Intelligent voice broadcast system and method for elevator

Abstract

According to the intelligent voice broadcast system and method for the elevator, elevator car environment data are collected in real time through a multi-source heterogeneous sensor set, and passenger density distribution and movement tracks are calculated through fusion of a millimeter wave radar and a 3D visual sensor; physiological features of a handrail contact area are detected through a biological impedance sensor so as to judge the health state of a passenger, and the current signal-to-noise ratio is dynamically calculated and the voice frequency band is optimized through an environment noise analysis module. Therefore, by constructing a technical system of'multi-modal perception-edge intelligent decision-cross-modal interaction 'and adopting innovative means such as millimeter-wave radar and biological impedance fusion sensing, deep reinforcement learning of a dynamic strategy and a phased array directional sound field, the problems of broadcast stiffness, poor scene adaptability and privacy potential safety hazards are solved. According to the system and method, accurate passenger feature recognition, dynamic voice content generation and multi-channel collaborative feedback are achieved through two-dimensional innovation, and the intelligent level and service efficiency of elevator voice interaction are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent elevators, and in particular, to an elevator intelligent voice broadcast system and method. Background Art

[0002] Currently, most elevator voice broadcast systems adopt pre-recorded fixed voice libraries and perform one-way information output based on a simple floor trigger mechanism. Existing technologies usually rely on basic infrared sensors or cameras for passenger detection. The voice broadcast content and form are single, lacking the dynamic perception ability of passenger characteristics, environmental status, and equipment operation parameters. Moreover, the interaction method is limited to traditional speaker playback, making it difficult to meet the perception needs of special groups and the intelligent service requirements in complex scenarios.

[0003] The traditional system has three core defects: First, the broadcast strategy is out of touch with passenger characteristics and the real-time environment, and it cannot adaptively adjust the speech rate, content, and interaction method. Second, it lacks a multi-modal data fusion decision-making mechanism, resulting in slow response to emergencies and the lack of personalized services. Third, data processing relies on a centralized architecture, posing a risk of privacy leakage and making it difficult to achieve low-latency edge computing, restricting the real-time performance and security of the system. Summary of the Invention

[0004] The present invention aims to solve at least one of the technical problems in the related technologies to some extent.

[0005] To this end, the present invention aims to propose an elevator intelligent voice broadcast system and method. By constructing a technical system of "multi-modal perception-edge intelligent decision-making-cross-modal interaction" and adopting innovative means such as millimeter-wave radar and bio-impedance fusion sensing, deep reinforcement learning dynamic strategy, and phased array directional sound field, the problems of rigid broadcast, poor scene adaptability, and privacy and security risks in the existing technology are solved. The dual-dimensional innovation of the system and method realizes accurate identification of passenger characteristics, dynamic generation of voice content, and multi-channel collaborative feedback, significantly improving the intelligent level and service efficiency of elevator voice interaction.

[0006] To achieve the above object, the present invention proposes an elevator intelligent voice broadcast system and method, including the following steps:

[0007] S1. Real-time collect the car environment data through a multi-source heterogeneous sensor group, where

[0008] a. Use millimeter-wave radar and 3D vision sensors for fusion calculation of passenger density distribution and movement trajectory;

[0009] b. Detect the physiological characteristics of the handrail contact area through a bio-impedance sensor to judge the passenger's health status;

[0010] c. Use the environmental noise analysis module to dynamically calculate the current signal-to-noise ratio and optimize the voice frequency band;

[0011] S2. Build a dynamic broadcast strategy based on deep reinforcement learning, where:

[0012] a. Input elevator operation parameters, passenger behavior characteristics, and building space data into the strategy network;

[0013] b. When a pregnant woman or a disabled person is detected in the car, the safety warning broadcast priority is automatically raised to the highest level;

[0014] c. Predict the running jitter intensity based on the real-time load change rate and dynamically adjust the rhythm interval of voice broadcast;

[0015] S3, generating context-aware speech content, wherein:

[0016] a. Start the pre-announcement system 0.5 seconds before the elevator door opens, generating a shortened version of the "approaching floor" reminder tone;

[0017] b. When it is recognized that the passenger has not selected a floor for a long time, an active inquiry statement "Which area do you want to go to" is triggered;

[0018] c. Combine the passenger's historical elevator data to insert a personalized greeting when arriving at the target floor;

[0019] S4. Perform cross-modal collaborative feedback, where:

[0020] a. When an emergency notification is broadcast, the piezoelectric film on the inner wall of the car is activated to generate a tactile warning signal;

[0021] b. When the ambient noise exceeds 65 decibels, it automatically switches to the bone conduction speaker for directional audio transmission.

[0022] Specifically, the passenger identification method in step S2b includes: extracting the spatiotemporal feature vector of the passenger when walking by using a gait analysis algorithm;

[0023] Combine the Bluetooth broadcast signal of the smart bracelet to perform identity cross-verification;

[0024] When a medical emergency personnel is identified, the authority restrictions on priority access to the floor are automatically lifted.

[0025] Specifically, the personalized greeting generation method of step S3c includes:

[0026] Establish a mapping relationship between the employee database and elevator behavior characteristics;

[0027] When a registered user is identified, the preset greeting preference is called;

[0028] On the birthday, the greeting message is automatically inserted and the car lighting system is linked to change the light color.

[0029] Specifically, it also includes the intelligent handling process when the elevator fails:

[0030] When abnormal vibration is detected, the three-dimensional sound field positioning guidance system is activated;

[0031] Generate real-time treatment suggestions through the speech synthesis engine, and the speech speed is automatically adjusted according to the carbon dioxide concentration in the cabin;

[0032] Activate the voice-controlled emergency unlocking protocol when the access control system fails.

[0033] An elevator intelligent voice broadcast system, implemented using any one of the elevator intelligent voice broadcast methods of claims 1-4, comprising:

[0034] Multimodal sensing array, including a radar-vision fusion sensor unit deployed on the top of the car with sub-centimeter spatial resolution, a bioelectric detection circuit integrated in the handrail that can measure heart rate variability index, and a 16-channel acoustic acquisition module arranged around the car;

[0035] Edge computing decision center, including passenger intention prediction model, button operation sequence analysis based on LSTM neural network, speech synthesis quality evaluator, real-time calculation of speech intelligibility index and dynamic adjustment of synthesis parameters;

[0036] Adaptive output device group, including deformable directional speaker array, which can adjust the sound beam focus angle according to the passenger position, intelligent light-transmitting glass, which can display augmented reality navigation signs during voice broadcast, and distributed tactile feedback matrix, which supports multi-point differentiated vibration encoding;

[0037] The cloud-based collaborative management platform includes the sharing of abnormal event knowledge base across elevator groups and a digital twin-based broadcast strategy simulation test environment.

[0038] Specifically, it is characterized in that the bioelectric detection circuit in the multimodal sensing array comprises:

[0039] Flexible electrode array, using silver nanowire composite material to achieve curved surface conformity;

[0040] Impedance spectrum analysis unit, which can detect bioelectric signals in the frequency range of 0.1-100kHz;

[0041] Safety isolation circuit ensures that the detection voltage is always lower than 5mV.

[0042] Specifically, the deformable directional speaker array in the adaptive output device group includes:

[0043] A phased array module consisting of 64 micro transducers;

[0044] Beamforming controller based on passenger iris positioning;

[0045] An ultrasonic carrier modulation unit for realizing the directional transmission of audible sound.

[0046] Specifically, the cloud collaborative management platform includes:

[0047] A voice log evidence storage system based on blockchain;

[0048] A model optimization pipeline supporting federated learning;

[0049] A remote diagnosis interface for the elevator voice system, which can adjust the voice priority weights of each floor online.

[0050] Additional aspects and advantages of the present invention will be given in part in the following description, become apparent in part from the following description, or be learned through the practice of the present invention. Description of the Drawings

[0051] The above and / or additional aspects and advantages of the present invention will become apparent and be readily understood from the following description of the embodiments in conjunction with the drawings, where:

[0052] Figure 1 Is a flowchart of the elevator intelligent voice broadcast method of the present invention;

[0053] Figure 2 Is a schematic diagram of the elevator intelligent voice broadcast system of the present invention.

[0054] In the figure: 201, radar-vision fusion sensing unit; 202, bioelectric detection circuit; 203, 16-channel acoustic acquisition module; 204, edge computing decision center; 205, deformable directional speaker array; 206, intelligent light-transmitting glass; 208, cloud collaborative management platform. Detailed Embodiments

[0055] The embodiments of the present invention will be described in detail below. Examples of the embodiments are shown in the drawings, where the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the drawings are exemplary and are intended to explain the present invention and should not be construed as limiting the present invention. On the contrary, the embodiments of the present invention include all variations, modifications, and equivalents falling within the spirit and scope of the appended claims.

[0056] The elevator intelligent voice broadcast system and method according to the embodiments of the present invention will be described below in conjunction with the drawings.

[0057] As Figure 1 - Figure 2As shown, the elevator intelligent voice broadcast system and method of the embodiment of the present invention, the multi-source heterogeneous sensor group includes a data fusion module of a millimeter wave radar array (working frequency 76-81GHz) and a binocular 3D vision sensor, wherein the millimeter wave radar detects the macroscopic movement trend of passengers through the Doppler effect, the 3D vision sensor realizes fine posture analysis through feature point cloud matching, and the passenger density distribution is calculated in real time by the DBSCAN clustering algorithm. The bioimpedance sensor uses a four-electrode method to measure the skin conductivity and capacitance changes at the contact part of the handrail, and extracts the heart rate variability index (HRV) in combination with the PPG signal. When the HRV standard deviation is detected to exceed 50ms for 3 consecutive seconds, a health warning is triggered. The environmental noise analysis module dynamically selects the 1.6-3.2kHz frequency band with the best signal-to-noise ratio for voice enhancement through the improved Bark spectrum sub-band division method.

[0058] It should be noted that the fusion of millimeter-wave radar and 3D vision described in this embodiment uses the Kalman filter algorithm for spatiotemporal calibration to solve the problem of data asynchrony between sensors. The bioimpedance detection circuit sets a contact impedance compensation loop to eliminate the measurement deviation caused by changes in hand humidity, and uses a sliding window method to filter instantaneous interference signals. The noise analysis module introduces a human ear auditory masking effect model to ensure that the speech band optimization conforms to the psychoacoustic characteristics, and at the same time eliminates interference in specific frequency bands (such as elevator motor noise) through an adaptive filter.

[0059] The gait analysis algorithm uses the OpenPose framework to extract 25 skeletal key points of the human body, calculates feature vectors such as gait cycle and stride length through the spatiotemporal graph convolutional network (ST-GCN), and performs cosine similarity matching with the preset medical personnel gait feature library. The identity cross-verification module monitors the device fingerprint (such as MAC address hash value) in the Bluetooth 5.1 broadcast packet, uses the elliptic curve cryptography (ECC) algorithm to transmit the temporary session key, and matches and verifies it with the pre-stored whitelist database. When a medical emergency personnel is identified, a permission upgrade instruction is sent to the elevator control system to remove the floor selection restriction and activate the emergency pass mode.

[0060] It should be noted that the gait recognition described in this embodiment sets a dynamic threshold adjustment mechanism. When the elevator acceleration exceeds 0.5m / s 2 The matching tolerance is automatically relaxed to 0.85 (default 0.92) when the user is logged in. Bluetooth authentication uses a two-factor authentication mechanism, which requires both device fingerprint matching and dynamic password authentication. The medical staff identity database and the hospital HIS system are synchronized in real time through the API to ensure the timeliness of permission information and set the automatic expiration time of permissions (such as 30 minutes after the end of the operation).

[0061] The personalized greeting generation module constructs user portraits through the mapping relationship between the employee database and elevator riding behavior characteristics, including frequently used floors, elevator riding time periods, and interaction preferences. When a registered user is recognized, the preset greeting preference settings (such as language type, volume level) are called, and a greeting statement is dynamically synthesized through a natural language generation (NLG) engine. On the user's birthday, a birthday greeting is automatically inserted and the car lighting system is linked. The light color is adjusted to the RGB value preset by the user (such as warm yellow #FFD700) through PWM dimming.

[0062] It should be noted that the user portrait data described in this embodiment is processed using differential privacy technology to ensure irreversible anonymization of personal information. The NLG engine is based on the Transformer architecture, supports mixed Chinese and English broadcasting, and can adapt to dialect features through transfer learning. The lighting linkage module controls the LED driver through the DMX512 protocol, and the light color transition time is set to 500 ms to avoid visual discomfort.

[0063] In the elevator fault handling process, the three-dimensional sound field positioning and guiding system determines the passenger's position through the direction of arrival (DOA) estimation algorithm and controls the phased array speaker to generate directional prompt voices. The speech synthesis engine calls the predefined template library according to the fault type and dynamically adjusts the speech rate in combination with real-time sensor data (such as carbon dioxide concentration): when the CO2 concentration > 1000 ppm, the speech rate is reduced to 0.75 times the normal speed to relieve passenger anxiety. The emergency unlocking protocol uses voiceprint recognition technology to verify the password and supports dynamic keys (such as "emergency password + timestamp hash").

[0064] It should be noted that the DOA estimation described in this embodiment uses the MUSIC algorithm, and the spatial resolution reaches ±5°. The voice template library is constructed based on a fault knowledge graph and supports multi-level association reasoning (such as "abnormal vibration → mechanical fault → it is recommended to stand against the wall"). The voiceprint recognition is equipped with an anti-noise microphone array, the signal-to-noise ratio threshold is set to 15 dB, and the false acceptance rate (FAR) is lower than 0.01%.

[0065] The radar-vision fusion sensing unit (201) of the multi-modal perception array uses FPGA to implement data preprocessing, and the sub-centimeter resolution is achieved through the sub-pixel interpolation algorithm. The heart rate variability index measurement of the bioelectric detection circuit (202) uses the Pan-Tompkins algorithm to extract the R-wave interval and calculates the power spectral density through the Welch method. The 16-channel acoustic acquisition module (203) is arranged with a circular microphone array, and the sound source localization is achieved through the GCC-PHAT algorithm.

[0066] It should be noted that the FPGA preprocessing module described in this embodiment integrates the CORDIC algorithm to accelerate coordinate transformation, and the delay is controlled within 5 ms. The heart rate detection is provided with a motion artifact compensation module to remove limb movement interference through accelerometer data. The sound source localization module supports the blind source separation technology and can extract the target voice when multiple people are speaking simultaneously.

[0067] The flexible electrode array is manufactured by 3D printing technology. The substrate material is a TPU-silver nanowire composite material (silver content 12%), and the electrode spacing is adjustable at 2 mm to adapt to different hand sizes. The impedance spectroscopy analysis unit uses the AD5933 chipset and can resolve impedance changes of 0.5% at a detection frequency of 10 kHz. The cell membrane capacitance parameters are inverted through the Cole-Cole model. The safety isolation circuit includes dual protection of an optocoupler relay and a self-recovery fuse, and the insulation resistance between the detection circuit and the main control system is > 100 MΩ.

[0068] It should be noted that the electrode surface described in this embodiment is coated with a PEDOT:PSS conductive polymer layer, and the contact impedance is stabilized below 5 kΩ. The impedance detection uses the multi-point frequency scanning method (1 kHz, 10 kHz, 100 kHz), and the skin capacitance value is inverted through an equivalent circuit model. The isolation circuit is provided with real-time leakage current monitoring, and when it exceeds 10 μA, the hardware power-off protection is triggered.

[0069] The phased array module of the deformable directional speaker array (205) consists of 64 micro transducers, with a working frequency of 20 - 40 kHz. The beamforming controller calculates the phase delay parameters based on passenger iris localization (error < 1°). The ultrasonic carrier modulation unit uses DSB-SC technology to modulate the voice signal to a 40 kHz carrier and demodulates it into audible sound through the air nonlinear effect.

[0070] It should be noted that the iris localization described in this embodiment uses a near-infrared camera (wavelength 850 nm) and an improved Daugman algorithm, and the recognition speed is < 200 ms. The beamforming supports dynamic focusing, and the sound pressure level fluctuation is controlled within the range of ±2 dB. The ultrasonic modulation is provided with automatic gain control (AGC) to ensure that the MOS score of voice clarity within a transmission distance of 3 m is > 4.0.

[0071] The blockchain evidence storage system of the cloud collaborative management platform (208) adopts the Hyperledger Fabric framework. Each voice log generates a Merkle tree hash and writes it into the smart contract. The federated learning pipeline realizes model parameter aggregation through Paillier homomorphic encryption. The local training cycle of each elevator node is set to 5 rounds, and the remote diagnosis interface supports the SNMP protocol and can dynamically adjust the floor voice weight (for example, the priority of the fire escape floor is increased to Level 5).

[0072] It should be noted that in this embodiment, a privacy channel is set up in the blockchain, and sensitive data is only visible to authorized parties. The federated learning model adopts the FedAvg algorithm, the global model update interval is 24 hours, and the voice weight adjustment is based on a reinforcement learning strategy, and the broadcast frequency is optimized by combining historical pedestrian flow data.

[0073] In summary, animal experiments and clinical verification have shown that the traditional Chinese medicine composition exhibits significant beneficial effects in the treatment of herpes zoster and its sequelae pain: in the acute-phase treatment, it can effectively inhibit the replication of varicella-zoster virus (the virus load reduction rate ≥ 82.3%), shorten the skin lesion healing time to 4.5 days (32% shorter than acyclovir), and significantly reduce the levels of inflammatory factors IL-6 and TNF-α (the reduction rate ≥ 50%); for post-herpetic neuralgia, the combination of concentrated pills and topical patches can increase the mechanical pain threshold by 52% (up to 12.5 g), inhibit the expression of substance P in the spinal dorsal horn (the reduction rate is 30%), and the effective rate of pain relief reaches 88% (significantly better than 68% of gabapentin); in terms of safety, the acute toxicity test LD50 > 20 g / kg (non-toxic level), there is no sensitization reaction and toxic components such as aristolochic acid, and the total flavonoid and saponin contents reach 9.5% and 5.6% respectively, with the advantages of high-efficiency antiviral, long-acting analgesia and high safety. The comprehensive efficacy and safety are significantly better than the treatment regimens of traditional insect-containing drugs or chemical drugs.

[0074] Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limitations of the present invention. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention.

Claims

1. An intelligent voice broadcast method for an elevator, characterized in that, The following steps are involved: S1. Collect cabin environment data in real time through a multi-source heterogeneous sensor group, where: a. Use millimeter wave radar and 3D vision sensor fusion to calculate passenger density distribution and movement trajectory; b. Use bioimpedance sensors to detect the physiological characteristics of the handrail contact area to determine the health status of the passenger; c. Use the environmental noise analysis module to dynamically calculate the current signal-to-noise ratio and optimize the voice frequency band; S2. Build a dynamic broadcast strategy based on deep reinforcement learning, where: a. Input elevator operation parameters, passenger behavior characteristics, and building space data into the strategy network; b. When a pregnant woman or a disabled person is detected in the car, the safety warning broadcast priority is automatically raised to the highest level; c. Predict the running jitter intensity based on the real-time load change rate and dynamically adjust the rhythm interval of voice broadcast; S3, generating context-aware speech content, wherein: a. Start the pre-announcement system 0.5 seconds before the elevator door opens, generating a shortened version of the "approaching floor" reminder tone; b. When it is recognized that the passenger has not selected a floor for a long time, an active inquiry statement "Which area do you want to go to" is triggered; c. Combine the passenger's historical elevator data to insert a personalized greeting when arriving at the target floor; S4. Perform cross-modal collaborative feedback, where: a. When an emergency notification is broadcast, the piezoelectric film on the inner wall of the car is activated to generate a tactile warning signal; b. When the ambient noise exceeds 65 decibels, it automatically switches to the bone conduction speaker for directional audio transmission.

2. The elevator intelligent voice broadcast method according to claim 1, wherein The passenger identification method in step S2b includes: extracting the spatiotemporal feature vector of the passenger while walking by using a gait analysis algorithm; Combine the Bluetooth broadcast signal of the smart bracelet to perform identity cross-verification; When a medical emergency personnel is identified, the authority restrictions on priority access to the floor are automatically lifted.

3. The elevator intelligent voice broadcast method according to claim 1, wherein, The personalized greeting generation method of step S3c includes: Establish a mapping relationship between the employee database and elevator behavior characteristics; When a registered user is identified, the preset greeting preference is called; On the birthday, the greeting message is automatically inserted and the car lighting system is linked to change the light color.

4. The elevator intelligent voice broadcast method according to claim 1, wherein It also includes the intelligent handling process when the elevator fails: When abnormal vibration is detected, the three-dimensional sound field positioning guidance system is activated; Generate real-time treatment suggestions through the speech synthesis engine, and the speech speed is automatically adjusted according to the carbon dioxide concentration in the cabin; Activate the voice-controlled emergency unlocking protocol when the access control system fails.

5. An elevator intelligent voice broadcast system implemented by using the elevator intelligent voice broadcast method according to any one of claims 1-4, characterized in that, include: A multimodal sensing array, including a radar-vision fusion sensor unit (201) deployed on top of the car, with sub-centimeter spatial resolution, a bioelectric detection circuit (202) integrated in the handrail, capable of measuring heart rate variability index, and a 16-channel acoustic acquisition module (203) arranged around the car; An edge computing decision center (204), including a passenger intention prediction model, a button operation sequence analysis based on an LSTM neural network, a speech synthesis quality evaluator, a real-time calculation of a speech intelligibility index and a dynamic adjustment of synthesis parameters; The adaptive output device group includes a deformable directional speaker array (205) that can adjust the sound beam focusing angle according to the passenger position, an intelligent light-transmitting glass (206) that displays augmented reality navigation signs during voice announcements, and a distributed haptic feedback matrix (207) that supports multi-point differential vibration coding; The cloud collaborative management platform (208) includes the sharing of abnormal event knowledge bases across elevator groups and a simulation test environment for voice announcement strategies based on digital twins.

6. The elevator intelligent voice broadcast system according to claim 5, wherein The bioelectric detection circuit (202) in the multimodal perception array includes: A flexible electrode array that uses silver nanowire composite materials to achieve curved surface fitting; An impedance spectrum analysis unit that can detect bioelectric signals in the frequency range of 0.1 - 100 kHz; A safety isolation circuit that ensures the detection voltage is always lower than 5 mV.

7. The elevator intelligent voice broadcast system according to claim 5, characterized in that, The deformable directional speaker array (205) in the adaptive output device group includes: A phased array module composed of 64 micro transducers; A beamforming controller based on passenger iris positioning; An ultrasonic carrier modulation unit that realizes the directional transmission of audible sound.

8. The elevator intelligent voice broadcast system according to claim 5, characterized in that, The cloud collaborative management platform (208) includes: A voice log storage and verification system based on blockchain; A model optimization pipeline that supports federated learning; A remote diagnosis interface for the elevator voice system that can online adjust the voice priority weights of each floor.

Images