Method and device for warning dangerous events of elderly living alone based on household Internet of Things controller

Abstract

The embodiment of the invention discloses a method and a device for alarming dangerous events of elderly people living alone based on a home Internet of Things controller. A specific embodiment of the method comprises the following steps: performing human shape detection on each video frame image in a video frame image sequence; in response to determining that the human shape detection result meeting the first preset condition exists in the human shape detection result sequence, determining the video frame image corresponding to the human shape detection result meeting the first preset condition as an old person image; single old man image extraction is carried out on each old man image in the old man image group to generate single old man images; performing tumble detection on each single old man image in the single old man image group to generate a tumble detection result; controlling audio acquisition equipment to acquire audio of the old in the target area in real time; performing audio recognition on the audio of the old; and in response to determining that the audio recognition text represents the distress call text, initiating a voice request sound to the target rescue platform. According to the embodiment, the accuracy and timeliness of fall detection of the old people are improved.

Description

Technical Field

[0001] The embodiments of the present disclosure relate to the field of computers, and in particular to a method and device for warning of dangerous events for elderly people living alone based on a home Internet of Things controller. Background Art

[0002] Dangerous events for the elderly may refer to elderly people falling down or strangers breaking in. At present, the commonly used method for warning of dangerous events for the elderly is: the camera monitors the family room / ward in real time, and the monitoring personnel observe the video collected by the camera in real time to determine whether there is a dangerous event for the elderly.

[0003] However, the above-mentioned method of detecting dangerous events for the elderly usually has the following technical problems: the monitoring personnel observe the video collected by the camera in real time, which is inefficient and cannot warn of dangerous events for the elderly in time, which may easily cause safety hazards.

[0004] The above information disclosed in this Background section is only for enhancement of understanding of the background of the inventive concept and therefore it may contain information that does not form the prior art that is already known in this country to a person of ordinary skill in the art. Summary of the invention

[0005] The content of this disclosure is used to introduce concepts in a brief form, which will be described in detail in the detailed implementation section below. The content of this disclosure is not intended to identify the key features or essential features of the technical solution claimed for protection, nor is it intended to limit the scope of the technical solution claimed for protection.

[0006] Some embodiments of the present disclosure propose a dangerous event alarm method, device, electronic device and computer-readable medium for elderly people living alone based on a home Internet of Things controller to solve one or more of the technical problems mentioned in the above background technology section.

[0007] In a first aspect, some embodiments of the present disclosure provide a method for warning of dangerous events for elderly people living alone based on a home Internet of Things controller, the method comprising: in response to receiving a dangerous event detection instruction for elderly people living alone, controlling an associated camera device to collect monitoring video of a target area in real time; performing frame processing on the monitoring video to obtain a video frame image sequence; performing human shape detection on each video frame image in the video frame image sequence to generate a human shape detection result to obtain a human shape detection result sequence; in response to determining that a human shape detection result that meets a first preset condition exists in the human shape detection result sequence, determining the video frame image corresponding to the human shape detection result that meets the first preset condition as an elderly person image, and obtaining an elderly person image. image group; extracting a single elderly image from each elderly image in the elderly image group to generate a single elderly image and obtain a single elderly image group; performing fall detection on each single elderly image in the single elderly image group to generate a fall detection result and obtain a fall detection result group; in response to determining that there is a fall detection result that meets the second preset condition in the fall detection result group, controlling the associated audio acquisition device to collect the elderly audio in the target area in real time; performing audio recognition on the elderly audio to obtain an audio recognition text; in response to determining that the audio recognition text represents a call for help text, initiating a voice request to the target rescue platform, and controlling the associated alarm device to emit an alarm prompt sound.

[0008] In a second aspect, some embodiments of the present disclosure provide a dangerous event alarm device for elderly people living alone based on a home Internet of Things controller, the device comprising: a first control unit, configured to control an associated camera device to collect monitoring video of a target area in real time in response to receiving a dangerous event detection instruction for elderly people living alone; a framing unit, configured to perform frame processing on the monitoring video to obtain a video frame image sequence; a detection unit, configured to perform human shape detection on each video frame image in the video frame image sequence to generate a human shape detection result to obtain a human shape detection result sequence; a determination unit, configured to determine the video frame image corresponding to the human shape detection result that meets the first preset condition as an elderly image in response to determining that there is a human shape detection result that meets a first preset condition in the human shape detection result sequence to obtain an elderly image group; and extract a single The first control unit is configured to extract a single elderly image from each elderly image in the above-mentioned elderly image group to generate a single elderly image and obtain a single elderly image group; the second control unit is configured to perform fall detection on each single elderly image in the above-mentioned single elderly image group to generate a fall detection result and obtain a fall detection result group; the second control unit is configured to control the associated audio acquisition device to collect the elderly audio in the target area in real time in response to determining that there is a fall detection result that meets the second preset condition in the above-mentioned fall detection result group; the recognition unit is configured to perform audio recognition on the above-mentioned elderly audio to obtain an audio recognition text; the third control unit is configured to initiate a voice request to the target rescue platform in response to determining that the above-mentioned audio recognition text represents a call for help text, and control the associated alarm device to emit an alarm prompt sound.

[0009] In a third aspect, some embodiments of the present disclosure provide an electronic device comprising: one or more processors; a storage device on which one or more programs are stored, and when the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any implementation manner of the above-mentioned first aspect.

[0010] In a fourth aspect, some embodiments of the present disclosure provide a computer-readable medium having a computer program stored thereon, wherein when the program is executed by a processor, the method described in any implementation manner of the above-mentioned first aspect is implemented.

[0011] The above-mentioned embodiments of the present disclosure have the following beneficial effects: through the dangerous event alarm method for elderly people living alone based on the home Internet of Things controller of some embodiments of the present disclosure, the accuracy and timeliness of the alarm for dangerous events for the elderly are improved. Specifically, the reason for the easy safety hazard is that the monitoring personnel observe the video collected by the camera in real time, which is inefficient and cannot timely warn the elderly of dangerous events. Based on this, in some embodiments of the present disclosure, the dangerous event alarm method for elderly people living alone based on the home Internet of Things controller, first, in response to receiving a dangerous event detection instruction, controls the associated camera device to collect the monitoring video of the target area in real time. Secondly, the above-mentioned monitoring video is frame-processed to obtain a video frame image sequence; each video frame image in the above-mentioned video frame image sequence is subjected to human shape detection to generate a human shape detection result, and obtains a human shape detection result sequence. In this way, it is convenient to determine whether there is an elderly person in the area, and if there is an elderly person, a fall detection is performed. Then, in response to determining that there is a human shape detection result that meets the first preset condition in the above-mentioned human shape detection result sequence, the video frame image corresponding to the human shape detection result that meets the first preset condition is determined as an elderly image, and an elderly image group is obtained. In this way, the video frame image with a human shape image can be taken out. Afterwards, a single elderly image is extracted from each elderly image in the above-mentioned elderly image group to generate a single elderly image and obtain a single elderly image group; a fall detection is performed on each single elderly image in the above-mentioned single elderly image group to generate a fall detection result and obtain a fall detection result group. Thus, it is possible to detect whether the elderly have fallen through the video image, thereby improving the accuracy and timeliness of the fall detection. Then, in response to determining that there is a fall detection result that satisfies the second preset condition in the above-mentioned fall detection result group, the associated audio acquisition device is controlled to collect the elderly audio in the target area in real time; the above-mentioned elderly audio is audio recognized to obtain an audio recognition text. Thus, when the elderly fall, the elderly audio can be collected to determine whether the elderly are in urgent need of help. Finally, in response to determining that the above-mentioned audio recognition text represents a call for help text, a voice request is initiated to the target rescue platform, and the associated alarm device is controlled to emit an alarm prompt tone. Thus, a request can be initiated to the rescue platform in a timely manner to rescue the elderly in a timely manner. Thus, the accuracy and timeliness of the alarm for dangerous events for the elderly are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the accompanying drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and that components and elements are not necessarily drawn to scale.

[0013] Figure 1 It is a flowchart of some embodiments of the method for alarming dangerous events for elderly people living alone based on a home IoT controller according to the present disclosure;

[0014] Figure 2 It is a structural schematic diagram of some embodiments of the dangerous event alarm device for elderly people living alone based on the home Internet of Things controller according to the present disclosure;

[0015] Figure 3 is a schematic diagram of the structure of an electronic device suitable for implementing some embodiments of the present disclosure;

[0016] Figure 4 It is a schematic diagram of the architecture of a home Internet of Things according to the method for warning of dangerous events for elderly people living alone based on a home Internet of Things controller disclosed in the present invention;

[0017] Figure 5 This is a schematic internal test operation page screenshot about emergency contacts in the dangerous event alarm method for elderly people living alone based on a home IoT controller according to the present disclosure;

[0018] Figure 6 It is a schematic internal device test operation page screenshot of the alarm device in the dangerous event alarm method for elderly people living alone based on the home Internet of Things controller according to the present disclosure;

[0019] Figure 7 This is a screenshot of a schematic function setting test operation page regarding human detection in the dangerous event alarm method for elderly people living alone based on a home Internet of Things controller according to the present invention. DETAILED DESCRIPTION

[0020] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments set forth herein. On the contrary, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0021] It should also be noted that, for ease of description, only the parts related to the invention are shown in the drawings. In the absence of conflict, the embodiments and features in the embodiments of the present disclosure can be combined with each other.

[0022] It should be noted that the concepts such as "first" and "second" mentioned in the present disclosure are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0023] It should be noted that the modifications of "one" and "plurality" mentioned in the present disclosure are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0024] The names of the messages or information exchanged between multiple devices in the embodiments of the present disclosure are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0025] The present disclosure will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0026] Figure 1 The flowchart of some embodiments of the method for alarming dangerous events for elderly people living alone according to the present disclosure is shown. The flowchart of some embodiments of the method for alarming dangerous events for elderly people living alone based on the home Internet of Things controller according to the present disclosure is shown. The method for alarming dangerous events for elderly people living alone based on the home Internet of Things controller comprises the following steps:

[0027] Step 101, in response to receiving a dangerous event detection instruction, controlling the associated camera device to collect monitoring video of the target area in real time.

[0028] In some embodiments, the execution subject (e.g., a computing device) of the method for warning of dangerous events for elderly people living alone based on a home IoT controller can control the associated camera device to collect monitoring videos of the target area in real time in response to receiving a dangerous event detection instruction for elderly people living alone. The target area may refer to an area that needs to be monitored for falls of the elderly. For example, the target area may refer to a room / living room where an elderly person living alone lives, or a public place. For example, monitoring videos of the target area within a preset time period can be collected in real time. The duration of the preset time period may be 5-10 minutes. The preset time period should not be too long so as to detect whether an elderly person has fallen in time. The dangerous event detection instruction for elderly people living alone may refer to an instruction to detect whether there is a dangerous event in the target area. Dangerous events may include, but are not limited to: elderly fall events, stranger intrusion events.

[0029] Figure 4 This is a schematic diagram of the architecture of the home Internet of Things according to the method for alarming dangerous events for elderly people living alone based on the home Internet of Things controller disclosed in the present invention, including: audio acquisition equipment, alarm equipment, control panel, contact terminal, Ethernet, buttons, camera (camera device), infrared lamp, white light lamp, home Internet of Things controller. Among them, the audio acquisition equipment and the alarm equipment are used for voice input and voice output. The control panel can be the operation panel of the home Internet of Things controller. The interconnection of various devices can be achieved through Ethernet and WIFI. The buttons can control the opening and closing of the audio acquisition equipment and the alarm equipment; the camera can collect video in real time; the infrared lamp and the white light lamp can help the camera to achieve night shooting.

[0030] Figure 6 It is a schematic screenshot of the internal device test operation page for the alarm device in the dangerous event alarm method for elderly people living alone based on the home Internet of Things controller according to the present invention, showing the operation page for the alarm setting of the alarm device.

[0031] Step 102, performing frame processing on the above surveillance video to obtain a video frame image sequence.

[0032] In some embodiments, the execution subject may perform frame processing on the surveillance video to obtain a video frame image sequence. For example, the surveillance video may be segmented frame by frame to obtain a video frame image sequence.

[0033] Step 103 , performing human shape detection on each video frame image in the above video frame image sequence to generate a human shape detection result, thereby obtaining a human shape detection result sequence.

[0034] In some embodiments, the execution subject may perform human shape detection on each video frame image in the video frame image sequence to generate a human shape detection result and obtain a sequence of human shape detection results. The human shape detection result may indicate whether there is an old man figure in the video frame image. For example, the human shape detection model may be a human key point detection model, a face recognition model, or a three-dimensional human posture recognition model. Human key point detection model: This type of model is mainly used to detect key points on the human body, such as joint positions, to achieve recognition and understanding of human postures. For example, the feature pyramid attention hourglass model has achieved the best detection effect on the MPII and LSP human key point detection data sets, and its effectiveness has been verified by improved comparative experiments. Face recognition model: The face recognition model is mainly used to detect and identify facial features, including detection, alignment, feature extraction, and classification steps. For example, the FaceNet model redefines a face recognition framework based on Triplet Loss by using Triplet Loss. The framework does not require an additional fully connected layer to perform classification tasks, but directly extracts facial features for similarity calculation. 3D human posture recognition model: This type of model is mainly used to predict the 3D coordinate position and angle of human joints, and achieves a more accurate description of human posture by adding depth information. For example, the 3DPose model realizes real-time 3D posture recognition of the human body.

[0035] For example, Figure 7 This is a screenshot of a schematic function setting test operation page for human detection in the dangerous event alarm method for elderly people living alone based on a home Internet of Things controller according to the present invention, showing the operation page for turning on the human function.

[0036] Step 104, in response to determining that there is a human shape detection result that meets the first preset condition in the human shape detection result sequence, the video frame image corresponding to the human shape detection result that meets the first preset condition is determined as an old person image, to obtain an old person image group.

[0037] In some embodiments, the execution subject may determine the video frame image corresponding to the human shape detection result satisfying the first preset condition as an old person image in response to determining that there is a human shape detection result satisfying the first preset condition in the human shape detection result sequence, and obtain an old person image group. The first preset condition may refer to: the human shape detection result indicates that there is a human shape old person in the video frame image.

[0038] Step 105 , extracting a single elderly image from each elderly image in the elderly image group to generate a single elderly image and obtain a single elderly image group.

[0039] In some embodiments, the above-mentioned execution subject can perform a single elderly image extraction on each elderly image in the above-mentioned elderly image group to generate a single elderly image and obtain a single elderly image group. Single elderly image extraction can refer to: extracting a human-shaped elderly image from the elderly image. For example, a single elderly image extraction can be performed on each elderly image through an image target extraction algorithm to generate a single elderly image. The image target extraction algorithm may include but is not limited to: LBP algorithm (Local Binary Patterns, local binary patterns), edge detection operator, feature extraction based on parameter measurement, feature extraction based on non-parametric measurement, feature extraction based on Karhumen-Loeve expansion, wavelet transform feature extraction and other methods. For another example, the outline of the human-shaped image in the elderly image can be marked first, and then the human-shaped image in the outline can be extracted.

[0040] Step 106: Perform fall detection on each single elderly person image in the single elderly person image group to generate a fall detection result and obtain a fall detection result group.

[0041] In some embodiments, the execution subject may perform fall detection on each single elderly image in the single elderly image group to generate a fall detection result and obtain a fall detection result group. For example, a pre-trained fall detection model may be used to perform fall detection on each single elderly image in the single elderly image group to generate a fall detection result. For example, the pre-trained fall detection model may be an SVM support vector machine model. The pre-trained fall detection model may include: a first-level SVM classifier and a second-level SVM classifier. The fall detection result may indicate whether the state of the elderly in the single elderly image is a fall state.

[0042] The first-level SVM classifier is used to determine whether the moving object is in a non-upright state (squatting, falling, etc.). The features such as the aspect ratio of the object, the minimum rectangular frame area containing the object, the minimum rectangular frame perimeter of the object, and the height of the moving object are extracted for classifier training and classification judgment. For image frames judged as non-upright by the first-level classifier, they are sent to the second-level SVM classifier for classification judgment. The second-level SVM classifier is used to distinguish whether the moving object is in a falling or other non-upright state. The features such as the Zernike moment feature, the height of the moving object, the width of the moving object, the contour area of ​​the moving object, and the contour perimeter of the moving object are extracted for classifier training and classification judgment.

[0043] In practice, the execution subject may perform fall detection on each single elderly person image in the single elderly person image group through the following steps to generate a fall detection result:

[0044] In the first step, the above-mentioned single elderly image is input into the pre-trained elderly upright state recognition model to generate the elderly upright state recognition result. For example, the elderly upright state recognition model can be an SVM (Support Vector Machine) classifier, which is used to determine whether the moving object (elderly) is in a non-upright (squatting, falling, etc.) state. For another example, the elderly upright state recognition model can be a MoveNet model, a PoseNet model, a DCPose model, and a DensePose model. Among them, MoveNet was developed by researchers and built using TensorFlow.js. It has super fast and highly accurate characteristics and can detect 17 key points of the human body. It has two versions: the Lightning version is for applications with low latency requirements, and the Thunder version is designed for applications that focus on achieving higher accuracy. PoseNet is a deep learning model that uses TensorFlow to detect different body parts and provides comprehensive skeletal information by adding other key points. It can detect postures in real time and can effectively perform single and multi-posture detection of the human body.

[0045] In the second step, in response to determining that the above-mentioned elderly upright state recognition result meets the first detection condition, the above-mentioned single elderly image is input into the pre-trained elderly fall recognition model to generate the elderly fall recognition result. The first detection condition may refer to: the elderly upright state recognition result represents the elderly non-upright state.

[0046] The above-mentioned elderly upright state recognition model can be trained by the following steps:

[0047] The first step is to obtain a training sample set of the upright state of the elderly, wherein the training samples of the upright state of the elderly in the training sample set of the upright state of the elderly include: sample elderly behavior images and sample upright state labels.

[0048] The second step is to determine the initial upright state recognition model for the elderly. The initial upright state recognition model for the elderly includes: an initial feature extraction network and an initial upright state recognition network. The initial feature extraction network includes: an initial convolution model sequence, and the initial convolution model in the initial convolution model sequence includes: an initial subconvolution model and an initial activation model.

[0049] Here, the input of the initial convolution sub-model included in the initial convolution model in the initial convolution model sequence may be: the output of the initial activation model included in the previous initial convolution model. The input of the initial convolution sub-model included in the first initial convolution model in the initial convolution model sequence may be: a sample elderly behavior image. The input of the initial activation model of the initial convolution model in the initial convolution model sequence may be: the output of the initial convolution sub-model included in the initial convolution model. For example, the number of initial convolution models in the initial convolution model sequence may be 22.

[0050] The initial upright state recognition network may be a model that takes the initial elderly behavior image features as input and takes the initial upright state detection results as output. The initial upright state recognition network is used to: filter and process the initial elderly behavior image features through a preset filtering algorithm to generate an initial upright behavior detection result. For example, the preset filtering algorithm may be: NMS (Non Maximum Suppression) algorithm.

[0051] The third step is to select a target elderly upright state training sample from the elderly upright state training sample set. The target elderly upright state training sample can be randomly selected from the elderly upright state training sample set.

[0052] The fourth step is to input the sample elderly behavior images included in the above-mentioned target elderly upright state training samples into the initial feature extraction network to obtain the initial elderly behavior image features.

[0053] The fifth step is to input the initial elderly behavior image features into the initial upright state recognition network to obtain the initial upright state detection results.

[0054] Step 6: Based on a preset loss function, determine the loss value between the initial upright state detection result and the corresponding sample upright state label. For example, the preset loss function may be a hinge loss function or a cross entropy loss function.

[0055] In the seventh step, in response to determining that the loss value is less than or equal to the preset difference value, the initial elderly upright state recognition model is determined as the trained elderly upright state recognition model.

[0056] Therefore, it is possible to detect relatively accurate information about the upright state of the elderly based on a relatively accurate model for recognizing the upright state of the elderly, so as to subsequently recognize whether the elderly have fallen. In addition, by recognizing the upright state of the elderly, it is possible to determine whether the elderly are squatting or bending over, so as to avoid abnormal warnings caused by directly judging that the elderly are in a state of falling.

[0057] Step 107: in response to determining that there is a fall detection result that meets the second preset condition in the fall detection result group, controlling the associated audio collection device to collect audio of the elderly in the target area in real time.

[0058] In some embodiments, the execution subject may control the associated audio acquisition device to collect the audio of the elderly in the target area in real time in response to determining that there is a fall detection result that satisfies the second preset condition in the fall detection result group. The second preset condition may be: the fall detection result indicates that the elderly has fallen. The audio acquisition device may be a device for collecting sound. For example, the audio acquisition device may be a network audio collector, a high-fidelity sound collector, or a camera-recorded audio collector.

[0059] Step 108, performing audio recognition on the elderly audio to obtain an audio recognition text.

[0060] In some embodiments, the above-mentioned execution subject can perform audio recognition on the above-mentioned elderly audio to obtain audio recognition text. For example, the above-mentioned elderly audio can be audio recognized by an audio recognition algorithm to obtain an audio recognition text. For example, the audio recognition algorithm may include: an algorithm based on dynamic time warping (DTW), a method based on a hidden Markov model (HMM) of a parameter model, and a method based on a vector quantization (VQ) of a non-parametric model. For another example, acoustic features include linear predictive coding (LPC), Mel-frequency cepstrum coefficients (MFCC), Mel-scale filter banks (FBank), etc.

[0061] In practice, the above-mentioned execution entity can perform audio recognition on the above-mentioned elderly audio through the following steps:

[0062] The first step is to preprocess the above-mentioned elderly audio to obtain the elderly audio filter signal.

[0063] The first step may include the following sub-steps:

[0064] The first sub-step is to perform pre-emphasis processing on the elderly audio to obtain an elderly audio emphasis signal. For example, the elderly audio can be pre-emphasized by a high-pass filter algorithm to obtain an elderly audio emphasis signal.

[0065] It should be noted that the sound an elderly person may make when falling down is relatively small. Therefore, it is necessary to amplify the elderly person's audio to enhance the signal strength of the elderly person's audio, thereby ensuring the accuracy of the elderly person's audio (cry for help) recognition.

[0066] The second sub-step is to perform frame processing on the above-mentioned elderly audio emphasis signal to obtain the elderly audio frame signal. The above-mentioned elderly audio emphasis signal can be frame processed according to a preset time interval to obtain the elderly audio frame signal.

[0067] The third sub-step is to perform windowing processing on the elderly audio signal to obtain an elderly audio window signal. For example, the elderly audio signal can be windowed by using a Hamming window function to obtain an elderly audio window signal.

[0068] The fourth sub-step is to transform the above-mentioned elderly audio window signal to obtain the elderly audio spectrum signal. For example, the above-mentioned elderly audio window signal can be transformed by a Fourier transform algorithm to obtain the elderly audio spectrum signal.

[0069] The fifth sub-step is to filter the above-mentioned elderly audio spectrum signal to obtain the elderly audio filter signal. For example, the above-mentioned elderly audio spectrum signal can be filtered through a filter to obtain the elderly audio filter signal.

[0070] In the second step, the above-mentioned elderly audio filter signal is input into the pre-trained speech recognition model to obtain the audio recognition text. Among them, the above-mentioned pre-trained speech recognition model includes: a classification network, an encoding network, an enhancement network and a decoding network. The classification network may include: a deep neural network (DNN), a recurrent neural network (RNN), a long short-term memory recurrent neural network (LSTM-RNN), a full sequence convolutional neural network (DFCNN), a feedforward sequence memory neural network (FSMN), and a deep separable convolutional neural network (DFSMN). The encoding network may include: a recurrent neural network (RNN), a long short-term memory network (LSTM), and a bidirectional long short-term memory network (BLSTM-RNN). The enhancement network may be: a PEL (PresuppositionEvidence Logic, enhanced prompt) model. The decoding network may be: a hidden Markov model (HMM) and a neural network-based model, such as a recurrent neural network (RNN) and a long short-term memory network (LSTM).

[0071] The pre-trained speech recognition model may be trained through the following sub-steps:

[0072] The first sub-step is to obtain a sample speech signal set and a sample speech text set, wherein the sample speech signals in the sample speech signal set correspond to the sample speech texts in the sample speech text set.

[0073] In the second sub-step, a sample speech signal is selected from the sample speech signal set as a target sample speech signal, and the following recognition training steps are performed:

[0074] 1. Input the target sample speech signal into the initial classification network included in the initial speech recognition model to obtain initial speech classification information. The initial speech recognition model also includes: an initial encoding network, an initial enhancement network and an initial decoding network.

[0075] 2. Input the initial speech classification information into the initial coding network included in the initial speech recognition model to obtain initial speech coding information.

[0076] 3. Input the initial speech coding information into the initial enhancement network included in the initial speech recognition model to obtain initial speech enhancement information.

[0077] It should be emphasized that classifying and enhancing voice signals can improve the accuracy of voice recognition for the elderly. Through voice classification, it can be determined whether the elderly are calling for help; through voice enhancement, the accuracy of voice recognition for the elderly can be improved when the elderly's voice is relatively low.

[0078] 4. Input the initial speech enhancement information into the initial decoding network included in the initial speech recognition model to obtain the initial speech text.

[0079] 5. Based on a preset loss function, determine the loss value of the initial speech text and the corresponding sample speech text. For example, the preset loss function may be a hinge loss function or a cross entropy loss function.

[0080] 6. In response to determining that the loss value is less than a preset loss value, the initial speech recognition model is determined as the trained speech recognition model.

[0081] Step 109 , in response to determining that the audio recognition text represents a distress call text, a voice request is initiated to a target rescue platform, and an associated alarm device is controlled to emit an alarm prompt tone.

[0082] In some embodiments, the above-mentioned execution subject may initiate a voice request to the target rescue platform in response to determining that the above-mentioned audio recognition text represents a call for help text, and control the associated alarm device to emit an alarm prompt tone. The call for help text may represent a text for calling for help. For example, the call for help text may include descriptions such as "help me" and "help my life". The target rescue platform may be a pre-set system platform for rescuing the elderly. For example, the target rescue platform may be a call for help platform for medical staff, or a call for help platform for emergency personnel in the target area. The voice request may refer to a request to talk to the on-duty personnel of the target rescue platform. The alarm device may be a voice prompt device. For example, the alarm device may be a speaker. The alarm prompt tone may be a prompt tone that indicates that a dangerous event exists in the target area.

[0083] Optionally, in response to determining that the audio recognition text represents a call for help text, it is identified whether there is a single elderly image representing a frontal facial image of the elderly in the single elderly image group. For example, a face recognition model can be used to identify whether there is a single elderly image representing a frontal facial image of the elderly in the single elderly image group. For example, the face recognition model can be a MTCNN (Multi-task Cascaded Convolutional Networks) model, a CenterFace model, or a RetinaFace model.

[0084] Optionally, in response to determining that there is a single elderly image representing a frontal face image of an elderly person in the single elderly image group, for each single elderly image representing a frontal face image of an elderly person, the following processing steps are performed:

[0085] The first step is to capture the face image in the single elderly image. For example, the face image in the single elderly image can be cut.

[0086] In the second step, the above face image is corrected to obtain a face correction image. For example, the above face image can be corrected by a face image correction algorithm to obtain a face correction image. For example, the face image correction algorithm may include multiple steps such as face detection, face alignment, and automatic calibration of scale, position and posture changes. Face detection and positioning: This is the first step of face correction. By using libraries such as MediaPipe, the positioning of the face and the positioning of the facial features in the face can be easily solved. The mp_face_detection.FaceDetection function in the MediaPipe library is used to detect faces in images or videos and return the coordinate information of the face, including the face position and the attributes of the bounding box, such as relative position and confidence score. This information is crucial for subsequent face alignment and correction. Face alignment and correction: After face detection, the face needs to be aligned and corrected to solve problems caused by different camera angles and different human movements. This includes calibration of scale, position and posture changes to ensure that the face image meets the optimal state for feature extraction. The matching method is an algorithm that extracts a face posture calibration image from a face image, and extracts the positions of the two eyes from it to achieve automatic calibration of the scale, position and posture changes of the face. Gamma correction: In addition, lighting adjustment for face images is also part of the correction. Gamma correction is an image processing technology used to adjust the brightness and contrast of an image and improve the visual effect of the image. In face processing, gamma correction can be used to adjust the lighting conditions of the face image, improve image quality, and make subsequent feature extraction and recognition more accurate. The third step is to perform compensation processing on the above-mentioned face correction image to obtain a compensated face image. For example, compensation processing may include but is not limited to: light compensation (uniform brightness adjustment), exposure compensation, and phase aberration compensation.

[0087] It should be noted that due to the uneven light distribution in the environment of elderly people living alone (for example, only one main light is on), the exposure of the collected facial images is insufficient; therefore, when identifying whether a stranger has intruded, the collected facial images need to be corrected and exposure compensated to ensure the clarity of the facial images and improve the accuracy of facial image recognition.

[0088] In the fourth step, the compensated face image is input into a pre-trained face recognition model to obtain a face recognition result. For example, the face recognition model can be a MTCNN (Multi-task Cascaded Convolutional Networks) model and a face online recognition system based on a convolutional neural network (CNN).

[0089] Optionally, deduplication processing is performed on each face recognition result to obtain a target face recognition result group. That is, deduplication processing is performed on each face recognition result representing the same face to obtain a target face recognition result group.

[0090] Optionally, for each target face recognition result in the above target face recognition result group, the following determination steps are performed:

[0091] The first step is to determine whether the face data corresponding to the target face recognition result is stored in the preset face database. The face data may include face features, face images, the name of the elderly, and emergency contact information. That is, it can be determined whether the face database stores face data with the same face name as the target face recognition result.

[0092] For example, Figure 5 This is a schematic screenshot of the internal test operation page regarding emergency contacts in the dangerous event alarm method for elderly people living alone based on the home Internet of Things controller disclosed in the present invention, showing the setting and operation page of the emergency contact information.

[0093] The second step is to call the terminal of the emergency contact corresponding to the face data in response to determining that the face database stores the face data corresponding to the target face recognition result. The terminal of the emergency contact may refer to the user's mobile phone terminal.

[0094] The third step is, in response to determining that the face database does not store face data corresponding to the target face recognition result, determining the user corresponding to the target face recognition result as an abnormal user, and controlling the associated alarm device to perform alarm processing.

[0095] Optionally, in response to determining that there is no human figure detection result satisfying a first preset condition in the human figure detection result sequence, the video acquisition time is determined.

[0096] In some embodiments, the execution subject may determine the video capture time in response to determining that there is no human figure detection result satisfying the first preset condition in the human figure detection result sequence.

[0097] Optionally, in response to determining that the duration between the current time and the video acquisition time is greater than or equal to a preset duration, the associated alarm device is controlled to perform alarm processing, and voice alarm information is sent to the terminal of each preset emergency contact.

[0098] In some embodiments, the execution subject may control the associated alarm device to perform alarm processing and send a voice alarm message to each preset emergency contact terminal in response to determining that the duration between the current time and the video acquisition time is greater than or equal to the preset duration. The voice alarm message may indicate that the elderly person has not appeared for a long time and needs to be inspected. The elderly person may refer to an elderly person / patient who should appear in the target area.

[0099] Optionally, in response to determining that there is a human figure detection result that meets the first preset condition in the human figure detection result sequence, the current position information of each elderly person in the surveillance video is identified to obtain a position information group.

[0100] In some embodiments, the above-mentioned execution subject can identify the current position information of each elderly person in the above-mentioned surveillance video in response to determining that there is a human shape detection result that meets the first preset condition in the above-mentioned human shape detection result sequence, and obtain a position information group. That is, the position information of each elderly person in the above-mentioned surveillance video in the target area can be identified to obtain a position information group. That is, the position information can represent the position of the elderly person in the target area. There is a dangerous area (calibrated dangerous area) in the target area. For example, the above-mentioned execution subject can identify the current position information of each elderly person in the above-mentioned surveillance video through the following image position detection algorithm (Hough transform, Harris corner detection algorithm, Canny algorithm).

[0101] Optionally, in response to determining that there is location information satisfying the abnormal area condition in the above location information group, the associated alarm device is controlled to perform alarm processing, and voice alarm information is sent to the terminal of each preset emergency contact.

[0102] In some embodiments, the execution subject may control the associated alarm device to perform alarm processing and send voice alarm information to the terminals of each preset emergency contact in response to determining that there is location information that meets the abnormal area condition in the location information group. The abnormal area condition may be that the location represented by the location information is within the dangerous area in the target area.

[0103] First, in response to determining that the audio recognition text represents a call for help text, it is identified whether there is a single elderly image representing the frontal face image of the elderly in the single elderly image group. In this way, it is easy to determine the identity information of the elderly. Then, in response to determining that there is a single elderly image representing the frontal face image of the elderly in the single elderly image group, for each single elderly image representing the frontal face image of the elderly, the following processing steps are performed: intercepting the face image in the single elderly image; correcting the face image to obtain a face correction image; compensating the face correction image to obtain a compensated face image; inputting the compensated face image into a pre-trained face recognition model to obtain a face recognition result; deduplicating each face recognition result to obtain a target face recognition result group. In this way, the identity information of each elderly person can be determined. Afterwards, for each target face recognition result in the above target face recognition result group, the following determination steps are performed: determine whether the face data corresponding to the above target face recognition result is stored in the preset face database; in response to determining that the face data corresponding to the above target face recognition result is stored in the above face database, call the terminal of the emergency contact corresponding to the above face data. In this way, the emergency contact of the fallen elderly person can be urgently contacted. Finally, in response to determining that the face data corresponding to the above target face recognition result is not stored in the above face database, the user corresponding to the above target face recognition result is determined as an abnormal user, and the associated alarm device is controlled to perform alarm processing. In this way, it is possible to determine whether a stranger has broken in by recognizing the face of the elderly in the target area.

[0104] Further references Figure 2 As an implementation of the methods shown in the above figures, the present disclosure provides some embodiments of a dangerous event alarm device for elderly people living alone based on a home Internet of Things controller. These embodiments of the dangerous event alarm device for elderly people living alone based on a home Internet of Things controller are similar to Figure 1 Corresponding to the method embodiments shown, the dangerous event alarm device for elderly people living alone based on the home Internet of Things controller can be specifically applied to various electronic devices.

[0105] like Figure 2As shown, in some embodiments, a dangerous event alarm device 200 for elderly people living alone based on a home Internet of Things controller includes: a first control unit 201, a framing unit 202, a detection unit 203, a determination unit 204, an extraction unit 205, a fall detection unit 206, a second control unit 207, an identification unit 208 and a third control unit 209. Among them, the first control unit 201 is configured to control the associated camera device to collect the monitoring video of the target area in real time in response to receiving the dangerous event detection instruction of the elderly living alone; the framing unit 202 is configured to perform frame processing on the above-mentioned monitoring video to obtain a video frame image sequence; the detection unit 203 is configured to perform human shape detection on each video frame image in the above-mentioned video frame image sequence to generate a human shape detection result and obtain a human shape detection result sequence; the determination unit 204 is configured to determine the video frame image corresponding to the human shape detection result that meets the first preset condition as an elderly image in response to determining that there is a human shape detection result that meets the first preset condition in the above-mentioned human shape detection result sequence, and obtain an elderly image group; the extraction unit 205 is configured to perform frame processing on the above-mentioned monitoring video to obtain a video frame image sequence of each elderly image in the above-mentioned elderly image group The image of a single elderly person is extracted to generate a single elderly person image and obtain a single elderly person image group; the fall detection unit 206 is configured to perform fall detection on each single elderly person image in the above-mentioned single elderly person image group to generate a fall detection result and obtain a fall detection result group; the second control unit 207 is configured to control the associated audio acquisition device to collect the elderly audio in the target area in real time in response to determining that there is a fall detection result that meets the second preset condition in the above-mentioned fall detection result group; the recognition unit 208 is configured to perform audio recognition on the above-mentioned elderly person audio to obtain an audio recognition text; the third control unit 209 is configured to initiate a voice request to the target rescue platform in response to determining that the above-mentioned audio recognition text represents a call for help text, and control the associated alarm device to emit an alarm prompt sound.

[0106] It can be understood that the various units recorded in the dangerous event alarm device 200 for elderly people living alone based on the home Internet of Things controller are similar to those in the reference Figure 1 Therefore, the operations, features and beneficial effects described above for the method are also applicable to the dangerous event alarm device 200 for elderly people living alone based on the home Internet of Things controller and the units contained therein, and will not be repeated here.

[0107] Reference below Figure 3, which shows a schematic diagram of the structure of an electronic device 300 (e.g., a computing device) suitable for implementing some embodiments of the present disclosure. The electronic devices in some embodiments of the present disclosure may include, but are not limited to, mobile terminals such as mobile phones, notebook computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 3 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0108] like Figure 3 As shown, the electronic device 300 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 301, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 302 or a program loaded from a storage device 308 into a random access memory (RAM) 303. In the RAM 303, various programs and data required for the operation of the electronic device 300 are also stored. The processing device 301, the ROM 302, and the RAM 303 are connected to each other via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0109] Typically, the following devices may be connected to the I / O interface 305: input devices 306 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; output devices 307 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; storage devices 308 including, for example, a magnetic tape, a hard disk, etc.; and communication devices 309. The communication devices 309 may allow the electronic device 300 to communicate with other devices wirelessly or by wire to exchange data. Although Figure 3 The electronic device 300 is shown with various devices, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead. Figure 3 Each block shown in the figure may represent one device, or may represent multiple devices as required.

[0110] In particular, according to some embodiments of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, some embodiments of the present disclosure include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In some such embodiments, the computer program can be downloaded and installed from the network through the communication device 309, or installed from the storage device 308, or installed from the ROM 302. When the computer program is executed by the processing device 301, the above-mentioned functions defined in the method of some embodiments of the present disclosure are executed.

[0111] It should be noted that the computer-readable medium recorded in some embodiments of the present disclosure may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In some embodiments of the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In some embodiments of the present disclosure, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer readable signal medium may also be any computer readable medium other than a computer readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0112] In some embodiments, the client and the server may communicate using any currently known or future developed network protocol such as HTTP (HyperText Transfer Protocol), and may be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), an internet (e.g., the Internet), and a peer-to-peer network (e.g., an ad hoc peer-to-peer network), as well as any currently known or future developed network.

[0113] The above-mentioned computer-readable medium may be included in the above-mentioned electronic device; or it may exist independently without being assembled into the electronic device. The above-mentioned computer-readable medium carries one or more programs. When the above-mentioned one or more programs are executed by the electronic device, the electronic device: in response to receiving a dangerous event detection instruction, controls the associated camera device to collect the monitoring video of the target area in real time; performs frame processing on the above-mentioned monitoring video to obtain a video frame image sequence; performs human shape detection on each video frame image in the above-mentioned video frame image sequence to generate a human shape detection result to obtain a human shape detection result sequence; in response to determining that there is a human shape detection result that meets the first preset condition in the above-mentioned human shape detection result sequence, determines the video frame image corresponding to the human shape detection result that meets the first preset condition as an old person image, and obtains an old person image. image group; extracting a single elderly image from each elderly image in the elderly image group to generate a single elderly image and obtain a single elderly image group; performing fall detection on each single elderly image in the single elderly image group to generate a fall detection result and obtain a fall detection result group; in response to determining that there is a fall detection result that meets the second preset condition in the fall detection result group, controlling the associated audio acquisition device to collect the elderly audio in the target area in real time; performing audio recognition on the elderly audio to obtain an audio recognition text; in response to determining that the audio recognition text represents a call for help text, initiating a voice request to the target rescue platform, and controlling the associated alarm device to emit an alarm prompt sound.

[0114] Computer program code for performing the operations of some embodiments of the present disclosure may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0115] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0116] The units described in some embodiments of the present disclosure may be implemented by software or by hardware. The described units may also be provided in a processor, for example, may be described as: a processor comprising: a first control unit, a framing unit, a detection unit, a determination unit, an extraction unit, a fall detection unit, a second control unit, an identification unit and a third control unit. Among them, the names of these units do not constitute limitations on the units themselves in certain circumstances, for example, the framing unit may also be described as "a unit that performs framing processing on the above-mentioned surveillance video to obtain a video frame image sequence".

[0117] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), and the like.

[0118] The above descriptions are only some preferred embodiments of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present disclosure is not limited to the technical solutions formed by a specific combination of the above-mentioned technical features, but should also cover other technical solutions formed by any combination of the above-mentioned technical features or their equivalent features without departing from the above-mentioned inventive concept. For example, the above-mentioned features are replaced with the technical features with similar functions disclosed in the embodiments of the present disclosure (but not limited to) and the technical solutions formed.

Claims

1. A dangerous event alarm method for elderly people living alone based on a home Internet of Things controller, comprising: In response to receiving a dangerous event detection instruction for an elderly person living alone, controlling an associated camera device to collect surveillance video of a target area in real time; Performing frame processing on the surveillance video to obtain a video frame image sequence; Performing human shape detection on each video frame image in the video frame image sequence to generate a human shape detection result, thereby obtaining a human shape detection result sequence; In response to determining that there is a human shape detection result that meets a first preset condition in the human shape detection result sequence, determining the video frame image corresponding to the human shape detection result that meets the first preset condition as an old person image, thereby obtaining an old person image group; Extracting a single elderly image from each elderly image in the elderly image group to generate a single elderly image and obtain a single elderly image group; Performing fall detection on each single elderly person image in the single elderly person image group to generate a fall detection result, thereby obtaining a fall detection result group; In response to determining that there is a fall detection result that meets a second preset condition in the fall detection result group, controlling an associated audio collection device to collect audio of an elderly person in a target area in real time; Performing audio recognition on the elderly audio to obtain audio recognition text; In response to determining that the audio recognition text represents a distress call text, a voice request is initiated to a target rescue platform, and an associated alarm device is controlled to emit an alarm prompt tone.

2. The method according to claim 1, wherein: The method further comprises: In response to determining that the audio recognition text represents a call for help text, identifying whether there is a single elderly person image representing a frontal facial image of an elderly person in the group of single elderly person images; In response to determining that there is a single elderly image representing a frontal face image of an elderly person in the single elderly image group, for each single elderly image representing a frontal face image of an elderly person, the following processing steps are performed: intercepting a face image from the single elderly person image; Performing correction processing on the face image to obtain a face-corrected image; Performing compensation processing on the face correction image to obtain a compensated face image; Inputting the compensated face image into a pre-trained face recognition model to obtain a face recognition result; De-duplicate each face recognition result to obtain a target face recognition result group; For each target face recognition result in the target face recognition result group, the following determination steps are performed: Determine whether face data corresponding to the target face recognition result is stored in a preset face database; In response to determining that face data corresponding to the target face recognition result is stored in the face database, a call is made to a terminal of an emergency contact corresponding to the face data.

3. The method according to claim 2, wherein: The determining step further comprises: In response to determining that the face database does not store face data corresponding to the target face recognition result, the user corresponding to the target face recognition result is determined to be an abnormal user, and the associated alarm device is controlled to perform alarm processing.

4. The method according to claim 1, wherein: The method further comprises: In response to determining that there is no human shape detection result satisfying a first preset condition in the human shape detection result sequence, determining a video acquisition time; In response to determining that the duration between the current time and the video acquisition time is greater than or equal to the preset duration, the associated alarm device is controlled to perform alarm processing, and voice alarm information is sent to the terminal of each preset emergency contact.

5. The method according to claim 1, wherein: The method further comprises: In response to determining that there is a human shape detection result that meets the first preset condition in the human shape detection result sequence, identifying the current position information of each elderly person in the monitoring video to obtain a position information group; In response to determining that there is location information satisfying the abnormal area condition in the location information group, the associated alarm device is controlled to perform alarm processing, and voice alarm information is sent to the terminal of each preset emergency contact.

6. The method according to claim 1, wherein: The performing fall detection on each single elderly image in the single elderly image group to generate a fall detection result includes: Inputting the single elderly image into a pre-trained elderly upright state recognition model to generate an elderly upright state recognition result; In response to determining that the elderly upright state recognition result meets the first detection condition, the single elderly image is input into a pre-trained elderly fall recognition model to generate an elderly fall recognition result.

7. The method according to claim 1, wherein: The step of performing audio recognition on the elderly audio to obtain audio recognition text includes: Preprocessing the elderly audio to obtain an elderly audio filter signal; The elderly audio filter signal is input into a pre-trained speech recognition model to obtain an audio recognition text, wherein the pre-trained speech recognition model includes: a classification network, a coding network, an enhancement network and a decoding network.

8. A dangerous event alarm device for elderly people living alone based on a home Internet of Things controller, comprising: A first control unit is configured to, in response to receiving a dangerous event detection instruction for an elderly person living alone, control an associated camera device to collect surveillance video of a target area in real time; A framing unit is configured to perform framing processing on the surveillance video to obtain a video frame image sequence; A detection unit is configured to perform human shape detection on each video frame image in the video frame image sequence to generate a human shape detection result and obtain a human shape detection result sequence; a determining unit configured to, in response to determining that there is a human shape detection result that satisfies a first preset condition in the human shape detection result sequence, determine the video frame image corresponding to the human shape detection result that satisfies the first preset condition as an old person image, and obtain an old person image group; an extraction unit configured to extract a single elderly image from each elderly image in the elderly image group to generate a single elderly image and obtain a single elderly image group; a fall detection unit configured to perform fall detection on each single elderly person image in the single elderly person image group to generate a fall detection result, thereby obtaining a fall detection result group; A second control unit is configured to, in response to determining that there is a fall detection result that meets a second preset condition in the fall detection result group, control an associated audio collection device to collect audio of an elderly person in a target area in real time; A recognition unit, configured to perform audio recognition on the elderly audio to obtain an audio recognition text; The third control unit is configured to initiate a voice request to a target rescue platform in response to determining that the audio recognition text represents a distress call text, and control an associated alarm device to emit an alarm prompt tone.

9. An electronic device, comprising: one or more processors; a storage device having one or more programs stored thereon; When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 7.

10. A computer readable medium having a computer program stored thereon, wherein: When the program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images