island, reef and maritime area rights and interests are strengthened. The strategic difference between the two directions reflects the different requirements of maritime communication for "accurate listening" and "standard speaking".

[0057] (3) Digital broadcast normalization. Before speech synthesis, the digit strings in the text to be synthesized are converted into the customary reading methods for radio calls according to the rules of the target language, with the longest match priority to apply the multi-digit combination rule first and then perform digit-by-digit replacement: English adopts the NATO / ICAO reading method (e.g., Zero, Wun, Too), Chinese adopts the standard radio reading method (e.g., "1" → "Yao"), Japanese converts the text to be synthesized into Kana to prevent the synthesizer from mispronouncing Chinese characters with Chinese pronunciation, and Korean adopts a similar conventional reading method. The conversion rules are stored in an editable glossary, which allows users to add, delete and modify online. This step enables the synthesized speech to broadcast key digit strings such as channel numbers, coordinates, and call signs in a radio customary manner, which significantly improves the recognizability under noisy channels.

[0058] Specifically, the communication interface, mutex state machine and degradation mechanism are defined in the speech interpretation broadcast and standard transmission system.

[0059] 1. Communication interface definition.

[0060] Interface 1: Local HTTP / WebSocket interface between the acceleration computing unit and the host unit.

[0061] The accelerated computing unit runs four independent inference microservices: noise reduction, speech recognition (Speech-to-Text, STT), translation, and synthesis (Text-to-Speech, TTS). STT, translation, and synthesis are implemented using the FastAPI framework and each listens to an independent port to interact with the main web application backend: the STT service receives a 16 kHz mono PCM audio segment (i.e., the original audio segment) and returns the recognized text and language identification; the translation service receives the text and translation strategy parameters (faithful literal translation or conforming to IMOSMCP normalization) and returns the translated speech; the TTS service receives the translated speech and timbre identifier and returns the synthesized audio stream; the noise reduction function serves STT and TTS, receiving the original audio and returning the noise-reduced and enhanced audio. In addition, the serial port proxy service uses the WebSocket protocol to push serial port events reported by the RF control unit to the host unit in real time, and receives control commands from the host unit and forwards them to the RF control unit via the serial port; the main web application backend is responsible for session management, terminology access, and mutual exclusion state machine maintenance. The Web UI on the host unit runs in a browser, calls various inference services via HTTP interfaces, subscribes to hardware events pushed by the serial port proxy service via WebSocket, and directly calls the local sound card for audio signal acquisition and playback. This layering concentrates the inference workload on the accelerated computing unit and the audio I / O and human-computer interaction interface on the host unit, with the two loosely coupled and replaceable independently.

[0062] Interface 2: Custom serial protocol interface between the host unit and the RF control unit.

[0063] The serial port parameters are 115200 baud, 8 data bits, no parity, 1 stop bit, UTF-8 encoding, and newline character as the message delimiter. Events reported by the RF control unit to the host unit include: READY (Power-on ready), SIGNAL_ON / SIGNAL_OFF (Call start event / Call end event, triggered based on channel level time accumulation), BTN_DOWN / BTN_UP (Voice input button pressed / Voice input button released), DIRECT_TX_DOWN / DIRECT_TX_UP (Manual direct transmission button pressed / Manual direct transmission button released), ACK:<command> and ERR:<reason> (Acknowledgment and exception reporting). Control commands sent by the host unit to the RF control unit include: LISTEN_ON / LISTEN_OFF (Call acquisition path establishment command / Call acquisition path disconnection command), PTT_ON / PTT_OFF (Transmit path establishment command / Transmit path shutdown command), and DIRECT_OFF (Active termination of manual direct transmission).

[0064] 2. Mutual exclusion state machine and degradation mechanism.

[0065] The host unit (web application backend) maintains a five-state mutually exclusive state machine: Idle, Receiving, Transmit, Voice In, and Direct Transmission (DIRECT_TX). State transitions are driven by events reported by the RF control unit. Before each transition, a path shutdown command corresponding to the old state is issued first, and then the path required for the new state is established, ensuring that only one audio path is active at any given time. Simultaneously, a degraded path for direct transmission is retained: when DIRECT_TX_DOWN is triggered, the RF control unit directly connects the "home microphone → transmitter" path and enables PTT, completely bypassing the listening and broadcasting links, ensuring that emergency manual communication is never locked under any circumstances.

[0066] In one exemplary embodiment, a method for interpreting and broadcasting messages for maritime VHF communications is provided. This method is applied to applications such as... Figure 3 The system shown includes a radio frequency control unit, a host unit, and an accelerated computing unit, which is a listening, translation, and broadcasting system, including a listening, translation, and broadcasting link.

[0067] like Figure 1 As shown, the listening and translation broadcasting link includes the following steps.

[0068] Step 11: The radio frequency control unit continuously samples the audio level of the radio communication terminal of the maritime VHF communication at a preset listening frequency. When the audio level is continuously higher than the preset level threshold and the duration reaches the preset confirmation time, it determines that the incoming call has started and reports the incoming call start event to the host unit. When the audio level is continuously lower than the preset level threshold and the duration reaches the preset silence time, it determines that the incoming call has ended and reports the incoming call end event to the host unit.

[0069] Step 12: After receiving the incoming call start event, the host unit sends an incoming call acquisition path establishment command to the radio frequency control unit. The radio frequency control unit connects the acquisition path from the radio communication terminal to the audio acquisition port of the host unit according to the incoming call acquisition path establishment command. The host unit acquires the incoming call audio received by the radio communication terminal through the acquisition path. After receiving the incoming call end event, the host unit stops acquiring the incoming call audio, performs voice activity detection on the acquired incoming call audio and segments it by sentence to obtain multiple raw audio segments, and sends each raw audio segment to the accelerated computing unit.

[0070] Step 13: The accelerated computing unit performs noise reduction processing on each original audio segment to obtain the corresponding enhanced audio segment. Then, it performs speech recognition, forced substitution correction, and faithful translation on each enhanced audio segment to obtain the corresponding translated speech. The original audio segments, the corresponding enhanced audio segments, and the translated speech are returned to the host unit. The host unit plays each original audio segment and the corresponding enhanced audio segment on the local player.

[0071] In one exemplary embodiment, a specification broadcasting method for maritime VHF communications is provided, which is applied to applications such as... Figure 3 The system shown includes a radio frequency control unit, a host unit, and an accelerated computing unit, which is a transcription and standard broadcasting system, including a native language transcription transmission link.

[0072] like Figure 2 As shown, the native language transcription transmission chain follows these steps.

[0073] Step 21: The host unit obtains the native language text of its personnel and sends the native language text to the accelerated computing unit.

[0074] Step 22: The accelerated computing unit performs a normalized translation on the native language text to obtain a normalized translation, and sends the normalized translation to the host unit.

[0075] Step 23: The host unit performs reading normalization processing on the numbers in the standard translation according to the target language rules to obtain the reading normalized text, and sends the reading normalized text to the accelerated computing unit.

[0076] Step 24: The accelerated computing unit performs speech synthesis on the read-out standardized text to obtain synthesized audio, and returns the synthesized audio to the host unit.

[0077] Step 25: The host unit sends a PTT enable command and a transmit path establishment command to the RF control unit; the RF control unit enables the radio communication terminal to enter the transmit state according to the PTT enable command, and connects the transmit path from the host unit's audio playback port to the radio communication terminal's transmit port according to the transmit path establishment command; after the radio communication terminal is in the transmit state and the transmit path has been established, the host unit plays the synthesized audio, and after the playback ends, sends a PTT release command and a transmit path shutdown command to the RF control unit; the RF control unit enables the radio communication terminal to exit the transmit state according to the PTT release command, and disconnects the transmit path from the host unit's audio playback port to the radio communication terminal's transmit port according to the transmit path shutdown command.

[0078] Compared to methods such as human onboard translators, reliance on cloud-based voice services, or simple splicing of existing modules, the advantages of the method in this application and their sources are as follows.

[0079] 1. Fully offline usable, even at sea without network access. Stable internet access is generally unavailable on long-range maritime voyages, rendering cloud-based translation solutions (such as cloud-based voice service products) completely ineffective. This method completes all noise reduction, recognition, translation, and synthesis locally within the accelerated computing unit, without relying on any cloud infrastructure. This effectiveness stems from the resource scheduling design that enables a single accelerated unit to handle the entire process through local collaboration and serial scheduling across three units.

[0080] 2. Automatic transmission and reception, no manual monitoring required. By leveraging channel level time accumulation monitoring (step 1) and transmission timing scheduling linked to a custom serial port protocol (step 12), the system can automatically translate and broadcast incoming calls and translate and transmit outgoing calls without manual intervention. This effect stems from an automatic hardware-software coupling method that binds hardware monitoring events and the main software flow into a mutually exclusive state machine using two communication interfaces.

[0081] 3. The error rate of proper nouns and number strings is significantly reduced. Triple language correction is used: three layers of language correction (forced replacement of terminology, IMO SMCP style constraints, and standardization of number reading) are used to correct errors in three stages: after recognition, during translation, and before synthesis. Its effect is significantly better than general translation solutions.

[0082] 4. Incoming messages are accurately understood and responses conform to conventions. Faithful translation of incoming messages ensures accurate understanding of the original meaning, while responses can be standardized using IMO SMCP to ensure that sent messages conform to maritime communication standards. The asymmetric configuration balances both "accurate listening" and "standardized speaking".

[0083] The above effect is the result of the synergy of multiple modules integrated according to the maritime communication scenario: individual recognition, translation or synthesis models are all public technologies, but only by integrating them with noise reduction, three-layer language correction, timbre cloning, and time accumulation listening, event-driven link and transmission timing linkage realized through two communication interfaces into a mutually exclusive state machine can the overall capability of "fully offline, automatic transmission and reception, low false alarm and strong standardization" be obtained.

[0084] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0085] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0086] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for interpreting and broadcasting standardized messages in maritime VHF communication, characterized in that, The listening and translation broadcasting and standard broadcasting method is applied to a listening and translation broadcasting and standard broadcasting system including a radio frequency control unit, a host unit and an accelerated computing unit. The listening and translation broadcasting and standard broadcasting method includes: executing the listening and translation broadcasting link and the native language transcription transmission link respectively. The audio-visualization and broadcasting link includes: The radio frequency control unit continuously samples the audio level of the radio communication terminal of the maritime VHF communication at a preset listening frequency. When the audio level is continuously higher than the preset level threshold and the duration reaches the preset confirmation time, it determines that the incoming call has started and reports the incoming call start event to the host unit. When the audio level is continuously lower than the preset level threshold and the duration reaches the preset silence time, it determines that the incoming call has ended and reports the incoming call end event to the host unit. After receiving a call start event, the host unit sends a call acquisition path establishment command to the radio frequency control unit. The radio frequency control unit connects the acquisition path from the radio communication terminal to the audio acquisition port of the host unit according to the call acquisition path establishment command. The host unit acquires the call audio received by the radio communication terminal through the acquisition path. After receiving a call end event, the host unit stops acquiring the call audio, performs voice activity detection on the acquired call audio and segments it by sentence to obtain multiple raw audio segments, and sends each raw audio segment to the acceleration computing unit. The accelerated computing unit performs noise reduction processing on each original audio segment to obtain the corresponding enhanced audio segment. Then, it sequentially performs speech recognition, forced substitution correction, and faithful translation on each enhanced audio segment to obtain the corresponding translated speech. The unit then returns each original audio segment, the corresponding enhanced audio segment, and the translated speech to the host unit. The host unit plays each original audio segment and the corresponding enhanced audio segment on the local player. The native language transcription transmission link includes: The host unit acquires the native language text of its personnel and sends the native language text to the accelerated computing unit; The accelerated computing unit performs a standardized translation on the native language text to obtain a standardized translation, and then sends the standardized translation to the host unit; The host unit performs normalization processing on the numbers in the standard translation according to the target language rules to obtain normalized text, and then sends the normalized text to the accelerated computing unit. The accelerated computing unit performs speech synthesis on the read-out standardized text to obtain synthesized audio, and returns the synthesized audio to the host unit. The host unit sends a PTT enable command and a transmit path establishment command to the radio frequency control unit. The radio frequency control unit enables the radio communication terminal to enter the transmit state according to the PTT enable command, and connects the transmit path from the host unit's audio playback port to the transmit port of the radio communication terminal according to the transmit path establishment command. After the radio communication terminal is in the transmit state and the transmit path has been established, the host unit plays the synthesized audio, and after the playback ends, sends a PTT release command and a transmit path shutdown command to the radio frequency control unit. The radio frequency control unit enables the radio communication terminal to exit the transmit state according to the PTT release command, and disconnects the transmit path from the host unit's audio playback port to the transmit port of the radio communication terminal according to the transmit path shutdown command.

2. The method for interpreting and broadcasting and standardizing information transmission for maritime VHF communication according to claim 1, characterized in that, The preset listening frequency is 100Hz, the preset level threshold is a 12-bit ADC count of 100 to 200, the preset confirmation time is 80ms, and the preset silence time is 2.5s.

3. The method for interpreting and broadcasting and standardizing maritime VHF communications according to claim 1, characterized in that, During the sentence segmentation process, the host unit has a speech activity detection trigger threshold of 0.5, a minimum silence duration of 300ms, a speech pre- and post-speech filling duration of 200ms, and a sampling rate of 16kHz.

4. The method for interpreting and broadcasting and standardizing maritime VHF communications according to claim 1, characterized in that, The noise reduction process uses the DeepFilterNetv3 model.

5. The method for interpreting and broadcasting and standardizing maritime VHF communications according to claim 1, characterized in that, The speech recognition uses the faster-whisper-large-v3-turbo model.

6. The method for interpreting and broadcasting and standardizing maritime VHF communications according to claim 1, characterized in that, The forced replacement correction includes replacing terms in the text obtained from speech recognition with standardized spellings based on a glossary and following the principle of longest match priority.

7. The method for interpreting and broadcasting and standardizing information transmission for maritime VHF communication according to claim 1, characterized in that, The host unit acquires the native language text of its personnel, including: When the radio frequency control unit detects that the voice input button has been pressed, it reports a voice input start event to the host unit. After receiving the voice input start event, the host unit sends a native language acquisition channel establishment command to the radio frequency control unit. The radio frequency control unit connects the audio acquisition port of the host unit according to the native language acquisition channel establishment command. The host unit acquires the native language audio of its personnel through the channel. When the radio frequency control unit detects that the voice input button has been released, it reports a voice input end event to the host unit. After receiving the voice input end event, the host unit stops acquiring native language audio and sends the acquired native language audio to the accelerated computing unit. The accelerated computing unit sequentially performs noise reduction processing, speech recognition, and forced substitution correction on the native language audio to obtain native language text, and sends the native language text to the host unit. Alternatively, the host unit may receive native language text directly input by its own personnel.

8. The method for interpreting and broadcasting and standardizing maritime VHF communications according to claim 1, characterized in that, The standardization process for broadcasting includes: using NATO / ICAO reading for English, using standard radio reading for Chinese, converting numbers in the standardized translation of Japanese to kana reading, and using the corresponding standard radio communication reading for Korean.

9. A method for interpreting and broadcasting messages for maritime VHF communication, characterized in that, The listening and translation broadcasting method is applied to a listening and translation broadcasting and standard broadcasting system including a radio frequency control unit, a host unit and an acceleration computing unit. The listening and translation broadcasting method includes: executing the listening and translation broadcasting link. The audio-visualization and broadcasting link includes: The radio frequency control unit continuously samples the audio level of the radio communication terminal of the maritime VHF communication at a preset listening frequency. When the audio level is continuously higher than the preset level threshold and the duration reaches the preset confirmation time, it determines that the incoming call has started and reports the incoming call start event to the host unit. When the audio level is continuously lower than the preset level threshold and the duration reaches the preset silence time, it determines that the incoming call has ended and reports the incoming call end event to the host unit. After receiving a call start event, the host unit sends a call acquisition path establishment command to the radio frequency control unit. The radio frequency control unit connects the acquisition path from the radio communication terminal to the audio acquisition port of the host unit according to the call acquisition path establishment command. The host unit acquires the call audio received by the radio communication terminal through the acquisition path. After receiving a call end event, the host unit stops acquiring the call audio, performs voice activity detection on the acquired call audio and segments it by sentence to obtain multiple raw audio segments, and sends each raw audio segment to the acceleration computing unit. The accelerated computing unit performs noise reduction processing on each original audio segment to obtain the corresponding enhanced audio segment. Then, it sequentially performs speech recognition, forced substitution correction, and faithful translation on each enhanced audio segment to obtain the corresponding translated speech. The unit then returns each original audio segment, the corresponding enhanced audio segment, and the translated speech to the host unit. The host unit plays each original audio segment and the corresponding enhanced audio segment on the local player.

10. A standard broadcasting method for maritime VHF communication, characterized in that, The standardized broadcasting method is applied to a listening and translation broadcasting and standardized broadcasting system including a radio frequency control unit, a host unit, and an accelerated computing unit. The standardized broadcasting method includes: executing a native language transcription transmission link. The native language transcription transmission link includes: The host unit acquires the native language text of its personnel and sends the native language text to the accelerated computing unit; The accelerated computing unit performs a standardized translation on the native language text to obtain a standardized translation, and then sends the standardized translation to the host unit; The host unit performs normalization processing on the numbers in the standard translation according to the target language rules to obtain normalized text, and then sends the normalized text to the accelerated computing unit. The accelerated computing unit performs speech synthesis on the read-out standardized text to obtain synthesized audio, and returns the synthesized audio to the host unit. The host unit sends a PTT enable command and a transmit path establishment command to the radio frequency control unit. The radio frequency control unit enables the maritime VHF radio communication terminal to enter the transmit state according to the PTT enable command, and connects the transmit path from the host unit's audio playback port to the transmit port of the radio communication terminal according to the transmit path establishment command. After the radio communication terminal is in the transmit state and the transmit path has been established, the host unit plays the synthesized audio, and after the playback ends, sends a PTT release command and a transmit path shutdown command to the radio frequency control unit. The radio frequency control unit enables the radio communication terminal to exit the transmit state according to the PTT release command, and disconnects the transmit path from the host unit's audio playback port to the transmit port of the radio communication terminal according to the transmit path shutdown command.

Images