Intelligent nursing medical record voice recording system

Abstract

The invention belongs to the technical field of medical informatization, and provides an intelligent nursing medical record voice recording system, which comprises a voice acquisition module, a voice processing module and a voice recording module, and is characterized in that the voice acquisition module acquires audio signals in real time through preset wake-up words; the voice recognition module converts the collected audio signals into text information; the timestamp generation module generates a timestamp and associates the timestamp with the text information to obtain associated data; the automatic recording module is used for extracting key information according to the associated data and automatically recording medication and treatment measure processes to obtain recorded data; the medical record generation module integrates the recorded data according to the standard format of the nursing medical record to generate a structured nursing medical record; through voice input and automatic processing functions, the time for manually inputting medical records by nursing personnel is greatly reduced, the risk of manual input errors and omission is reduced, the accuracy and integrity of medical record contents are ensured, and the medical service process is optimized.

Description

Technical Field

[0001] This invention belongs to the field of medical information technology, specifically an intelligent voice input nursing medical record system. Background Art

[0002] In today's healthcare industry, medical records are a crucial part of medical services. Traditional methods of entering nursing medical records rely primarily on manual input by medical staff, a process that is not only time-consuming but also prone to errors in high-pressure, fast-paced work environments. With the rapid development of medical knowledge and the increasing complexity of medical record content, healthcare workers need to maintain a high level of concentration during the data entry process to prevent the omission of critical information.

[0003] However, the traditional manual medical record entry process frequently consumes nurses' time and energy, preventing them from focusing on the actual care of patients. Nurses often have to squeeze out fragmented time to complete medical record entries amidst their busy work schedules, which can cause delays in updating medical records and consequently affect patients' subsequent treatment and care.

[0004] Furthermore, manual data entry is prone to errors and omissions. Due to the fast pace and large volume of information in nursing work, nurses may experience inaccurate or incomplete records due to fatigue, negligence, or other factors. This can not only mislead subsequent medical diagnoses and treatments, increasing medical risks, but also potentially affect doctor-patient communication and trust. According to relevant statistics, the error rate of manually entered medical records is as high as 10%-15%, including issues such as incorrect or omitted information and non-standard formatting, which greatly reduces the usability and reference value of medical records.

[0005] Currently, most hospitals still rely on paper documents and fragmented data systems for medical record keeping and management, which makes information retrieval difficult. Doctors spend a significant amount of time manually searching through medical records, especially when the patient's records are complex, making information retrieval inefficient. Furthermore, the unstructured format of medical record content makes systematic management and data analysis difficult, which to some extent affects the scientific rigor of medical decisions. Although some computerized electronic medical record systems have emerged in recent years, most existing solutions have failed to effectively integrate voice recognition technology, forcing healthcare workers to still invest considerable effort in tedious information input. In addition, these systems are often limited in functionality and lack intelligence, failing to automatically extract and organize information, further restricting workflow optimization.

[0006] Therefore, those skilled in the art have proposed an intelligent voice-input nursing medical record system, which aims to improve the quality of medical services and optimize medical processes by introducing voice recognition technology, automatic information extraction, and structured medical record generation. Summary of the Invention

[0007] To address the aforementioned technical problems, this invention provides an intelligent voice-input nursing medical record system to solve the issues raised in the background section.

[0008] An intelligent voice-based nursing medical record system includes:

[0009] The voice acquisition module is used to acquire audio signals in real time using a preset wake-up word;

[0010] A speech recognition module is used to convert the acquired audio signals into text information;

[0011] The timestamp generation module is used to generate timestamps and associate them with the text information to obtain associated data;

[0012] An automatic recording module is used to extract key information based on the associated data, automatically record the medication and treatment process, and obtain recorded data.

[0013] The medical record generation module is used to integrate the recorded data according to the standardized format of nursing medical records and generate structured nursing medical records.

[0014] Preferably, the voice acquisition module is further used for:

[0015] By continuously collecting ambient audio signals through a microphone device, a continuous audio signal x(t) that varies with time is obtained, where t is time;

[0016] The continuous audio signal x(t) is divided into multiple short frames, each with a duration of 10-30 milliseconds and a length of N sampling points. The audio signal of the nth frame is represented as x. n [m], where m = 0, 1, ..., N-1;

[0017] From each frame of audio signal x n Extracting the feature vector from [m], we obtain the feature vector f of the nth frame. n ;

[0018] The feature vector f n The feature template T of the preset wake word is compared with the Euclidean distance d(f) and the distance is calculated. n ,T):

[0019]

[0020] Where D is the dimension of the feature vector, f n,i T is the i-th element of the feature vector of the n-th frame. i This is the i-th element of the preset wake word feature template;

[0021] When the Euclidean distance d(f)n When T is less than the preset wake-up threshold θ, a wake-up word is detected.

[0022] When a wake word is detected, voice signals are collected in real time until a stop condition is met; the stop condition is no voice input within 2 minutes.

[0023] Preferably, the speech recognition module is further used for:

[0024] Using Mel-frequency cepstral coefficients (MFCC), from the audio signal x n Features are extracted from [m] to obtain feature F. n [e], where e is a time frame;

[0025] The feature F is represented by acoustic model A. n [e] mapped to a phoneme or acoustic unit is represented as:

[0026] P(x|F n [e])=A(F n [e])

[0027] Wherein, P(x|F n [e]) represents the probability of generating a specified phoneme given the features;

[0028] By constraining the output of the acoustic model A using a language model, we obtain:

[0029]

[0030] Where w is the word sequence, w i Let w be the i-th word in the word sequence, N be the number of words in the sequence, and P(w) be the probability of the entire word sequence w.

[0031] By combining acoustic and language models, the optimal word sequence W is obtained. * To maximize the probability word sequence:

[0032]

[0033] Among them, P(W|F n [e]) is a given feature F n The probability of the word sequence W under the condition [e], P(W) is the prior probability of the word sequence W, P(F) is the probability of the word sequence W under the condition [e]. n [e]|W) is to generate feature F given the word sequence W. n The conditional probability of [e];

[0034] By combining feature extraction, acoustic models, and language models, the input audio signal x is processed. n [m] is converted to the corresponding text information W n .

[0035] Preferably, the timestamp generation module is further used for:

[0036] The timestamp is calculated based on the number of milliseconds that have elapsed since the Unix epoch, which is a fixed starting point. The generated timestamp TS is represented as follows:

[0037] TS=T now -T0

[0038] Where T0 is the time point of the Unix epoch, T now This refers to the current time point;

[0039] The timestamp generation module is also used for:

[0040] The generated timestamp is associated with the text information obtained by the speech recognition module;

[0041] By generating a timestamp immediately after speech recognition is completed, and combining the timestamp and text information into a data structure, the associated data is obtained:

[0042] AD = (TS, W) n )

[0043] In this context, AD represents associated data.

[0044] Preferably, the automatic recording module is further used for:

[0045] Regarding the text information W n Preprocessing is performed to obtain the preprocessed text information W. n ′;

[0046] The preprocessed text information W n Extract key information from the sample, which includes drug information and treatment information, and is represented as: (Dr,Tr)=M(W) n ′), where Dr is drug information, Tr is treatment information, and M is the extraction model;

[0047] The extracted key information and the timestamp are structured to obtain the recorded data:

[0048] RD = (TS, Dr, Tr)

[0049] Here, RD represents recorded data.

[0050] Preferably, the medical record generation module is further used for:

[0051] The standardized format of nursing medical records is abstracted into a nursing medical record template Te, which is used to define the names and arrangement order of each part of the medical record;

[0052] Mapping each piece of information in the recorded data RD to its corresponding position in the standardized nursing medical record format is represented as follows:

[0053] MD = Map(RD, Te)

[0054] Where MD is the mapped data, and the structure of MD matches the standard format of nursing medical records; Map is the mapping function.

[0055] The mapped data MD is then populated into the nursing record template Te to generate a complete structured nursing record MR.

[0056]

[0057] in, For data population operations, fill in the information in MD into the corresponding positions according to the template requirements.

[0058] Compared with the prior art, the present invention has the following beneficial effects:

[0059] 1. This invention greatly reduces the time nursing staff spend manually entering medical records through voice input and automatic processing functions, allowing them to devote more time to patient care and improving the overall efficiency of medical services.

[0060] 2. This invention reduces the risk of errors and omissions in manual data entry by using voice recognition and automatic extraction of key information, ensuring the accuracy and completeness of medical record content, and helping to improve the accuracy of medical diagnosis and the rationality of treatment plans.

[0061] 3. This invention makes medical record management more standardized and convenient through structured medical record generation and the application of timestamps, enabling medical staff to quickly find and review patient nursing information, optimizing medical service processes, and improving the collaborative efficiency of medical teams. Attached Figure Description

[0062] Figure 1 This is a block diagram of the intelligent voice input nursing medical record system of the present invention. DETAILED DESCRIPTION

[0063] The embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and should not be construed as limiting the scope of the invention.

[0064] As attached Figure 1 As shown:

[0065] Example: This invention provides an intelligent voice-input nursing medical record system, comprising:

[0066] The voice acquisition module is used to acquire audio signals in real time by means of a preset wake-up word; it continuously collects ambient audio signals through a microphone device to obtain a continuous audio signal x(t) that changes over time, where t is time.

[0067] The continuous audio signal x(t) is divided into multiple short frames, each with a duration of 10-30 milliseconds and a length of N sampling points. The audio signal of the nth frame is represented as x. n [m], where m = 0, 1, ..., N-1;

[0068] From each frame of audio signal x n Extracting the feature vector from [m], we obtain the feature vector f of the nth frame. n ;

[0069] The feature vector f n The feature template T of the preset wake word is compared with the Euclidean distance d(f) and the distance is calculated. n ,T):

[0070]

[0071] Where D is the dimension of the feature vector, f n,i T is the i-th element of the feature vector of the n-th frame. i This is the i-th element of the preset wake word feature template;

[0072] When the Euclidean distance d(f) n When T is less than the preset wake-up threshold θ, a wake-up word is detected.

[0073] When a wake word is detected, voice signals are collected in real time until a stop condition is met; the stop condition is no voice input within 2 minutes.

[0074] By using preset wake words, the voice input of doctors or nurses can be collected in real time, accurately capturing the voice information of nursing staff and ensuring that no important nursing records are missed. The preset wake words can also avoid mis-collection, improve the accuracy and efficiency of data collection, increase the efficiency of medical record entry, reduce manual input time, and make it convenient for medical staff to record information in busy environments.

[0075] The speech recognition module converts the acquired audio signal into text information; it uses Mel-frequency cepstral coefficients (MFCC) to convert the audio signal x... n Features are extracted from [m] to obtain feature F. n [e], where e is a time frame;

[0076] Using acoustic model A to extract features F n [e] mapped to a phoneme or acoustic unit is represented as:

[0077] P(x|F n [e])=A(F n [e])

[0078] Wherein, P(x|F n [e]) represents the probability of generating a specified phoneme given the features;

[0079] By constraining the output of acoustic model A using the language model, we obtain:

[0080]

[0081] Where w is the word sequence, w i Let w be the i-th word in the word sequence, N be the number of words in the sequence, and P(w) be the probability of the entire word sequence w.

[0082] By combining acoustic and language models, the optimal word sequence W is obtained. * To maximize the probability word sequence:

[0083]

[0084] Among them, P(W|F n [e]) is a given feature F n The probability of the word sequence W under the condition [e], P(W) is the prior probability of the word sequence W, P(F) is the probability of the word sequence W under the condition [e]. n [e]|W) is to generate feature F given the word sequence W. n The conditional probability of [e];

[0085] By combining feature extraction, acoustic models, and language models, the input audio signal x is processed. n [m] is converted to the corresponding text information W n .

[0086] The collected speech signals are converted into text information, realizing the conversion from speech to text. By quickly and accurately converting speech into editable text, the speed of medical record entry is greatly improved, the time and workload of manual typing are reduced, and the error rate that may be caused by manual entry is also reduced.

[0087] The timestamp generation module is used to generate timestamps, associate them with text information, and obtain associated data. The timestamp is calculated based on the number of milliseconds that have elapsed since the Unix epoch, which is a fixed starting point. The generated timestamp TS is represented as follows:

[0088] TS=T now -T0

[0089] Where T0 is the time point of the Unix epoch, T now This refers to the current time point;

[0090] The generated timestamp is associated with the text information obtained by the speech recognition module; by generating a timestamp immediately after speech recognition is completed, and combining the timestamp and text information into a data structure, the associated data is obtained:

[0091] AD = (TS, W) n )

[0092] AD stands for Associated Data. By generating timestamps and associating them with text information, a time identifier is added to each record, giving nursing records a clear chronological order. This facilitates medical staff in viewing and tracing patients' nursing conditions at different points in time, providing accurate time references for medical decisions and helping to improve the quality and safety of medical services.

[0093] The automatic recording module is used to extract key information based on associated data, automatically record the process of medication and treatment measures, and obtain the recorded data.

[0094] For text information W n Preprocessing is performed to obtain the preprocessed text information W. n ′;

[0095] Preprocessed text information W n Extract key information from the sample. Key information includes drug information and treatment information. Drug information includes drug name, dosage, and frequency. Treatment information includes treatment method and timing. The extracted key information is represented as follows:

[0096] (Dr,Tr)=M(W n ′)

[0097] Where Dr represents drug information, Tr represents treatment information, and M represents the extraction model;

[0098] The extracted key information and timestamps are then structured to obtain the recorded data:

[0099] RD = (TS, Dr, Tr)

[0100] Here, RD represents recorded data. By extracting key information from associated data, the system automatically records medication and treatment procedures. It can automatically extract crucial medication and treatment information from large amounts of text data, avoiding omissions that may occur during manual searching and recording, ensuring the completeness and accuracy of nursing records, and also saving nursing staff time in organizing records.

[0101] The medical record generation module is used to integrate recorded data according to the standardized format of nursing medical records and generate structured nursing medical records; it abstracts the standardized format of nursing medical records into a nursing medical record template Te, which is used to define the names and arrangement order of each part of the medical record;

[0102] Mapping each piece of information in the recorded data RD to its corresponding location in the standardized nursing medical record format is represented as follows:

[0103] MD = Map(RD, Te)

[0104] Where MD is the mapped data, and the structure of MD matches the standard format of nursing medical records; Map is the mapping function.

[0105] The mapped data MD is then populated into the nursing record template Te to generate a complete structured nursing record MR.

[0106]

[0107] in, For data population operations, fill in the information in MD into the corresponding positions according to the template requirements.

[0108] Based on the standardized format of nursing medical records, the recorded data is integrated to generate structured nursing medical records, ensuring that the generated nursing medical records comply with industry norms and standards, have a unified format and clear structure, facilitate medical staff to view, manage and share, improve the readability and usability of medical records, and are conducive to the effective transmission of medical information and the control of medical quality.

[0109] As can be seen from the above, compared with the traditional manual entry of medical records, this system uses voice acquisition and recognition technology to achieve rapid entry of medical records, freeing up the hands of nursing staff and improving work efficiency. It is especially suitable for scenarios where nursing work is busy and requires rapid recording. Furthermore, it can automatically extract key information such as medication and treatment measures from the text converted from speech, whereas existing technologies may require manual screening and recording. This system reduces manual intervention and improves the accuracy and efficiency of information extraction. At the same time, it adds a timestamp to each record and links it with the text information, making the medical records have a clearer timeline, which makes it easier for medical staff to accurately understand the patient's condition development and the time sequence of nursing measures. This is a function that may be relatively lacking in existing nursing medical record systems. Moreover, the system can automatically integrate data to generate structured medical records according to the standardized format of nursing medical records, ensuring the standardization and consistency of medical records.

[0110] It should be noted that, in order to provide a concise description of exemplary embodiments, not all features of actual embodiments (i.e., those features that are not relevant to the best mode of carrying out the invention as currently considered, or those features that are not relevant to implementing the invention) may be omitted.

[0111] It should be understood that numerous specific implementation decisions can be made during the development of any actual implementation method, and in any engineering or design project. Such development efforts may be complex and time-consuming, but for those of ordinary skill in the art who benefit from this disclosure, the development effort will be a routine work of design, manufacturing, and production without requiring much experimentation.

[0112] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. An intelligent voice-input nursing medical record system, characterized in that, include: The voice acquisition module is used to acquire audio signals in real time using a preset wake-up word; A speech recognition module is used to convert the acquired audio signals into text information; A timestamp generation module is used to generate timestamps and associate the timestamps with the text information to obtain associated data; An automatic recording module is used to extract the medication and treatment process based on the associated data to obtain recorded data; The medical record generation module is used to integrate the recorded data according to the standardized format of nursing medical records and generate structured nursing medical records.

2. The intelligent voice input nursing medical record system as described in claim 1, characterized in that, The voice acquisition module is also used for: By continuously collecting ambient audio signals through a microphone device, a continuous audio signal x(t) that varies with time is obtained, where t is time; The continuous audio signal x(t) is divided into multiple short frames, each with a duration of 10-30 milliseconds and a length of N sampling points. The audio signal of the nth frame is represented as x. n [m], where m = 0, 1, ..., N-1; From each frame of audio signal x n Extracting the feature vector from [m], we obtain the feature vector f of the nth frame. n ; The feature vector f n The feature vector f is calculated by comparing it with the feature template T of the preset wake word. n The Euclidean distance d(f) between the feature template T and the feature template T n ,T): Where D is the dimension of the feature vector, f n,i T is the i-th element of the feature vector of the n-th frame. i This is the i-th element of the preset wake word feature template; When the Euclidean distance d(f) n When T is less than the preset wake-up threshold θ, a wake-up word is detected. When a wake word is detected, voice signals are collected in real time until a stop condition is met; the stop condition is no voice input within 2 minutes.

3. The intelligent voice input nursing medical record system as described in claim 2, characterized in that, The speech recognition module is also used for: Using Mel-frequency cepstral coefficients (MFCC), from the audio signal x n Features are extracted from [m] to obtain feature F. n [e], where e is a time frame; The feature F is represented by acoustic model A. n [e] mapped to a phoneme or acoustic unit is represented as: P(x|F n [e])=A(F n [e]) Wherein, P(x|F n [e]) represents the probability of generating a specified phoneme given the features; By constraining the output of the acoustic model A using a language model, we obtain: Where w is the word sequence, w i Let w be the i-th word in the word sequence, N be the number of words in the sequence, and P(w) be the probability of the entire word sequence w. By combining acoustic and language models, the optimal word sequence W is obtained. * To maximize the probability word sequence: Among them, P(W|F n [e]) is a given feature F n The probability of the word sequence W under the condition [e], P(W) is the prior probability of the word sequence W, P(F) is the probability of the word sequence W under the condition [e]. n [e]|W) is to generate feature F given the word sequence W. n The conditional probability of [e]; By combining feature extraction, acoustic models, and language models, the input audio signal x is processed. n [m] is converted to the corresponding text information W n .

4. The intelligent voice input nursing medical record system as described in claim 1, characterized in that, The timestamp generation module is used for: Calculate the timestamp TS: TS=T now -T0 Where T0 is the time point of the Unix epoch, T now This refers to the current time point; The timestamp generation module is also used for: The generated timestamp is associated with the text information obtained by the speech recognition module; By generating a timestamp immediately after speech recognition is completed, and combining the timestamp and text information into a data structure, the associated data is obtained: AD=(TS,W n ) In this context, AD represents associated data.

5. The intelligent voice-input nursing medical record system as described in claim 1, characterized in that, The automatic recording module is also used for: Regarding the text information W n Preprocessing is performed to obtain the preprocessed text information W. n ′; The preprocessed text information W n Extract key information from the sample, which includes drug information and treatment information, and is represented as: (Dr,Tr)=M(W) n ′), where Dr is drug information, Tr is treatment information, and M is the extraction model; The extracted key information and the timestamp are structured to obtain the recorded data: RD = (TS, Dr, Tr) Here, RD represents recorded data.

6. The intelligent voice-input nursing medical record system as described in claim 1, characterized in that, The medical record generation module is also used for: The standardized format of nursing medical records is abstracted into a nursing medical record template Te, which is used to define the names and arrangement order of each part of the medical record; Mapping each piece of information in the recorded data RD to its corresponding position in the standardized nursing medical record format is represented as follows: MD = Map(RD, Te) Where MD is the mapped data, and the structure of MD matches the standard format of nursing medical records; Map is the mapping function. The mapped data MD is then populated into the nursing record template Te to generate a complete structured nursing record MR. in, For data population operations, fill in the information in MD into the corresponding positions according to the template requirements.

Images