A voice reminding method and system based on nuclear power high-risk operation prompt

Abstract

The present application belongs to the technical field of artificial intelligence, and specifically relates to a voice reminding method and system based on nuclear power high-risk operation prompt, comprising the following steps: step 1, text preprocessing; step 2, feature extraction; step 3, acoustic modeling; and step 4, training and optimization.The present application has the beneficial effect of improving work efficiency and accuracy: through the voice intelligent reminding system, the staff does not need to rely on text notification or artificial voice broadcast, can directly hear the important step reminder of the high-risk operation, reduces the time and error of information transmission, avoids the possibility of text understanding or interpretation error, reduces the error of the staff when performing the high-risk operation, and improves the work efficiency.

Description

Technical Field

[0001] This invention belongs to the field of artificial intelligence technology, specifically relating to a voice reminder method and system based on high-risk nuclear power plant operations. Background Technology

[0002] In the nuclear power industry, the execution of high-risk operations is a complex and crucial process. To ensure the accurate and safe execution of work orders, providing effective reminder methods is essential. Currently, most power plant reminder systems rely primarily on text or audio notifications, requiring users to use electronic devices to view relevant information about high-risk operations. However, this method suffers from issues such as untimely information delivery and limited effectiveness. Summary of the Invention

[0003] The purpose of this invention is to provide a voice reminder method and system based on high-risk nuclear power operations. It achieves voice broadcast reminders for important steps in work orders through end-to-end voice synthesis, and has the characteristics of high efficiency, accuracy and safety. It can be widely used in the nuclear power industry and other high-risk operation fields to improve work efficiency and safety.

[0004] The technical solution of the present invention is as follows: A voice reminder method based on high-risk nuclear power plant operations, comprising the following steps:

[0005] Step 1: Text preprocessing;

[0006] Step 2: Feature extraction;

[0007] Step 3: Acoustic modeling;

[0008] Step 4: Training and optimization.

[0009] In step 1, the input work order text is preprocessed, including word segmentation and punctuation processing, to convert the text into a discrete sequence of words.

[0010] In step 2, the text sequence is encoded using a Transformer network structure, which includes multiple encoder layers. Each encoder layer contains a self-attention mechanism and a feedforward neural network. The self-attention mechanism models the relationship between words and captures global contextual information. The feedforward neural network is used to perform non-linear transformations on the features at each position. After processing by multiple encoder layers, the feature representation of each position is obtained.

[0011] In step 3, the feature representation output by the encoder is used as input, and acoustic modeling is performed by the decoder to generate a speech waveform. The decoder consists of multiple layers, each containing a self-attention mechanism, an encoder-decoder attention mechanism, and a feedforward neural network. The self-attention mechanism is used to model the context information inside the decoder, the encoder-decoder attention mechanism is used to associate the output of the encoder with the current position of the decoder, the feedforward neural network is used to perform nonlinear transformation on the features at each position, and finally, a linear layer maps the output of the decoder to a speech waveform.

[0012] In step 4, the speech synthesis model is trained using a training dataset with the target speech waveform. During the training process, the difference between the model output and the target speech waveform is calculated using the cross-entropy loss function, and the model parameters are updated using the backpropagation algorithm.

[0013] A voice reminder system based on high-risk nuclear power plant operations includes a model training module, a high-risk reminder setting module, a speech synthesis module, a speech broadcasting module, a reminder strategy module, and a user interaction module.

[0014] The model training module uses nuclear power plant work order text data and speech datasets. The data is derived from work orders in the nuclear power field, containing specific data and technical terms related to nuclear power. It trains a speech synthesis model based on a Transformer network structure using a training dataset with target speech waveforms. The training process includes the following:

[0015] Data preparation: Collect text data of nuclear power work orders, including historical work orders and standard operating procedure texts;

[0016] Voice dataset: Record or acquire voice data corresponding to the work order text;

[0017] Data preprocessing: The text data is segmented and punctuated, and the speech data is feature extracted;

[0018] Model training:

[0019] Model selection: The Transformer network structure, including encoder and decoder, is adopted;

[0020] Training process: The Transformer model is trained using a training dataset with the target speech waveform. The difference between the model output and the target speech waveform is calculated using the cross-entropy loss function, and the model parameters are updated using the backpropagation algorithm.

[0021] The high-risk alert setting module analyzes the work order data sent by the user through the terminal, receives and parses the work orders to be executed, and can accept input in text or phonetic character format, including:

[0022] Data reception and parsing:

[0023] User input: Provide a user interface or API interface for users to input work order data;

[0024] Data parsing: Parse the input work order data to extract key steps and important information.

[0025] The aforementioned speech synthesis module employs a Transformer network neural network model to convert important steps or phonetic characters in the work order into speech. The specific neural network model includes an encoder-decoder architecture, where the encoder converts the text sequence into an intermediate representation, and the decoder generates the corresponding speech waveform based on the intermediate representation. An attention mechanism is used between the encoder and decoder to enhance the focus on input information and improve the accuracy of the generated speech. The neural network model architecture uses a Transformer network structure and is trained using a cross-entropy loss function and optimization algorithms.

[0026] The voice broadcast module generates voice messages and plays them through an audio playback device or a voice presentation device. The voice broadcast module supports different audio output methods. The reminder strategy module controls the timing and frequency of voice broadcasts based on the importance, urgency, and execution sequence of the work order to ensure the accuracy and timeliness of the reminder information.

[0027] The beneficial effects of this invention are as follows: improved work efficiency and accuracy: through the voice intelligent reminder system, staff do not need to rely on text notifications or manual voice broadcasts, and can directly hear reminders of important steps in high-risk operations, reducing the time and error of information transmission, avoiding the possibility of misunderstanding or interpreting text, reducing errors made by staff when performing high-risk operations, and improving work efficiency.

[0028] Enhancing the safety of nuclear power plant operations: Through the intelligent voice reminder system, reminders for important steps can be promptly conveyed to staff, reducing potential safety hazards in high-risk operations and improving the safety of the work site.

[0029] Application of end-to-end speech synthesis technology: By using end-to-end speech synthesis technology combined with specific neural network models, direct conversion from text to speech is achieved, eliminating intermediate steps in traditional speech synthesis and improving the naturalness and accuracy of synthesized speech.

[0030] Modular system design: The system of this invention adopts a modular design, which not only makes the system structure clear and the function of each module clear, but also facilitates the expansion and maintenance of the system, and improves the system's flexibility and scalability. Attached Figure Description

[0031] Figure 1 A flowchart of a voice reminder method based on high-risk nuclear power plant operations provided by the present invention;

[0032] Figure 2 This is a schematic diagram of a voice reminder system based on high-risk nuclear power plant operations provided by the present invention. Detailed Implementation

[0033] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0034] The present invention provides a voice reminder method and system based on high-risk nuclear power plant operation prompts. When an AWT work order is started, it provides accurate, timely and efficient voice prompts for high-risk operations, ensuring the stability of high-risk nuclear power plant operations and improving safety production efficiency.

[0035] This invention relates to the field of natural language processing, and provides a voice reminder method and system for high-risk nuclear power plant operations. The solution is based on a nuclear power plant workflow management information system, and the invention uses a Transformer network structure. Through the application and practice of this method and system, the implementation efficiency of high-risk nuclear power plant operations is improved, solving the problem that traditional methods involve excessive user viewing of prompts and subsequent processing, failing to meet practical needs.

[0036] Compared to traditional text or voice notifications, this method achieves accuracy and timeliness in information delivery through voice prompts. Workers no longer need to rely on electronic devices to view relevant information; they can receive crucial step-by-step instructions for high-risk operations simply through voice broadcasts, thus improving work efficiency.

[0037] By employing the Transformer network structure and natural language processing techniques, this method can better model the relationships between texts and capture global contextual information, thereby improving the quality and accuracy of speech synthesis.

[0038] Through the reminder strategy module, this method can intelligently control the timing and frequency of voice broadcasts based on the characteristics of the work order and the sequence of execution steps, ensuring the accuracy and timeliness of reminder information and reducing unnecessary interference.

[0039] The user interaction module offers personalized settings options, allowing users to customize it according to their needs and preferences, thereby improving the user experience.

[0040] The reason for choosing the Transformer network structure in this method is that Transformer has achieved significant results in natural language processing tasks and has the following advantages:

[0041] 1. Capturing Long-Distance Dependencies: Traditional recurrent neural networks (RNNs) are prone to gradient vanishing or exploding problems when processing long sequences, making it difficult to capture long-distance dependencies. Transformers, however, can directly model the relationships between words through self-attention, without relying on a fixed-length window, thus better capturing long-distance dependencies.

[0042] 2. Strong Parallel Computing Capability: Due to the temporal dependency of RNNs, they need to be computed sequentially, making parallel computation difficult and limiting their efficiency on large-scale data. In contrast, the self-attention mechanism in Transformers allows for parallel computation, resulting in faster model training and the ability to handle larger datasets.

[0043] 3. Global Contextual Information: Traditional RNN models process words one by one, with each word only seeing the preceding context information. In contrast, the self-attention mechanism in Transformer models can model all words at once, capturing global contextual information and contributing to the generation of more accurate and fluent speech.

[0044] 4. Parameter Sharing and Efficiency: Traditional RNN models require storing and computing a large number of intermediate states when processing long sequences, resulting in numerous model parameters and low computational efficiency. Transformer, through its self-attention mechanism, achieves global information exchange, avoiding state propagation common in RNNs, resulting in better parameter sharing and higher computational efficiency.

[0045] like Figure 1 As shown, a voice reminder method based on high-risk nuclear power plant operations includes the following steps:

[0046] Step 1: Text Preprocessing

[0047] The input work order text is preprocessed, including word segmentation and punctuation processing, to transform the text into a discrete sequence of words.

[0048] Word segmentation: The input work order text is segmented into word sequences. For example, the input text "Please check reactor pressure" is segmented into "Please", "Check", "Reactor", and "Pressure".

[0049] Punctuation processing: This feature handles punctuation marks in text, allowing you to choose whether to remove or retain them, depending on your specific needs. For example, you can remove periods and commas to simplify subsequent processing.

[0050] Word vector representation: Converting the segmented sequence of words into word vectors. Pre-trained word embedding models (such as Word2Vec, GloVe, or BERT) can be used to represent each word as a fixed-dimensional vector. For example, the word "check" might be represented as a 300-dimensional vector.

[0051] Step 2: Feature Extraction

[0052] The text sequence is encoded using a Transformer network structure, which includes multiple encoder layers, each containing a self-attention mechanism and a feedforward neural network. The self-attention mechanism models the relationships between words and captures global contextual information. The feedforward neural network performs a non-linear transformation on the features at each position. After processing by multiple encoder layers, a feature representation for each position is obtained.

[0053] In the feature extraction step, the text sequence is encoded using a Transformer network structure. The Transformer network consists of multiple encoder layers, each containing a self-attention mechanism and a feedforward neural network.

[0054] Self-attention mechanisms are used to calculate the importance of each word in the sequence, capturing global contextual information. The formula is as follows:

[0055]

[0056] Where Q (query matrix), K (key matrix), and V (value matrix) are obtained from the input word vectors through linear transformation. k It is a scaling factor, usually equal to the dimension of the key vector.

[0057] Feedforward neural networks are used to perform non-linear transformations on the features at each location. The formula is as follows:

[0058] FFN(x) = max(0, xW1+b1)W2+b2

[0059] Where: W1 and W2 are weight matrices. b1 and b2 are bias vectors. max(0,x) is the ReLU activation function. The calculation steps are as follows:

[0060] The first layer of linear transformation and ReLU activation: After the input feature vector x undergoes a linear transformation xW1+b1, the ReLU activation function is applied.

[0061] The second linear transformation: the activated output undergoes another linear transformation (max(0, xW1+b1))W2+b2.

[0062] After processing through multiple encoder layers, feature representations are obtained for each position. These feature representations contain relationships between words and contextual information.

[0063] Step 3: Acoustic Modeling

[0064] The feature representation output from the encoder is used as input, and the decoder performs acoustic modeling to generate a speech waveform. The decoder also consists of multiple layers, each containing a self-attention mechanism, an encoder-decoder attention mechanism, and a feedforward neural network. The self-attention mechanism models the contextual information within the decoder. The encoder-decoder attention mechanism associates the encoder's output with the decoder's current position. The feedforward neural network performs a non-linear transformation on the features at each position. Finally, a linear layer maps the decoder's output to a speech waveform.

[0065] The acoustic modeling part uses a decoder to convert the feature representation output by the encoder into a speech waveform. The decoder also consists of multiple layers, each containing a self-attention mechanism, an encoder-decoder attention mechanism, and a feedforward neural network.

[0066] The encoder-decoder attention mechanism is used to associate the encoder's output with the decoder's current position. The formula is as follows:

[0067]

[0068] Here, Q is the query matrix input to the decoder, and K and V are the key matrix and value matrix output by the encoder.

[0069] The feedforward neural network in the decoder, similar to that in the encoder, performs a non-linear transformation on the features at each location. Finally, a linear layer maps the decoder's output to a speech waveform.

[0070] Output = Linear(x)

[0071] Here, x is the output feature of the decoder, which is converted into a speech signal by the linear layer.

[0072] Specifically as follows:

[0073] Linear Transformation: The feature vector x output by the decoder undergoes a linear transformation y = xW + b.

[0074] Where W is the weight matrix and b is the bias vector.

[0075] Generate speech waveforms: Use the output of the linear layer as speech waveform data.

[0076] The speech synthesis method based on Transformer networks utilizes an encoder-decoder structure to effectively convert text input into speech output. By combining self-attention mechanism and feedforward neural network, it captures global contextual information of the text and generates high-quality speech prompts.

[0077] Step 4: Training and Optimization

[0078] The speech synthesis model is trained using a training dataset containing the target speech waveform. During training, the difference between the model output and the target speech waveform is calculated using the cross-entropy loss function, and the model parameters are updated using the backpropagation algorithm.

[0079] The self-attention mechanism is expressed as follows:

[0080]

[0081] Where Q is the query vector, K is the key vector, V is the value vector, and d k It is the vector dimension.

[0082] The encoder-decoder attention mechanism is represented as:

[0083] MultiHead(Q,K,V)=Concat(h1,...,hn)W O

[0084] hi=Attention(QWQi,KWKi,VWVi)

[0085] Where h is the number of heads, W Qi W Ki W Vi W O It is a parameter matrix.

[0086] The objective function uses the cross-entropy loss function to measure the difference between the model output and the target speech waveform.

[0087] LOSS=-i∑tilog(pi)

[0088] Where ti is the i-th value of the target speech waveform, and pi is the i-th value of the model output.

[0089] like Figure 2 As shown, a voice reminder system based on high-risk nuclear power plant operations is based on artificial intelligence and natural language processing technology, and includes: a model training module, a high-risk reminder setting module, a speech synthesis module, a speech broadcasting module, a reminder strategy module, and a user interaction module.

[0090] Model training module: The training data for nuclear power work order text data and speech datasets comes from work orders in the nuclear power field, containing nuclear power-specific data and proper nouns, and has nuclear power characteristics. The speech synthesis model based on the Transformer network structure is trained using a training dataset with target speech waveforms.

[0091] A speech synthesis model based on a Transformer network structure was trained using text and speech datasets from nuclear power plant work orders, along with a training dataset containing target speech waveforms. The main technical methods are as follows:

[0092] 1. Data preparation:

[0093] Nuclear power plant work order text data: Collect a large amount of text data of nuclear power plant work orders, including historical work orders and standard operating procedure texts.

[0094] Voice dataset: Record or acquire voice data corresponding to the work order text to ensure that the voice data accurately reflects the text content.

[0095] Data preprocessing: The text data is segmented and punctuated, and the speech data is subjected to feature extraction, such as MFCC (Mel frequency cepstral coefficient) feature extraction.

[0096] 2. Model Training:

[0097] Model selection: The Transformer network structure is adopted, which includes two main parts: encoder and decoder.

[0098] Training process: The Transformer model is trained using a training dataset containing the target speech waveform. The difference between the model output and the target speech waveform is calculated using the cross-entropy loss function, and the model parameters are updated using the backpropagation algorithm.

[0099] Optimization algorithms: Commonly used optimization algorithms, such as Adam, are used to accelerate model convergence.

[0100] High-risk alert setting module: By analyzing the work order data sent by the user through the terminal, it receives and parses the work orders to be executed. The work orders can be entered in text form or phonetic character form.

[0101] The system analyzes work order data sent by users through terminals, receives and parses work orders to be executed, and accepts input in text or phonetic character format. Key technical methods include:

[0102] Data reception and parsing:

[0103] User input: Provide a user interface (GUI) or API interface for users to input work order data.

[0104] Data parsing: Parse the input work order data, extract key steps and important information, and use natural language processing techniques such as named entity recognition (NER) and dependency parsing.

[0105] Speech Synthesis Module: Utilizing end-to-end speech synthesis technology, this module employs a Transformer network neural network model to convert key steps or phonetic characters in work orders into natural and fluent speech. The specific neural network model includes an encoder-decoder architecture, where the encoder converts the text sequence into an intermediate representation, and the decoder generates the corresponding speech waveform based on this intermediate representation. An attention mechanism can be used between the encoder and decoder to enhance the focus on input information and improve the accuracy of the generated speech. The neural network model architecture uses a Transformer network structure, combined with a cross-entropy loss function and optimization algorithms for training.

[0106] The main technical means include:

[0107] 1. Text encoding:

[0108] Encoder: The preprocessed text sequence is input into the encoder to generate an intermediate representation. The encoder contains multiple layers of self-attention mechanisms and feedforward neural networks, with each layer extracting features from the input sequence.

[0109] 2. Speech generation:

[0110] Decoder: The decoder receives the intermediate representation output by the encoder and combines the self-attention mechanism and the encoder-decoder attention mechanism to generate the corresponding speech waveform.

[0111] Linear layer: The output of the decoder is mapped to a speech waveform through a linear layer.

[0112] 3. Model training:

[0113] Loss function: The cross-entropy loss function is used to measure the difference between the generated speech and the target speech.

[0114] Optimization process: The model parameters are optimized using gradient descent and backpropagation algorithms.

[0115] Voice broadcast module: This module plays the generated voice messages through audio playback or voice presentation devices so that staff can hear important reminders about high-risk operations. The voice broadcast module supports different audio output methods, such as speakers and headphones. Key technical means include:

[0116] Audio devices: Supports multiple audio output methods, such as speakers, headphones, etc.

[0117] Playback control: Implement audio playback control functions, such as play, pause, stop, etc., through audio playback libraries (such as PyAudio, ALSA, etc.).

[0118] Reminder Strategy Module: Based on the importance, urgency, and execution sequence of work orders, a reminder strategy module is designed to control the timing and frequency of voice announcements, ensuring the accuracy and timeliness of reminder information. Key technical methods include:

[0119] 1. Strategy Design:

[0120] Priority settings: Define the importance and urgency of different types of work orders, and set the priority of reminders.

[0121] Time control: Control the timing and frequency of voice broadcasts based on the sequence of execution steps and the actual operation.

[0122] 2. Scheduling Algorithm: A priority queue scheduling algorithm is used to manage reminder tasks, ensuring that high-priority tasks are broadcast in a timely manner.

[0123] User interaction module: Customizable based on user preferences, allowing for settings such as volume level, whether to enable / disable, and selection of male or female voice prompts. Key technical methods include:

[0124] 1. User Settings:

[0125] Settings interface: Provides a user interface for users to personalize settings, including volume, voice gender selection, enabling or disabling reminders, etc.

[0126] 2. Configure storage:

[0127] Storage method: User settings can be saved in local configuration files or cloud databases for the system to read and apply at runtime.

[0128] Data interaction between system modules: The preprocessing module converts the work order text into a word sequence and passes it to the feature extraction module. The intermediate representation generated by the feature extraction module is passed to the acoustic modeling module to generate a speech waveform. The speech waveform generated by the acoustic modeling module is passed to the speech playback module for actual playback. The model training module provides a trained model for the speech synthesis module to use. The work order data parsed by the high-risk warning setting module is passed to the speech synthesis module. The reminder strategy module controls the timing and frequency of the speech playback module. The settings parameters of the user interaction module affect the behavior of the reminder strategy module and the speech playback module.

[0129] Example:

[0130] A voice reminder method based on high-risk nuclear power plant operations includes the following steps:

[0131] Step 001: Model Training. This step includes: Text Preprocessing: Preprocessing the input work order text, including word segmentation and punctuation processing, converting the text into a discrete word sequence. Feature Extraction, Acoustic Modeling, Training, and Optimization. The specific neural network model adopts an encoder-decoder architecture, where the encoder converts the text sequence into an intermediate representation, and the decoder generates the corresponding speech waveform based on the intermediate representation. To enhance the attention to input information and the accuracy of generation, the model incorporates an attention mechanism. The system uses the cross-entropy loss function and optimization algorithms to train the neural network model to improve the quality and accuracy of speech synthesis.

[0132] Step 002: Work Order Data Analysis. In this step, the system receives and analyzes work order data sent from user terminals. The system can parse work orders to be executed, which can be entered in text or phonetic character format. Through analysis of the work order data, the system can extract key step information required to execute high-risk operations.

[0133] Step 003: Reminder Strategy, comprising two parts: 1. Analyzing whether the work order's data format meets the requirements. If not, the user is reminded that the work order's data format does not meet the requirements. 2. Based on the work order's importance, urgency, and the sequence of execution steps, the system designs a reminder strategy module to control the timing and frequency of voice broadcasts, ensuring the accuracy and timeliness of reminder information. Through flexible reminder strategies, the system can provide effective voice prompts for high-risk operations based on actual conditions, improving the efficiency and safety of operation execution.

[0134] Step 004: User Interaction. The system provides a user interaction module, allowing users to customize reminder methods based on their preferences and settings. Users can adjust the volume, enable or disable the voice reminder function, and select parameters such as male or female voice versions to meet their individual needs and preferences.

[0135] Step 005: Speech Synthesis. In the speech synthesis module, the system employs end-to-end speech synthesis technology, using a neural network model with a Transformer network structure. This model converts important step text or phonetic characters in the work order into natural and fluent speech.

[0136] Step 006: Voice broadcast. The generated voice message is played through an audio playback device or a voice presentation device so that staff can hear the important steps of the high-risk operation. The voice broadcast module supports different audio output methods, such as speakers and headphones, to adapt to different usage scenarios and needs.

Claims

1. A voice prompting method based on nuclear power high-risk operation prompt, characterized in that, Comprising the following steps: Step 1: text preprocessing; The step 1 in the input of the work order text preprocessing, including word segmentation, punctuation processing, the text is converted into discrete word sequence; Step 2: feature extraction; The step 2 in the use of the Transformer network structure to encode the text sequence, the Transformer network structure includes multiple encoder layers, each encoder layer contains self-attention mechanism and feedforward neural network, self-attention mechanism modeling the relationship between words and capturing global context information, feedforward neural network for nonlinear conversion of the features of each position, after the processing of multiple encoder layers, the feature representation of each position is obtained; Step 3: acoustic modeling; The step 3 in the feature representation output by the encoder as input, acoustic modeling is carried out through the decoder to generate speech waveform, the decoder is composed of multiple layers, each layer contains self-attention mechanism, encoder-decoder attention mechanism and feedforward neural network, self-attention mechanism is used to model the context information within the decoder, encoder-decoder attention mechanism is used to associate the output of the encoder with the current position of the decoder, feedforward neural network is used to nonlinearly transform the features of each position, finally, the output of the decoder is mapped to the speech waveform through a linear layer; Step 4: training and optimization; The step 4 in the use of training data set with target speech waveform to train the speech synthesis model, during the training process, the difference between the model output and the target speech waveform is calculated using cross entropy loss function, and the parameters of the model are updated through back propagation algorithm; Step 5: work order data analysis, extract the important step information required for the execution of high-risk operation, and remind, including 1, analyze whether the data format of the work order meets the requirements, if it does not meet the requirements, remind the user that the data format of the work order does not meet the prompt requirements; 2, according to the importance, urgency and execution step of the work order, remind the occasion and frequency of voice broadcast, and then perform voice broadcast.

2. A voice reminder system based on the method of claim 1, characterized by: Including model training module, high risk prompt setting module, speech synthesis module, voice broadcast module, reminder strategy module and user interaction module.

3. The voice reminder system based on nuclear power high-risk operation prompt according to claim 2, characterized in that: The model training module trains the speech synthesis model based on the Transformer network structure by training the nuclear power work order text data and the speech data set, which is derived from the work order in the nuclear power field, contains nuclear power specific data and proper nouns, and uses the training data set with target speech waveform to train the speech synthesis model based on the Transformer network structure, which includes the following: Data preparation: collect the text data of the nuclear power work order, including historical work order and standard operation procedure text; Speech data set: record or obtain the speech data corresponding to the work order text; Data preprocessing: word segmentation, punctuation processing on text data, and feature extraction on speech data; Model training: Model selection: Adopting the Transformer network structure, including encoder and decoder; Training process: Using the training dataset with target speech waveforms, training the Transformer model, using the cross-entropy loss function to calculate the difference between the model output and the target speech waveform, and updating the model parameters through the back propagation algorithm.

4. The voice reminder system based on nuclear power high-risk operation prompt according to claim 2, characterized in that: The high-risk prompt setting module analyzes the work order data sent by the user through the terminal, receives and parses the to-be-executed work order, which can be input in text form or phonetic character form, including: Data reception and analysis: User input: Provide a user interface or API interface for users to input work order data; Data analysis: Analyze the input work order data and extract key steps and important information.

5. The voice reminder system based on nuclear power high-risk operation prompt according to claim 2, characterized in that: The speech synthesis module adopts a Transformer network structure neural network model to convert important step text or phonetic characters in the work order into speech. The specific neural network model includes an encoder-decoder architecture, where the encoder converts text sequences into intermediate representations, and the decoder generates corresponding speech waveforms based on the intermediate representations. The attention mechanism is used between the encoder and the decoder to enhance the attention to the input information and the accuracy of the generated speech. The neural network model architecture adopts the Transformer network structure and is trained using the cross-entropy loss function and optimization algorithm.

6. The voice reminder system based on nuclear power high-risk operation prompt according to claim 2, characterized in that: The voice broadcast module plays the generated voice through audio playback devices or voice presentation devices. The voice broadcast module supports different audio output methods. The reminder strategy module controls the timing and frequency of voice broadcast based on the importance, urgency, and execution order of the work order to ensure the accuracy and timeliness of the reminder information.

Images