Medical intercom system and method based on audio processing

Abstract

The present invention relates to the field of medical voice communication technology, and discloses a medical intercom system and method based on audio processing, the system comprising: an audio acquisition module, an audio positioning module, a demand determination module, an instruction generation module, an instruction distribution module, a response personnel determination module, an instruction sending module and a response processing module. By implementing the present invention, the demand type data and the emergency level in the call audio are automatically parsed by intelligent speech recognition and semantic analysis technology, and accurate allocation and priority processing of resources are achieved, ensuring that key needs are responded to in a timely manner, accurately locating the patient's position and identifying the patient's needs, and quickly matching the most suitable response team.

Description

Technical Field

[0001] The present invention relates to the technical field of medical voice communication, and in particular to a medical intercom system and method based on audio processing. Background Art

[0002] The core of hospital intercom technology is to realize voice communication between medical staff. Specifically, the intercom captures the voice signals of medical staff through the microphone and converts them into electrical signals. Then, these electrical signals are processed and transmitted through the communication protocol software, and finally transmitted to the receiver's intercom through the network or wireless signal. The receiver's intercom then converts the electrical signal into a sound wave signal and plays it out through the speaker to realize voice communication between the two parties.

[0003] With the continuous advancement of medical technology and the increasing diversification of patient needs, cross-department collaboration and efficient communication have become important factors in improving the quality of medical services and patient satisfaction. Traditional medical intercom systems seem to be unable to cope with the situation. Most of these systems are confined to the basic call and answer framework, lacking deep intelligent processing capabilities and flexible scheduling mechanisms. Specifically, they have shortcomings in accurately locating the patient's location. They can only vaguely perceive the call signal, but cannot accurately lock the specific location of the patient, forcing the response team to consume precious time in the process of searching for the patient, which directly weakens the timeliness of the rescue operation. At the same time, in terms of identifying patient needs, traditional systems also seem to be stretched thin. They can only receive the most basic call signals and cannot intelligently analyze the specific types of services required by patients, such as emergency treatment, daily care, professional consultation, etc., as well as the degree of urgency, resulting in uneven resource allocation and failure of the priority processing mechanism. Faced with the complex situation of tight beds or patients with special needs, traditional systems seem to be even more unable to cope with the situation. It is difficult to effectively coordinate the response team across departments, and the scheduling process is cumbersome and lengthy, and inefficient. In addition, due to the lack of accurate assessment of the real-time load and availability of responders, the system is often unable to quickly match the most appropriate rescue force, thus missing out on precious treatment opportunities. More seriously, information transmission during cross-department collaboration is often delayed or misunderstood, further hindering the smooth progress of team collaboration and affecting the overall treatment experience of patients and the overall quality of medical services. Summary of the invention

[0004] In view of this, the present invention provides a medical intercom system and method based on audio processing to solve the problem of how to accurately locate the patient's position and identify the patient's needs so as to quickly match the most appropriate response team.

[0005] In a first aspect, the present invention provides a medical intercom system based on audio processing, the system comprising: an audio acquisition module, used to collect call audio of a target object using a microphone matrix arranged in a department; an audio positioning module, used to determine the identity data and location data of the target object based on the call audio; a demand determination module, used to determine the demand type data and demand urgency level data of the target object based on the call audio; an instruction generation module, used to compose an intercom scheduling instruction based on the identity data, location data, demand type data and demand urgency level data of the target object; an instruction allocation module, used to determine at least one target response team based on the intercom scheduling instruction; a response personnel determination module, used to obtain the availability data and load data of the target response team, and determine the target response personnel based on the intercom scheduling instruction, availability data and load data; an instruction sending module, used to send the intercom scheduling instruction to the intercom device of the target response personnel; a response processing module, used to receive the response audio sent by the target response personnel through the intercom device, and play the response audio using the microphone matrix.

[0006] The medical intercom system based on audio processing provided in this embodiment, through the audio acquisition module, uses a high-sensitivity microphone matrix to fully capture the call audio of the target object in the department, ensures the collection of information without omission, and provides a solid foundation for subsequent accurate processing. Through the audio positioning module, the identity data and precise location of the target object are quickly determined by advanced audio analysis technology, effectively shortening the time for the response team to find patients and improving the response efficiency. Through the demand determination module, the demand type data and emergency level in the call audio are automatically parsed by intelligent speech recognition and semantic analysis technology to achieve accurate allocation and priority processing of resources, ensuring that key needs are responded to in a timely manner. Through the instruction generation module, detailed intercom scheduling instructions are generated by comprehensive multivariate data of the target object, providing comprehensive information support for scheduling decisions and promoting smooth cross-department collaboration. Through the instruction allocation module, at least one target response team is quickly matched and allocated with an intelligent scheduling algorithm, optimizing resource allocation and ensuring efficient use of rescue forces. Through the response personnel determination module, the most suitable response personnel are accurately selected based on the real-time team availability and load data, avoiding waste of resources and ensuring response speed and quality. Through the command sending module, the intercom dispatch command is quickly transmitted to the intercom device of the target responder using instant messaging technology, reducing information transmission delay and improving response agility. Through the response processing module, the microphone matrix is ​​used as the medium to play the responder's reply audio, ensuring clear transmission and wide coverage of information, and enhancing communication effectiveness.

[0007] In an optional embodiment, the medical intercom system based on audio processing also includes an instruction display module, which includes a first display unit and a second display unit, wherein the first display unit is used to use a total display device to display all intercom scheduling instructions in a preset area and processing progress data of each intercom scheduling instruction; the second display unit is used to use multiple sub-display devices to respectively display the intercom scheduling instructions sent to each response team and the processing progress data of each intercom scheduling instruction, and the sub-display devices correspond one-to-one to the response teams.

[0008] In an optional implementation, the audio positioning module includes:

[0009] The pre-processing unit is used to filter and denoise the call audio. The denoising formula is as follows:

[0010] ;

[0011] Among them, Y(k) represents the amplitude spectrum estimation of the denoised call audio, X(k) represents the amplitude spectrum of the call audio in the frequency domain, k represents the frequency index, that is, different frequency components, |X(k)| represents the absolute value of X(k), N(k) represents the estimation of the amplitude spectrum of the call audio in the frequency domain, α represents the over-subtraction factor, α is a real number less than or equal to 1, when α=1, it means that the noise is completely subtracted; when α<1, it means that the noise is partially subtracted; a voiceprint extraction unit is used to extract the voiceprint of the preprocessed call audio to obtain a voiceprint to be matched; a voiceprint matching unit is used to match the voiceprint to be matched with the pre-stored voiceprint corresponding to each grid point in the pre-stored voiceprint map to obtain a matching result; a confirmation unit is used to determine the identity data and location data of the target object based on the matching result, the identity data includes the pre-stored case data of the target object, and the location data includes the pre-stored bed location and the current call location of the target object.

[0012] In an optional implementation, the demand determination module includes: a demand type determination unit, which is used to extract the call text of the call audio, use the pre-trained named entity to identify the key entities in the call text, match the key entities with the predefined demand type dictionary, and obtain demand type data, the demand type data includes first-level demands and second-level demands, the first-level demands include medical demands and non-medical demands, and the second-level demands are subordinate demands of the first-level demands; a sentiment scoring unit, which is used to calculate the sentiment score based on the call text, and the calculation formula is as follows:

[0013] ;

[0014] Among them, Sentiment Score represents the sentiment score of the call text, n is the total number of words in the call text, word_scorei is the sentiment score of the i-th word, and the sentiment score of the word is obtained by querying the sentiment dictionary, weighti is the weight of the i-th word, and the weight of the word is calculated based on word frequency-inverse document frequency; the urgency scoring unit is used to locate the emergency words in the call text based on the call text and the predefined emergency vocabulary table, and calculate the urgency score of each emergency word; the demand urgency level evaluation unit is used to determine the demand urgency level data based on the sentiment score and the urgency score, and the calculation formula is as follows:

[0015] ;

[0016] Among them, Emergency Level represents the emergency level data, γ and δ are weight coefficients used to adjust the relative importance of sentiment score and emergency word score in the emergency level calculation, Sentiment Score is the sentiment score, which is a continuous value, m is the number of emergency words, and urgency_scorej is the urgency score of the jth emergency word.

[0017] In an optional embodiment, the instruction allocation module includes: a first allocation unit, which is used to determine the department to which the target object currently belongs based on the intercom scheduling instruction when it is detected that the first-level demand in the intercom scheduling instruction is a medical demand, and determine the medical demand response team based on the second-level demand in the intercom scheduling instruction, and take the response team corresponding to the currently affiliated department and the medical demand response team as the target response team; a second allocation unit, which is used to determine the target response team based on the second-level demand in the intercom scheduling instruction when it is detected that the first-level demand in the intercom scheduling instruction is a non-medical demand.

[0018] In an optional embodiment, the response personnel determination module includes: a priority calculation unit, which is used to obtain all unfinished intercom scheduling instructions in the target response team, calculate the pending response priority data of each unfinished intercom scheduling instruction, and obtain a pending response priority data set; a demand type statistics unit, which is used to obtain the pending response demand type data of each unfinished intercom scheduling instruction, and obtain a pending response demand type data set; a determination unit, which is used to determine the target response personnel based on the pending response priority data set, the pending response demand type data set, the availability data and the load data, wherein the availability data includes the skill data of each response personnel in the target response team, and the load data includes the number of target objects assigned to each response personnel in the target response team as the target response personnel.

[0019] In an optional embodiment, the medical intercom system based on audio processing also includes: a soothing module, which is used to generate soothing audio based on intercom scheduling instructions and emotional scores when no response audio emitted by the target responder through the intercom device is detected within a preset time, and use a microphone matrix to play the soothing audio.

[0020] In a second aspect, the present invention provides a medical intercom method based on audio processing, the method comprising: using a microphone matrix arranged in a department to collect call audio of a target object; determining the identity data and location data of the target object based on the call audio; determining the demand type data and demand urgency level data of the target object based on the call audio; forming an intercom scheduling instruction based on the identity data, location data, demand type data and demand urgency level data of the target object; determining at least one target response team based on the intercom scheduling instruction; obtaining the availability data and load data of the target response team, and determining the target response personnel based on the intercom scheduling instruction, availability data and load data; sending the intercom scheduling instruction to the intercom device of the target response personnel; receiving the response audio emitted by the target response personnel through the intercom device, and playing the response audio using the microphone matrix. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0022] Figure 1 is a principle block diagram of a specific example of a medical intercom system based on audio processing according to an embodiment of the present invention;

[0023] Figure 2 It is a flowchart of a specific example of a medical intercom method based on audio processing according to an embodiment of the present invention. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0025] The core of hospital intercom technology is to realize voice communication between medical staff. Specifically, the intercom captures the voice signals of medical staff through the microphone and converts them into electrical signals. Then, these electrical signals are processed and transmitted through the communication protocol software, and finally transmitted to the receiver's intercom through the network or wireless signal. The receiver's intercom then converts the electrical signal into a sound wave signal and plays it out through the speaker to realize voice communication between the two parties.

[0026] With the continuous advancement of medical technology and the increasing diversification of patient needs, cross-department collaboration and efficient communication have become important factors in improving the quality of medical services and patient satisfaction. Traditional medical intercom systems seem to be unable to cope with the situation. Most of these systems are confined to the basic call and answer framework, lacking deep intelligent processing capabilities and flexible scheduling mechanisms. Specifically, they have shortcomings in accurately locating the patient's location. They can only vaguely perceive the call signal, but cannot accurately lock the specific location of the patient, forcing the response team to consume precious time in the process of searching for the patient, which directly weakens the timeliness of the rescue operation. At the same time, in terms of identifying patient needs, traditional systems also seem to be stretched thin. They can only receive the most basic call signals and cannot intelligently analyze the specific types of services required by patients, such as emergency treatment, daily care, professional consultation, etc., as well as the degree of urgency, resulting in uneven resource allocation and failure of the priority processing mechanism. Faced with the complex situation of tight beds or patients with special needs, traditional systems seem to be even more unable to cope with the situation. It is difficult to effectively coordinate the response team across departments, and the scheduling process is cumbersome and lengthy, and inefficient. In addition, due to the lack of accurate assessment of the real-time load and availability of responders, the system is often unable to quickly match the most appropriate rescue force, thus missing out on precious treatment opportunities. More seriously, information transmission during cross-department collaboration is often delayed or misunderstood, further hindering the smooth progress of team collaboration and affecting the overall treatment experience of patients and the overall quality of medical services.

[0027] In view of the technical problems mentioned in the background technology, the present application embodiment provides a medical intercom system based on audio processing, see Figure 1 As shown, the system includes: an audio acquisition module, an audio positioning module, a demand determination module, an instruction generation module, an instruction allocation module, a responder determination module, an instruction sending module and a response processing module. The functions of each of the above components are introduced below.

[0028] The audio collection module is used to collect the call audio of the target object using the microphone matrix set up in the department.

[0029] Specifically, the microphone matrix should be arranged reasonably according to the specific layout and size of the department. Generally, microphones should be evenly distributed in all corners or key positions of the department to ensure that the entire area can be covered and audio signals from all directions can be captured. The audio acquisition module is responsible for collecting audio signals such as patients' calls and medical staff's conversations in the medical environment. Using a high-sensitivity microphone array, directional pickup and noise suppression are supported. The sampling rate determines the frequency resolution of the audio signal, and the bit depth determines the dynamic range and signal-to-noise ratio of the audio signal. When selecting these parameters, they need to be reasonably configured according to the application scenario and sound quality requirements. As an example, the sampling rate of the microphone array can be 48kHz, the bit depth is 24bit, and the signal-to-noise ratio is greater than or equal to 70dB.

[0030] The audio positioning module is used to determine the identity data and location data of the target object based on the call audio.

[0031] Specifically, the audio positioning module includes:

[0032] The pre-processing unit is used to filter and denoise the call audio. The denoising formula is as follows:

[0033] ;

[0034] Among them, Y(k) represents the amplitude spectrum estimation of the denoised call audio, X(k) represents the amplitude spectrum of the call audio in the frequency domain, k represents the frequency index, that is, different frequency components, |X(k)| represents the absolute value of X(k), N(k) represents the estimation of the amplitude spectrum of the call audio in the frequency domain, α represents the over-subtraction factor, α is a real number less than or equal to 1, when α=1, it means that the noise is completely subtracted; when α<1, it means that the noise is partially subtracted.

[0035] Furthermore, the filtering and denoising processing in the preprocessing unit are audio processing that can significantly improve the quality of the audio signal and provide a better foundation for subsequent processing. The call audio is converted from the time domain to the frequency domain, and its amplitude spectrum X(k) is calculated using the fast Fourier transform. Frequency domain processing can better separate signals and noise, because noise and signals often have different characteristics in the frequency domain. The minimum statistics method is used to estimate the noise amplitude spectrum N(k) in the call audio. According to the provided formula Y(k) = |X(k)| -α·N(k), the denoised audio amplitude spectrum is calculated. Among them, α is the over-reduction factor, which is used to control the intensity of denoising. The smaller the value of α, the weaker the denoising effect, and the more original audio information is retained; the larger the value of α, the stronger the denoising effect, but some useful audio information will be lost. The appropriate α value is selected through historical data training. In practical applications, multiple attempts and adjustments will be made according to the characteristics of the noise and the requirements for sound quality, and parameters such as smoothing factor and update rate are used to control the sensitivity and stability of noise estimation. When processing audio frames, the smoothness and real-time performance of the processing are controlled by setting overlap and frame skipping. The denoised amplitude spectrum Y(k) is combined with the phase information of the original audio and converted back to the time domain using inverse fast Fourier transform to obtain the denoised call audio.

[0036] The voiceprint extraction unit is used to extract the voiceprint of the preprocessed call audio to obtain the voiceprint to be matched.

[0037] Furthermore, the preprocessed audio signal is divided into multiple frames of fixed length, such as 20-30 milliseconds, and there can be partial overlap between frames, such as 10 milliseconds, to ensure continuity. A Hamming window is applied to each frame of the audio signal to reduce edge effects. The acoustic feature parameters of each frame of audio are calculated, and these parameters can reflect the speaker's individual characteristics such as the shape of the vocal tract and pronunciation habits. The acoustic features here include: Mel frequency cepstral coefficients, linear prediction coefficients, fundamental frequency and resonance peaks. In order to overcome the rate differences and speaking styles between different speakers, the extracted features are subjected to dynamic time warping. The most representative features are selected from the large number of extracted acoustic features to reduce the amount of calculation and improve recognition efficiency. The principal component analysis technology is used to further reduce the feature dimension while retaining as much information as possible that is useful for recognition. The extracted and processed feature vectors are combined into a voiceprint to be matched, and the voiceprint to be matched is used as an identifier of the identity of the target object.

[0038] The voiceprint matching unit is used to match the voiceprint to be matched with the pre-stored voiceprints corresponding to each grid point in the pre-stored voiceprint map to obtain a matching result.

[0039] Furthermore, the main purpose of the voiceprint matching unit is to compare the voiceprint to be matched extracted from the preprocessed call audio with the pre-stored voiceprint map to identify the identity and approximate location of the target object. After the voiceprint extraction unit is completed, the feature vector of the acoustic features of the voiceprint to be matched is obtained. The pre-stored voiceprint map is a database that stores pre-stored voiceprint templates corresponding to each grid point in the hospital (such as wards, corridors, etc.). These templates are obtained by collecting and processing audio signals of specific locations (such as beds, nurse stations, etc.) in advance. The voiceprint template of each grid point contains the voiceprint features of multiple known patients and the background noise features of the location. The template matching method is used to compare the voiceprint to be matched with each grid point voiceprint template in the voiceprint map. The similarity or distance metric, such as Euclidean distance, cosine similarity, etc., is calculated to evaluate the matching degree between the voiceprint to be matched and the voiceprint templates of each grid point. According to the set threshold or sorting mechanism, the grid point voiceprint template with the highest matching degree is selected as the candidate matching result. If there are multiple candidate results, further screening or comprehensive judgment is required based on the location information and time information. Output the matching results, including the most matching grid point identifier and the corresponding voiceprint template information.

[0040] The confirmation unit is used to determine the identity data and location data of the target object based on the matching result, the identity data includes the pre-stored case data of the target object, and the location data includes the pre-stored bed location and current call location of the target object.

[0041] Furthermore, the confirmation unit further confirms the identity data and precise location data of the target object based on the matching results of the voiceprint matching unit, combined with case data, bed location and other information. The grid point identifier in the matching result is used to query the case database or patient information system associated with it. If the grid point has been associated with a specific patient case, the identity data and case information of the patient are directly extracted as the identity data of the target object. If it is not directly associated with the patient, it is necessary to obtain the identity through AI voice interaction with the patient. Combine the grid point identifier and hospital layout information to determine the bed number of the target object.

[0042] The demand determination module is used to determine the demand type data and demand urgency level data of the target object based on the call audio.

[0043] Specifically, the demand determination module includes:

[0044] The demand type determination unit is used to extract the call text of the call audio, use the pre-trained named entity recognition to identify the key entities in the call text, match the key entities with the predefined demand type dictionary, and obtain the demand type data. The demand type data includes the first-level demand and the second-level demand. The first-level demand includes medical demand and non-medical demand, and the second-level demand is the subordinate demand of the first-level demand.

[0045] Furthermore, medical needs cover all needs directly related to medical treatment, disease diagnosis, health monitoring, etc. Medical needs are the most core and important part of the daily work of hospitals or medical institutions because they are directly related to the life safety and physical health of patients. Non-medical needs are opposite to medical needs. Non-medical needs mainly involve patients' daily life care, psychological support, environmental needs, etc. Although these needs do not directly involve medical treatment, they are also crucial for the overall rehabilitation and hospitalization experience of patients. As a subordinate of the first-level needs, the second-level needs provide more specific and detailed content. The second-level needs correspond to different categories in the first-level needs, and further refine the specific needs of each category. For medical needs, the second-level needs include: emergency treatment, such as emergency treatment for cardiac arrest, dyspnea, etc.; drug treatment, including drug prescription, medication guidance, drug side effect monitoring, etc.; surgical treatment, involving various surgical operations and their preoperative preparation and postoperative care; rehabilitation care, professional care and rehabilitation training for patients in the rehabilitation period. For non-medical needs, the second-level needs include: life care, such as dietary arrangements, personal hygiene, excretion assistance, etc.; psychological support, providing psychological counseling, emotional comfort, family communication and other services; environmental needs, adjusting the ward environment, providing necessary auxiliary equipment, etc. Use a pre-trained named entity recognition model that can identify key entities in the text, such as names of people, places, institutions, symptoms, disease names, drug names, etc. Ensure that the NER model is properly fine-tuned for the medical field to improve the accuracy of identifying medical-related entities. Extract all identified entities and classify them into different entity types, such as patient information, symptom descriptions, drug needs, etc., and output the identified key entities and their types. Obtain the identified key entities and their types and the predefined demand type dictionary, traverse all identified key entities, and preliminarily determine whether the call belongs to medical needs or non-medical needs based on their types and context information. Use rules or machine learning models to assist in judgment to ensure accuracy. For calls determined to be medical needs, further subdivide them into specific subordinate needs, such as first aid needs, consultation needs, and drug purchase needs, based on entities such as symptoms, diseases, and drug needs. Use predefined requirement type dictionaries for exact or fuzzy matching to ensure coverage of as many cases as possible. Requirement type data, including first-level and second-level requirements.

[0046] The sentiment scoring unit is used to calculate the sentiment score based on the call text. The calculation formula is as follows:

[0047] ;

[0048] Among them, Sentiment Score represents the sentiment score of the call text, n is the total number of words in the call text, word_scorei is the sentiment score of the i-th word, the sentiment score of the word is obtained by querying the sentiment dictionary, weighti is the weight of the i-th word, and the weight of the word is calculated based on the term frequency-inverse document frequency.

[0049] Furthermore, a dictionary containing a large number of words and their corresponding sentiment scores is pre-constructed. Positive numbers represent positive sentiment, negative numbers represent negative sentiment, and zero represents neutral sentiment. The sentiment dictionary is generated manually, integrated from existing resources, or automatically using machine learning models. To maintain the accuracy of sentiment analysis, the sentiment dictionary is updated regularly to include new words and phrases, especially terms specific to the medical field. The call text is segmented into individual words or phrases. For Chinese text, word segmentation is also required. Common stop words in the text, such as "的", "是", "在", etc., which do not contribute substantially to sentiment analysis, are removed. For English text, the words can be reduced to their base form (lemmatization) or their stems can be extracted (stemming) to reduce the impact of lexical variations on the sentiment score. For each preprocessed word, its corresponding sentiment score word_scorei is queried in the sentiment dictionary. Calculate the frequency of each word's occurrence in the call text. Estimate the importance of the word in the entire dataset or corpus. If a word appears frequently in multiple documents, its IDF value is low; conversely, if a word appears rarely, its IDF value is high. Multiply TF and IDF to obtain the weight weighti of each word, which helps to highlight words that are important in the text but not common in the entire dataset. According to the formula, multiply the sentiment scores of all words by their corresponding weights and sum them to obtain the sentiment score SentimentScore of the entire call text.

[0050] An urgency scoring unit is used to locate the urgent words in the call text based on the call text and a predefined urgent vocabulary list, and calculate the urgency scores of each urgent word.

[0051] Furthermore, a series of words and phrases representing emergency situations are collected in advance, and these words may include "first aid", "fire", "heart attack", "serious injury" and the like. Emergency words are classified according to the degree of urgency, and an urgency score is assigned to each word. This score can be a numerical value or other measure that can represent the difference in urgency. A predefined emergency vocabulary is obtained. The call text is preprocessed, including steps such as word segmentation, stop word removal, lemma restoration / stemming, etc., to ensure that the emergency words can be accurately identified. The preprocessed call text is matched with the emergency vocabulary to locate the emergency words appearing in the text. In order to improve the accuracy of the match, fuzzy matching techniques such as regular expressions or string similarity algorithms can be considered to capture variants or spelling errors of the words. For each located emergency word, a corresponding score is directly assigned according to its urgency score in the emergency vocabulary. If the call text contains multiple emergency words, a weighted summation or maximum value can be considered to calculate the comprehensive urgency score. When weighted summation is performed, weights can be assigned according to the urgency of the words and the importance of the context. In some cases, the urgency of the emergency words is affected by the context. Therefore, the urgency scoring unit needs to combine context analysis to more accurately assess urgency. Urgent words containing negative words, such as "no heart attack", need special treatment to reduce their urgency scores or ignore them completely.

[0052] The demand urgency level evaluation unit is used to determine the demand urgency level data based on the sentiment score and the urgency score. The calculation formula is as follows:

[0053] ;

[0054] Among them, Emergency Level represents the emergency level data, γ and δ are weight coefficients used to adjust the relative importance of sentiment score and emergency word score in the emergency level calculation, Sentiment Score is the sentiment score, which is a continuous value, m is the number of emergency words, and urgency_scorej is the urgency score of the jth emergency word.

[0055] Furthermore, the two weight coefficients γ and δ are used to adjust the relative importance of the sentiment score and the urgency word score in the calculation of the urgency level. If the system focuses more on responding to the appearance of urgency words, δ can be set relatively high; if the sentiment score is also crucial for judging the degree of urgency, the value of γ should also be adjusted accordingly. The sentiment score is a continuous value that indicates the strength of the emotional tendency expressed in the call text. This score is calculated by the sentiment scoring unit and usually ranges from -1 (extremely negative) to 1 (extremely positive). m is the number of urgency words, that is, the number of words in the call text that are successfully matched by the urgency vocabulary. urgency_scorej is the urgency score of the jth emergency word, determined according to the definition of the word in the emergency vocabulary. The urgency score can be a simple accumulation of the urgency scores of all emergency words, or a weighted accumulation, where the weights can be assigned according to the urgency of the emergency words, contextual importance, or other factors. Emergency Level is the final emergency level data, obtained according to a given calculation formula. The calculation formula combines the sentiment score and the urgency score, and balances the impact of the two on the emergency level by adjusting the weight coefficients γ and δ. If no emergency words are recognized in the call text, but the sentiment score indicates that the caller is very anxious or panic, the emergency level evaluation unit still needs to give an appropriate emergency level so that the system can respond appropriately. Similarly, if the call text contains multiple high-urgency words, but the sentiment score is relatively low, the evaluation unit also needs to comprehensively consider these two types of information to determine the emergency level.

[0056] The instruction generation module is used to compose an intercom dispatch instruction based on the identity data, location data, demand type data and demand emergency level data of the target object.

[0057] Specifically, organize the target object's name, age, gender, and pre-stored case data, such as allergy history, previous medical history and other key information. Confirm the target object's current exact location, including ward number, bed number, and whether it is in a movable state, such as whether a stretcher or wheelchair is needed for transfer. According to the analysis results of the demand determination module, clarify the first-level and second-level needs. Emergency level data evaluation: Determine the urgency of the demand based on the emergency level data calculated based on sentiment score and emergency vocabulary score. Construct the command content, and the command header includes basic information such as command generation time and system identification for subsequent tracking and recording. Clearly mark the target object's name, identity characteristics (such as the bracelet number worn, bed number) and emergency contact information. Describe in detail the target object's current location, department, whether it is in a safe area, and whether special transfer measures are required. Clearly indicate the first-level and second-level needs. For medical needs, it is also necessary to include possible disease descriptions, required medical equipment, etc.; for non-medical needs, the required service types are detailed, such as life care, psychological support, etc. Clearly mark the urgency level of the demand to guide the response team's priority processing. Use a unified instruction format, including text description, digital coding, or a combination of the two, to ensure that instructions can be accurately parsed between different systems or devices. Encode key information to reduce the possibility of misinterpretation and ambiguity. Key information includes demand type, urgency level, etc. Use the instruction generation module to automatically generate intercom dispatch instructions based on the integrated data. Record the time the instruction was sent, the recipient, and the time the instruction was received to facilitate subsequent tracking of the execution of the instruction. Establish a feedback mechanism for the response team on the execution of the instruction to ensure that the instruction is effectively executed and adjusted or supplemented when necessary.

[0058] The instruction distribution module is used to determine at least one target response team based on the intercom scheduling instruction.

[0059] Specifically, the instruction distribution module includes:

[0060] The first allocation unit is used to determine the department to which the target object currently belongs based on the intercom scheduling instruction when it is detected that the first-level demand in the intercom scheduling instruction is a medical demand, and to determine the medical demand response team based on the second-level demand in the intercom scheduling instruction, and to take the response team corresponding to the current department and the medical demand response team as the target response team.

[0061] Furthermore, the first allocation unit is mainly responsible for handling situations where the first-level demand in the intercom dispatch instruction is a medical demand. In this scenario, the unit first identifies the department where the target object is currently located, and then further determines the medical demand response team that needs to be mobilized based on the specific content of the second-level demand. The current department is usually the one that knows the patient's condition best, and its corresponding response team can quickly assess the patient's condition and provide preliminary treatment or rescue measures based on professional knowledge. At the same time, according to the specific content of medical needs, such as trauma treatment, cardiovascular disease treatment, etc., mobilizing the corresponding professional medical team can ensure that the patient receives the most accurate and effective treatment.

[0062] The second allocation unit is used to determine a target response team based on a second level demand in the intercom dispatch instruction when it is detected that the first level demand in the intercom dispatch instruction is a non-medical demand.

[0063] Furthermore, the second allocation unit is responsible for handling the situation where the first-level demand in the intercom dispatch instruction is a non-medical demand. In this scenario, the unit directly determines the non-medical response team that needs to be mobilized based on the specific content of the second-level demand. These teams are responsible for non-medical work such as daily life care and psychological support for patients, which plays an important role in improving patients' hospitalization experience and rehabilitation effects. As an example, in a ward, a patient made a call through the medical intercom system because of postoperative pain and inability to turn over by himself. After the system confirms the patient's identity and location through the audio positioning module, the demand determination module analyzes that the first-level demand is "non-medical demand" and the second-level demand is "life care". After receiving this information, the second allocation unit directly selects the nursing team as the target response team based on the second-level demand "life care" and sends an intercom dispatch instruction to the intercom device of the nursing team. After receiving the instruction, the nursing team immediately goes to the ward to provide patients with life care services such as turning over and pain relief. By automatically allocating response teams for non-medical needs, the non-medical workload of medical staff can be reduced, allowing them to focus more on medical treatment.

[0064] The response personnel determination module is used to obtain the availability data and load data of the target response team, and determine the target response personnel based on the intercom scheduling instructions, availability data and load data.

[0065] Specifically, the responder determination module includes:

[0066] The priority calculation unit is used to obtain all unfinished intercom scheduling instructions in the target response team, calculate the response priority data of each unfinished intercom scheduling instruction, and obtain a response priority data set.

[0067] Furthermore, the priority calculation unit is responsible for obtaining all unfinished intercom scheduling instructions in the target response team, calculating the priority data to be responded to of each instruction through a series of algorithms, and finally generating a priority data set to be responded to. The priority calculation unit extracts all currently unfinished intercom scheduling instructions from the system database, and each intercom scheduling instruction includes but is not limited to the identity data, location data, demand type data and demand urgency level data of the target object. Different weight values ​​are assigned to the instructions according to the demand urgency level data. The higher the urgency level, the greater the weight value. Different demand types have different effects on priority. For example, medical demand has a higher priority weight. Considering the instruction generation time, the priority of instructions that have not been processed for a long time can be appropriately increased. Refer to the processing time and effect of similar historical cases to fine-tune the priority. The weighted sum method or hierarchical analysis method is used to comprehensively calculate the above factors to obtain the priority to be responded to of each instruction. The generated priority data set is stored in the form of a list or database table for subsequent processing and query. The calculated priority data set to be responded to is output to the determination unit as input. The system status is monitored in real time, the priority of the newly added or changed instructions is recalculated, and the data set is updated.

[0068] The demand type statistical unit is used to obtain the pending response demand type data of each unfinished intercom scheduling instruction, and obtain a pending response demand type data set.

[0069] Furthermore, the demand type statistics unit is responsible for collecting and counting the demand type data to be responded to for each unfinished intercom scheduling instruction to form a demand type data set to be responded to. This step helps to analyze the current demand distribution and provide a basis for resource allocation and personnel arrangement. Extract demand type data from the unfinished intercom scheduling instructions obtained by the priority calculation unit. Classify the demand type of each instruction according to the classification standards of the first-level demand and the second-level demand. Count the number of instructions under each category, and count the number of specific demand subtypes under each category, such as heart disease emergency and trauma treatment in the medical type. Generate a demand type data set to be responded to, including the number of instructions in each category and subcategory. Display the demand type distribution in the form of a chart for intuitive understanding. Monitor the changes in instructions in the system in real time, re-count the demand type of new or changed instructions, and update the data set.

[0070] A determination unit is used to determine the target response personnel based on a set of priority data to be responded to, a set of type data of demand to be responded to, availability data and load data, wherein the availability data includes the skill data of each response personnel in the target response team, and the load data includes the number of target objects allocated to each response personnel in the target response team as a target response personnel.

[0071] Further, the determination unit determines the target responder after comprehensive evaluation based on the data set of priority to be responded, the data set of demand type to be responded, and the availability data and load data of the target response team. The output results of the priority calculation unit and the demand type statistics unit are obtained, and the availability data and load data of the target response team are integrated. The responder's skill data is used to evaluate whether he meets the professional requirements of the instruction demand type. The number of target objects and workload currently assigned to the responder are considered to avoid over-allocation leading to low response efficiency. When the skill matching and load are reasonable, the responder corresponding to the instruction with high priority to be responded is given priority. A greedy algorithm is used to weigh and decide between multiple evaluation criteria to assign the most suitable target responder to each unfinished instruction. The target responder allocation plan is output, including the responder information corresponding to each instruction. A feedback mechanism is provided to collect the responder's feedback on the allocation plan for subsequent optimization and adjustment. The system status and the work status of the target response team are monitored in real time, and the newly added or changed instructions and the status changes of the responders are dynamically adjusted and optimized.

[0072] The command sending module is used to send the intercom dispatch command to the intercom device of the target responding personnel.

[0073] Specifically, before sending, the content of the instruction is reviewed to ensure that it is correct and then confirmed to be sent. According to the configuration of the target response team, the intercom dispatch instruction is sent to relevant personnel through multiple channels such as the intercom system, SMS, and email.

[0074] The response processing module is used to receive the response audio sent by the target responder through the intercom device and play the response audio using the microphone matrix.

[0075] Specifically, the beamforming technology of the microphone matrix is ​​used to transmit the received response audio to the target object or designated area, improving the pertinence and efficiency of audio transmission. When processing and playing audio signals, the privacy protection principle is strictly followed to avoid leaking the personal privacy information of the target object and responder. The reception time, processing process, playback effect and feedback of the target object of the response audio are recorded in detail to facilitate subsequent analysis and tracing.

[0076] In an optional manner of some embodiments, the medical intercom system based on audio processing further includes an instruction display module, which includes a first display unit and a second display unit, wherein:

[0077] The first display unit is used to display all intercom scheduling instructions in a preset area and processing progress data of each intercom scheduling instruction using a general display device.

[0078] Specifically, the first display unit is used as an overview interface to provide comprehensive monitoring and overview of the intercom scheduling instructions in the entire medical intercom system. The display interface of the total display device is divided into multiple areas, each of which can represent different departments, floors or preset areas, so as to intuitively view the intercom scheduling status of each area. A real-time refresh mechanism is adopted to ensure that the displayed information is always up to date. All intercom scheduling instructions in the preset area are displayed in a list form, and each instruction includes basic information such as timestamp, target object, demand type, and emergency level. Different colors are used for coding according to the emergency level or processing status of the intercom scheduling instruction for quick identification. A progress bar is equipped for each intercom scheduling instruction to display its processing progress in real time, including various stages such as assigned team, response personnel confirmation, processing, and completion. A timestamp is displayed next to each processing stage to record the start and end time of the stage, so as to track and evaluate the processing efficiency. Filtering and sorting functions are provided to allow staff to filter intercom scheduling instructions according to conditions such as time, emergency level, department, etc., or sort them in a specific order. Click on an intercom dispatch instruction to expand and view more detailed information, including detailed information of the target object, historical call records, feedback from the processing team and responders, etc.

[0079] The second display unit is used to use multiple sub-display devices to respectively display the intercom scheduling instructions sent to each response team and the processing progress data of each intercom scheduling instruction, and the sub-display devices correspond to the response teams one by one.

[0080] The second display unit focuses on providing each response team with a dedicated command display interface to improve the coordination efficiency and response speed within the team. Each response team is allowed to customize the layout and style of the display interface according to actual needs to improve the readability and ease of use of information. Only the intercom dispatch instructions sent to the response team are displayed to reduce the interference of irrelevant information. Automatic sorting according to the urgency level or processing status of the intercom dispatch instructions ensures that team members give priority to important and urgent instructions. Keep in sync with the first display unit and update the processing progress data of the intercom dispatch instructions in real time to ensure that team members are aware of the latest situation. Allow team members to submit processing status through the sub-display device, such as received, processing, completed, etc., which will be automatically synchronized to the first display unit and other related systems. Integrate internal communication functions on the sub-display device to facilitate rapid communication and collaboration between team members. When allocating tasks within the team, the availability data and load data of team members can be directly viewed through the sub-display device to make reasonable allocation decisions. Through the above-mentioned refinement, the first display unit and the second display unit can respectively provide comprehensive, real-time and personalized information display and interactive experience for the manager and response team of the medical intercom system, thereby improving the operating efficiency and response speed of the entire system.

[0081] In some optional embodiments, the medical intercom system based on audio processing further includes:

[0082] The soothing module is used to generate soothing audio based on the intercom scheduling instructions and emotion scores when no response audio is detected from the target responder through the intercom device within a preset time, and to play the soothing audio using the microphone matrix.

[0083] Specifically, the comfort module plays an important role in the medical intercom system, especially in emergency situations. When the target responder fails to respond in time, playing the comfort audio can effectively relieve the target object's anxiety and remind the system administrator or other response team to pay attention to the current situation. Set a preset time threshold, such as 1 minute, to monitor whether the response audio sent by the target responder through the intercom device is received within the specified time after the intercom dispatch instruction is sent. When no response audio is detected within the preset time, the comfort module will be activated. Read the intercom dispatch instruction that has not been responded to, and analyze the key information therein, such as the identity, location, demand type and urgency level of the target object. Combined with historical data or preset rules, evaluate the emotional state of the target object when making a call and give an emotional score. This score can be estimated based on factors such as tone, speech speed, and vocabulary selection in the call audio. Customize the content of the comfort audio according to the content and emotional score of the intercom dispatch instruction. For example, for urgent calls with high emotional scores, the comfort audio contains sentences such as "We are dealing with it urgently, please stay calm, help will arrive soon"; while for non-urgent calls with low emotional scores, it focuses on encouraging and supportive words. Using advanced speech synthesis technology, customized text content is converted into natural and fluent speech. During the synthesis process, the intonation, speed, volume and other parameters of the speech can be adjusted to better match the emotional color of the soothing audio. According to the location information of the target object, the most suitable microphone array or specific speaker is selected to play the soothing audio to ensure that the audio can be clearly conveyed to the target object. The volume and sound quality of the soothing audio are optimized before playing to ensure that it can be clearly heard in the environment where the target object is located and does not cause discomfort. Set the number of loop playbacks or duration of the soothing audio, and stop playing when specific conditions are met, such as receiving a response audio or manually stopping it by the system administrator. Record the playback time, number of playbacks and the target object's reaction to the soothing audio, such as whether to continue calling, emotional changes, etc., to provide a basis for subsequent improvements and optimizations. Key information such as the triggering, audio generation, playback and feedback of the soothing module is recorded in the system log for subsequent analysis and tracking. Through the above refinement, the soothing module can automatically generate and play appropriate soothing audio when the target responder fails to respond in time to relieve the target object's anxiety and remind the system administrator or other response team to pay attention to the current situation.

[0084] The medical intercom system based on audio processing provided in this embodiment, through the audio acquisition module, uses a high-sensitivity microphone matrix to fully capture the call audio of the target object in the department, ensures the collection of information without omission, and provides a solid foundation for subsequent accurate processing. Through the audio positioning module, the identity data and precise location of the target object are quickly determined by advanced audio analysis technology, effectively shortening the time for the response team to find patients and improving the response efficiency. Through the demand determination module, the demand type data and emergency level in the call audio are automatically parsed by intelligent speech recognition and semantic analysis technology to achieve accurate allocation and priority processing of resources, ensuring that key needs are responded to in a timely manner. Through the instruction generation module, detailed intercom scheduling instructions are generated by comprehensive multivariate data of the target object, providing comprehensive information support for scheduling decisions and promoting smooth cross-department collaboration. Through the instruction allocation module, at least one target response team is quickly matched and allocated with an intelligent scheduling algorithm, optimizing resource allocation and ensuring efficient use of rescue forces. Through the response personnel determination module, the most suitable response personnel are accurately selected based on the real-time team availability and load data, avoiding waste of resources and ensuring response speed and quality. Through the command sending module, the intercom dispatch command is quickly transmitted to the intercom device of the target responder using instant messaging technology, reducing information transmission delay and improving response agility. Through the response processing module, the microphone matrix is ​​used as the medium to play the responder's reply audio, ensuring clear transmission and wide coverage of information, and enhancing communication effectiveness.

[0085] The above is an embodiment of the medical intercom system based on audio processing provided by the present application. The following is an introduction to other embodiments of the medical intercom system based on audio processing provided by the present application. Please refer to the following for details.

[0086] The embodiment of the present invention also discloses a medical intercom method based on audio processing, such as Figure 2 As shown, the method includes: using a microphone matrix set in the department to collect call audio of the target object; determining the identity data and location data of the target object based on the call audio; determining the demand type data and demand urgency level data of the target object based on the call audio; forming an intercom scheduling instruction based on the identity data, location data, demand type data and demand urgency level data of the target object; determining at least one target response team based on the intercom scheduling instruction; obtaining the availability data and load data of the target response team, and determining the target response personnel based on the intercom scheduling instruction, availability data and load data; sending the intercom scheduling instruction to the intercom device of the target response personnel; receiving the response audio sent by the target response personnel through the intercom device, and playing the response audio using the microphone matrix.

[0087] The technical solution of the present invention has the following advantages:

[0088] The medical intercom method based on audio processing provided in this embodiment uses a highly sensitive microphone matrix to fully capture the call audio of the target object in the department, ensure the collection of information without omission, and provide a solid foundation for subsequent accurate processing. By quickly determining the identity data and precise location of the target object with advanced audio analysis technology, the time for the response team to find the patient is effectively shortened, and the response efficiency is improved. By automatically parsing the demand type data and emergency level in the call audio with intelligent speech recognition and semantic analysis technology, accurate allocation and priority processing of resources are achieved, ensuring that key needs are responded to in a timely manner. By generating detailed intercom scheduling instructions based on the multivariate data of the comprehensive target object, comprehensive information support is provided for scheduling decisions, and smooth cross-department collaboration is promoted. By quickly matching and allocating at least one target response team with an intelligent scheduling algorithm, resource allocation is optimized to ensure efficient use of rescue forces. Through the response personnel determination module, the most suitable response personnel are accurately selected based on the real-time team availability and load data, avoiding resource waste and ensuring response speed and quality. By quickly conveying the intercom scheduling instructions to the intercom equipment of the target response personnel with instant messaging technology, information transmission delay is reduced and response agility is improved. By using the microphone matrix as the medium to play the responders' response audio, the clear transmission and wide coverage of information can be ensured, thus enhancing the communication effect.

[0089] The implementation methods of each step in the medical intercom method based on audio processing provided in the embodiment of the present invention have been described in detail in any of the above-mentioned system embodiments, so they will not be repeated here.

[0090] The audio processing-based medical intercom system in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

Claims

1. A medical intercom system based on audio processing, characterized in that: The system comprises: An audio collection module is used to collect the call audio of the target object using a microphone matrix arranged in the department; An audio positioning module, used to determine the identity data and location data of the target object based on the call audio; A demand determination module, used to determine the demand type data and demand urgency level data of the target object based on the call audio; An instruction generation module, used for forming an intercom dispatch instruction based on the identity data, location data, demand type data and demand emergency level data of the target object; An instruction distribution module, used for determining at least one target response team based on the intercom dispatch instruction; A response personnel determination module, used to obtain the availability data and load data of the target response team, and determine the target response personnel based on the intercom scheduling instruction, the availability data and the load data; An instruction sending module, used for sending the intercom dispatch instruction to the intercom device of the target responder; A response processing module, used for receiving the response audio sent by the target responder through the intercom device, and playing the response audio using the microphone matrix; The requirements determination module includes: a demand type determination unit, configured to extract a call text of the call audio, identify key entities in the call text using pre-trained named entities, match the key entities with a predefined demand type dictionary, and obtain demand type data, wherein the demand type data includes first-level demands and second-level demands, wherein the first-level demands include medical demands and non-medical demands, and the second-level demands are subordinate demands of the first-level demands; The sentiment scoring unit is used to calculate the sentiment score based on the call text, and the calculation formula is as follows: ； Wherein, Sentiment Score represents the sentiment score of the call text, n is the total number of words in the call text, word_score i is the sentiment score of the i-th word. The sentiment score of the word is obtained by querying the sentiment dictionary. i is the weight of the i-th word. The weight of the word is calculated based on word frequency-inverse document frequency; An urgency scoring unit, configured to locate the emergency words in the call text based on the call text and a predefined emergency vocabulary list, and calculate the urgency score of each emergency word; The demand urgency level evaluation unit is used to determine the demand urgency level data based on the emotion score and the urgency score, and the calculation formula is as follows: ； Among them, Emergency Level represents the emergency level data, γ and δ are weight coefficients used to adjust the relative importance of sentiment score and emergency word score in the emergency level calculation, Sentiment Score is the sentiment score, which is a continuous value, m is the number of emergency words, urgency_score j is the urgency score of the jth urgent word.

2. The system according to claim 1, characterized in that The medical intercom system based on audio processing further includes an instruction display module, which includes a first display unit and a second display unit, wherein: A first display unit, used to display all intercom scheduling instructions in a preset area and processing progress data of each intercom scheduling instruction using a general display device; The second display unit is used to use multiple sub-display devices to respectively display the intercom scheduling instructions sent to each response team and the processing progress data of each intercom scheduling instruction, and the sub-display devices correspond to the response teams one by one.

3. The system according to claim 1, characterized in that The audio positioning module includes: The preprocessing unit is used to perform filtering and denoising on the call audio. The denoising formula is as follows: ； Wherein, Y(k) represents the amplitude spectrum estimation of the denoised call audio, X(k) represents the amplitude spectrum of the call audio in the frequency domain, k represents the frequency index, that is, different frequency components, |X(k)| represents the absolute value of X(k), N(k) represents the estimation of the amplitude spectrum of the call audio in the frequency domain, α represents the over-subtraction factor, α is a real number less than or equal to 1, when α=1, it means that the noise is completely subtracted; when α<1, it means that the noise is partially subtracted; A voiceprint extraction unit, used to extract the voiceprint of the preprocessed call audio to obtain a voiceprint to be matched; A voiceprint matching unit, used to match the voiceprint to be matched with the pre-stored voiceprints corresponding to each grid point in the pre-stored voiceprint map to obtain a matching result; A confirmation unit is used to determine the identity data and location data of the target object based on the matching result, the identity data including the pre-stored case data of the target object, and the location data including the pre-stored bed location and current call location of the target object.

4. The system according to claim 1, characterized in that Instruction distribution module, including: A first allocation unit is used for, when it is detected that the first-level demand in the intercom scheduling instruction is a medical demand, determining the department to which the target object currently belongs based on the intercom scheduling instruction, and determining the medical demand response team based on the second-level demand in the intercom scheduling instruction, and taking the response team corresponding to the currently affiliated department and the medical demand response team as the target response team; The second allocation unit is used to determine a target response team based on a second level demand in the intercom dispatch instruction when it is detected that the first level demand in the intercom dispatch instruction is a non-medical demand.

5. The system according to claim 1, characterized in that The responder identification module includes: A priority calculation unit, used to obtain all unfinished intercom scheduling instructions in the target response team, calculate the priority data to be responded to of each unfinished intercom scheduling instruction, and obtain a set of priority data to be responded to; The demand type statistical unit is used to obtain the pending response demand type data of each unfinished intercom scheduling instruction, and obtain a pending response demand type data set; A determination unit is used to determine the target response personnel based on the pending response priority data set, the pending response requirement type data set, the availability data and the load data, wherein the availability data includes the skill data of each response personnel in the target response team, and the load data includes the number of target objects allocated to each response personnel in the target response team as the target response personnel.

6. The system according to claim 4, characterized in that Medical intercom system based on audio processing also includes: The soothing module is used to generate soothing audio based on the intercom scheduling instruction and the emotion score when no response audio emitted by the target responder through the intercom device is detected within a preset time, and to play the soothing audio using the microphone matrix.

7. A medical intercom method based on audio processing, characterized in that: The method comprises: The microphone matrix installed in the department is used to collect the target object's call audio; determining identity data and location data of the target object based on the call audio; Determining the demand type data and demand urgency level data of the target object based on the call audio; Composing an intercom dispatch instruction based on the identity data, location data, demand type data and demand emergency level data of the target object; Determine at least one target response team based on the intercom dispatch instruction; Acquire the availability data and load data of the target response team, and determine the target response personnel based on the intercom dispatch instruction, the availability data and the load data; Sending the intercom dispatch instruction to the intercom device of the target responder; Receiving a response audio signal from the target responder through the intercom device, and playing the response audio signal using the microphone matrix; The determining of the demand type data and demand urgency level data of the target object based on the call audio includes: Extracting a call text of the call audio, using pre-trained named entities to identify key entities in the call text, matching the key entities with a predefined demand type dictionary, and obtaining demand type data, wherein the demand type data includes first-level demands and second-level demands, wherein the first-level demands include medical demands and non-medical demands, and the second-level demands are subordinate demands of the first-level demands; The sentiment score is calculated based on the call text, and the calculation formula is as follows: ； Wherein, Sentiment Score represents the sentiment score of the call text, n is the total number of words in the call text, word_score i is the sentiment score of the i-th word. The sentiment score of the word is obtained by querying the sentiment dictionary. i is the weight of the i-th word. The weight of the word is calculated based on word frequency-inverse document frequency; Based on the call text and a predefined emergency vocabulary list, the emergency words in the call text are located, and the urgency score of each emergency word is calculated; The demand urgency level data is determined based on the sentiment score and the urgency score, and the calculation formula is as follows: ； Among them, Emergency Level represents the emergency level data, γ and δ are weight coefficients used to adjust the relative importance of sentiment score and emergency word score in the emergency level calculation, Sentiment Score is the sentiment score, which is a continuous value, m is the number of emergency words, urgency_score j is the urgency score of the jth urgent word.

Images