Multi-source fusion alarm method and system for emergency intelligent broadcasting system

Abstract

The application relates to the technical field of acoustic signal processing and emergency broadcasting, and discloses a multi-source fusion alarm method and system for an emergency intelligent broadcasting system. The method comprises the following steps: acquiring multi-path environmental audio data collected by a distributed microphone array and a target audio signal to be broadcast; performing spatial energy spectrum analysis according to the multi-path environmental audio data to obtain an environmental noise distribution graph representing noise intensities of different coordinate points in space; determining a gain compensation parameter according to the environmental noise distribution graph and a discretized guide node, and performing gain adjustment on the target audio signal to obtain a to-be-projected audio signal; performing phased beamforming processing according to the to-be-projected audio signal and a calculated phase delay parameter to obtain a directional projection vector and drive a loudspeaker array to emit sound waves, thereby forming a spatialized guide sound field on an escape path. The method can realize accurate directional projection and dynamic path guidance of evacuation instructions in a complex high-noise environment.

Description

Technical Field

[0001] This invention relates to the fields of acoustic signal processing and emergency broadcasting technology, and in particular to a multi-source fusion alarm method and system for an emergency intelligent broadcasting system. Background Technology

[0002] Currently, emergency broadcasting systems are becoming increasingly important as core infrastructure for ensuring public safety and evacuating people during emergencies. In critical scenarios such as fires and earthquakes, timely, clear, and directional evacuation guidance is directly related to the safety of trapped individuals.

[0003] In existing technologies, emergency broadcasting typically relies on omnidirectional loudspeakers for fixed-power wide-area sound coverage, or simply combines on-site smart sensors to collect global volume data, mechanically increasing the overall output power of the broadcasting equipment. However, through derivation and operational practice analysis of these existing technologies, it has been found that they have insurmountable limitations when dealing with complex and ever-changing disaster scenes. On the one hand, sudden events such as fires or panic-stricken crowds are often accompanied by extremely high-decibel and unevenly distributed dynamic background noise. Traditional non-directional omnidirectional broadcasting is easily drowned out by localized high noise. Simply and blindly increasing the global volume will not only cause severe acoustic reverberation and howling, but will also further exacerbate the panic of the crowd, leading to a precipitous drop in the intelligibility of instructions. On the other hand, existing technologies severely lack the ability to control the spatial dimension of the sound field. Sound waves spread aimlessly and cannot be dynamically anchored to safe escape routes in physical space. Even if trapped people hear the alarm, they cannot locate the sound. In extreme environments with limited vision, such as dense smoke or power outages, they are very likely to deviate from the safe passage, making it impossible to form effective spatial evacuation guidance.

[0004] In summary, existing technologies suffer from the technical problem that emergency evacuation instructions are easily overwhelmed and path guidance fails. Summary of the Invention

[0005] This invention provides a multi-source fusion alarm method and system for emergency intelligent broadcasting systems, in order to solve the technical problems in the prior art where emergency evacuation instructions are easily overwhelmed and path guidance fails.

[0006] Firstly, in order to solve the above-mentioned technical problems, the present invention provides a multi-source fusion alarm method for an emergency intelligent broadcasting system, comprising:

[0007] Acquire multi-channel environmental audio data collected by a distributed microphone array, and acquire the target audio signal to be broadcast;

[0008] Spatial energy spectrum analysis is performed on the multi-channel environmental audio data to obtain an environmental noise distribution map characterizing the noise intensity at different coordinate points in space.

[0009] Obtain the escape path coordinates of the target area, and discretize the escape path coordinates into multiple consecutive guide nodes;

[0010] The gain compensation parameters for each of the guiding nodes are calculated based on the environmental noise distribution map, and the gain of the target audio signal to be broadcast is adjusted using the gain compensation parameters to obtain the audio signal to be projected.

[0011] The phase delay parameter is calculated based on the spatial geometric relationship between each array unit in the speaker array and the guide node;

[0012] The phased beamforming process is performed on the audio signal to be projected according to the phase delay parameter to obtain a directional projection vector pointing to the guiding node;

[0013] The speaker array is controlled to emit sound waves according to the directional projection vector, forming a spatialized guiding sound field that is dynamically distributed along the escape path with the guiding nodes.

[0014] Secondly, the present invention provides a multi-source fusion alarm system for an emergency intelligent broadcasting system, comprising:

[0015] The data acquisition module is used to acquire multi-channel environmental audio data collected by a distributed microphone array and to acquire the target audio signal to be broadcast.

[0016] The noise analysis module is used to perform spatial energy spectrum analysis based on the multi-channel environmental audio data to obtain an environmental noise distribution map characterizing the noise intensity at different coordinate points in space.

[0017] The path discretization module is used to obtain the escape path coordinates of the target area and discretize the escape path coordinates into multiple consecutive guide nodes;

[0018] The strategy adjustment module is used to calculate the gain compensation parameters for each of the guiding nodes based on the environmental noise distribution map, and to adjust the gain of the target audio signal to be broadcast using the gain compensation parameters to obtain the audio signal to be projected.

[0019] The delay calculation module is used to calculate the phase delay parameter based on the spatial geometric relationship between each array unit in the speaker array and the guide node;

[0020] A beamforming module is used to perform phased beamforming processing on the audio signal to be projected according to the phase delay parameter to obtain a directional projection vector pointing to the guiding node;

[0021] The deployment module is used to control the loudspeaker array to emit sound waves according to the directional projection vector, forming a spatialized guiding sound field that is dynamically distributed along the escape path with the guiding nodes.

[0022] Compared with the prior art, the present invention has the following beneficial effects:

[0023] (1) This invention effectively solves the problem of local submersion caused by uneven noise distribution in complex scenarios by collecting multiple ambient audio signals through a distributed microphone array and constructing a spatial noise distribution map. By upgrading single-point sampling to spatial multi-point sampling and using spatial interpolation algorithms to generate high-dimensional noise heatmaps, the system can accurately calculate gain compensation parameters for each discrete guiding node. This design avoids acoustic reverberation and crowd panic caused by blindly increasing the global volume, realizes on-demand allocation and precise enhancement of evacuation command energy, and significantly improves command intelligibility in extremely noisy environments.

[0024] (2) This invention completely breaks through the technical bottleneck of traditional broadcast sound field diffusion and lack of directionality by discretizing the escape path into guiding nodes and combining it with phased beamforming technology. By establishing the spatial geometric mapping relationship between each array unit and the discrete guiding nodes, and calculating the precise phase delay parameters, the sound wave energy can be focused and locked onto a specific safe location like a searchlight. This dynamic spatial guidance mechanism, which follows the path, provides a clear auditory beacon for trapped personnel in visually limited environments such as dense smoke and power outages, significantly improving the efficiency of evacuation and the accuracy of directional guidance.

[0025] (3) This invention ensures the robustness and reliability of evacuation schemes in complex acoustic environments through subband modulation signal fusion and a closed-loop optimization mechanism based on acoustic simulation. By using subband modulation technology to accurately embed the command speech into the spectral holes of the target audio, the signal distortion problem caused by the irreversibility of speech recognition features is solved. At the same time, by introducing gradient descent optimization with a preset clarity index and real-time feedback compensation from the receiver sensor, the system has the ability to self-correct dynamic environments (such as wind speed changes and obstacle movement), realizing real-time, closed-loop, and refined control of spatial guidance effects, and providing highly practical underlying technical support for intelligent alarms in emergency scenarios. Attached Figure Description

[0026] Figure 1 This is a schematic flowchart of a multi-source fusion alarm method for an emergency intelligent broadcasting system provided in the first embodiment of the present invention;

[0027] Figure 2 This is a schematic diagram of the structure of a multi-source fusion alarm system for an emergency intelligent broadcasting system provided in the second embodiment of the present invention. Detailed Implementation

[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] Reference Figure 1 The first embodiment of the present invention provides a multi-source fusion alarm method for an emergency intelligent broadcasting system, comprising the following steps:

[0030] S11, acquire multi-channel environmental audio data collected by a distributed microphone array, and acquire the target audio signal to be broadcast;

[0031] S12, Spatial energy spectrum analysis is performed on the multi-channel environmental audio data to obtain an environmental noise distribution map characterizing the noise intensity at different coordinate points in space;

[0032] S13, obtain the escape path coordinates of the target area, and discretize the escape path coordinates into multiple consecutive guide nodes;

[0033] S14, calculate the gain compensation parameters for each of the guiding nodes based on the environmental noise distribution map, and use the gain compensation parameters to adjust the gain of the target audio signal to be broadcast, so as to obtain the audio signal to be projected;

[0034] S15, calculate the phase delay parameter based on the spatial geometric relationship between each array unit in the speaker array and the guide node;

[0035] S16, Perform phased beamforming processing on the audio signal to be projected according to the phase delay parameter to obtain a directional projection vector pointing to the guiding node;

[0036] S17, the loudspeaker array is controlled to emit sound waves according to the directional projection vector, forming a spatialized guiding sound field that is dynamically distributed along the escape path with the guiding nodes.

[0037] In step S11, multi-channel environmental audio data collected by a distributed microphone array is acquired, and the target audio signal to be broadcast is acquired.

[0038] In one implementation, this embodiment deploys multiple omnidirectional microphone units on the load-bearing walls and ceiling physical nodes of the target area, and uses a precise time protocol to perform clock synchronization operations on all omnidirectional microphone units to construct the distributed microphone array. This embodiment obtains the upper limit value of the acoustic high-frequency range of the target area, multiplies the upper limit value by a constant value of two to obtain the fundamental Nyquist frequency; from the standard sampling frequency range set configured in the hardware, the smallest range value greater than the fundamental Nyquist frequency is extracted and determined as the preset sampling frequency. This embodiment drives the distributed microphone array to continuously perform analog-to-digital conversion and timestamp alignment marking on the physical changes in sound pressure in the surrounding environment according to the preset sampling frequency, generating the multi-channel environmental audio data containing multi-channel discrete time series.

[0039] It should be noted that in this embodiment, a pre-set alarm digital waveform encoded in pulse code modulation format is read from the storage medium of the emergency command center via the system's internal communication bus, or a real-time digital voice data stream is received from the command and dispatch terminal. This embodiment performs decoding and buffering operations on the pre-set alarm digital waveform or the real-time digital voice data stream to determine it as the target audio signal to be broadcast.

[0040] For example, for an indoor target area with a length and width of 10 meters, this embodiment deploys a distributed microphone array consisting of 25 omnidirectional microphone units in a grid pattern with a spatial spacing of 2 meters within its physical space. After clock synchronization calibration, each omnidirectional microphone unit collects the changes in sound wave amplitude at a preset sampling frequency of 44.1 kHz, outputting multi-channel environmental audio data containing 25 independent timing channels. Simultaneously, the system reads a pre-recorded digital evacuation voice in pulse code modulation format from the memory of the fire control host through the communication interface, extracts it directly, and uses it as the target audio signal to be broadcast.

[0041] In step S12, spatial energy spectrum analysis is performed based on the multi-channel environmental audio data to obtain an environmental noise distribution map characterizing the noise intensity at different coordinate points in space, including:

[0042] The local power spectral density of each sampling point is obtained by performing a fast Fourier transform on each of the environmental audio data streams.

[0043] Obtain the spatial coordinates of each array unit in the microphone array;

[0044] Spatial interpolation fitting is performed based on the local power spectral density and the spatial location coordinates to generate an environmental noise distribution map covering the target area.

[0045] In one implementation, this embodiment preprocesses the acquired multi-channel environmental audio data in discrete time-series form, employing a windowing truncation algorithm to extract observation windows with a length of 1024 sampling points. Within each observation window, this embodiment applies a Hanning window function to suppress spectral leakage and utilizes a Fast Fourier Transform algorithm to map the time-domain signal to the frequency domain, calculating the quotient of the square of the magnitude of each frequency component and the normalization coefficient to obtain the local power spectral density of each sampling point.

[0046] It should be noted that in this embodiment, the absolute values ​​of each omnidirectional microphone unit in the microphone array in the target area in the three-dimensional Cartesian coordinate system are extracted by retrieving the hardware topology mapping table pre-stored in the system memory, and these values ​​are determined as the spatial position coordinates of each array unit in the microphone array.

[0047] It is worth noting that the specific implementation process of generating the environmental noise distribution map using the spatial interpolation algorithm is as follows: In this embodiment, the target area is divided into a discrete grid of points with a step size of 0.2 meters. For each discrete grid point, the predicted sound pressure level is calculated using the inverse distance weighted interpolation (IDW) algorithm. Specifically, this embodiment calculates the spatial distance between the current discrete grid point and all array elements, uses the negative square of the spatial distance as a weighting factor, and performs a weighted average calculation on the local power spectral density corresponding to each array element. This embodiment encapsulates the calculated values ​​of each grid point into a spatial matrix and outputs the environmental noise distribution map.

[0048] For example, for an indoor area of ​​10m × 10m, this embodiment divides it into 50 × 50 computational grids. If the local power spectral density collected by a certain array unit is 65dB after logarithmic transformation, and the unit is located at coordinates (3.5, 4.2) meters, this embodiment calculates its contribution weight to the surrounding grid points. By fitting the energy contributions of 25 collection points, a noise thermal distribution matrix with a resolution of 0.2 meters and containing 2500 sound pressure feature values ​​is generated, thus obtaining the environmental noise distribution map.

[0049] In step S13, the escape path coordinates of the target area are obtained, and the escape path coordinates are discretized into multiple consecutive guide nodes.

[0050] In one implementation, this embodiment retrieves geographic information system data of the target area to obtain three-dimensional topological features including corridors, escape doors, and stairwells. These three-dimensional topological features are then discretized into a three-dimensional grid navigation map containing multiple spatial nodes and connected edges. This embodiment analyzes real-time alarm points from the fire alarm controller and, using these real-time alarm points as centers, updates the passage impedance weights of the corresponding spatial nodes in the three-dimensional grid navigation map according to a preset danger radiation radius.

[0051] It is worth noting that the method for determining the preset dangerous radiation radius is as follows: In this embodiment, the fire source heat release rate parameter fed back by the fire alarm controller is obtained, and a preset fire source radiation heat ratio coefficient and a preset human body heat radiation flux critical threshold are obtained; the fire source heat release rate parameter is multiplied by the preset fire source radiation heat ratio coefficient to obtain the effective radiation power; the human body heat radiation flux critical threshold is multiplied by the constant pi and then by the constant value four to obtain the radiation attenuation denominator; the effective radiation power is divided by the radiation attenuation denominator to obtain the quotient value, and the square root operation is performed on the quotient value to calculate the thermal radiation critical distance that personnel cannot tolerate, and the thermal radiation critical distance is fixed as the preset dangerous radiation radius. The preset fire source radiation heat ratio coefficient is calibrated to 0.3 based on the characteristics of combustibles in historical common building fires; the preset human body heat radiation flux critical threshold is calibrated to 2.5 kilowatts per square meter based on the pain tolerance threshold of human skin in an unprotected state in fire safety engineering.

[0052] This embodiment utilizes the A* path search algorithm, using a weighted sum of the passage impedance weights and the geometric Euclidean distance from the current node to the safety exit as a heuristic cost function, to search the 3D grid navigation map for a node sequence that connects the starting point to the safety exit and minimizes the cumulative value of the heuristic cost function. This embodiment then uses a cubic B-spline interpolation algorithm to perform spatial curve fitting on the node sequence, generating a continuous spatial trajectory curve, which is then used to determine the coordinates of the escape path.

[0053] It should be noted that the process of discretizing the escape path coordinates into multiple consecutive guiding nodes is as follows: In this embodiment, the beamforming resolution parameter of the loudspeaker array is obtained, that is, the minimum coverage diameter that the main lobe of the beam can form within the target coverage area. Specifically, in the anechoic chamber, a test signal is input to the loudspeaker array, and a standard sound level meter is used to measure the sound pressure level at the target distance along a direction perpendicular to the main beam axis. The spatial span corresponding to the decrease in sound pressure level by 3 dB is determined as the minimum coverage diameter of the main lobe of the beam. In this embodiment, the minimum coverage diameter is multiplied by a constant value of 0.5 to obtain the discrete step length. In this embodiment, resampling calculation is performed on the escape path coordinates, and a coordinate sampling value is marked every discrete step length on the continuous spatial trajectory curve to obtain the multiple consecutive guiding nodes.

[0054] It is worth noting that the escape path coordinates are represented as a continuous geometric curve, while at the physical acoustic level, a phased array beam can only focus on a specific point in space at any given time. By transforming the curve characteristics into a discrete sequence of nodes, this embodiment establishes a mathematical mapping from the macroscopic path to the microscopic acoustic wave projection target, thereby supporting the accurate phase delay calculation for a single guiding node in subsequent steps.

[0055] For example, the system detects a fire in the middle of a corridor, and the alarm controller reports a current fire source heat release rate parameter of 5000 kW. The system first multiplies 5000 kW by a fire source radiation heat ratio coefficient of 0.3, obtaining an effective radiation power of 1500 kW. Then, the system multiplies the critical threshold of 2.5 kW per square meter of human body heat radiation flux by 4π, calculating a radiation attenuation denominator of approximately 31.4. The system divides 1500 by 31.4 and takes the square root, calculating a preset danger radiation radius of approximately 6.9 meters. In the A* algorithm, the system sets the passage impedance weight of navigation grid nodes within 6.9 meters of the fire source to infinity, and replans a 30-meter-long corridor escape route avoiding this high-temperature radiation zone. If the main lobe coverage diameter of the speaker array at the corresponding distance is 2 meters, this embodiment multiplies 2 by 0.5 to calculate a distance step length of 1 meter. The system extracts a three-dimensional spatial coordinate point every 1 meter along the fitted curve trajectory, and finally generates a node sequence containing 31 ordered spatial coordinate points, which are then identified as the multiple consecutive guide nodes.

[0056] In step S14, gain compensation parameters for each of the guiding nodes are calculated based on the environmental noise distribution map, and the gain of the target audio signal to be broadcast is adjusted using the gain compensation parameters to obtain the audio signal to be projected, including:

[0057] Extract the local noise level corresponding to the coordinates of each of the guiding nodes in the environmental noise distribution map;

[0058] The required sound pressure level for each guiding node is determined by calculating the local noise level and the preset target signal-to-noise ratio.

[0059] Using an adaptive gain control algorithm, the target audio signal to be broadcast is dynamically compressed and power compensated according to the sound pressure level requirement value to obtain the audio signal to be projected.

[0060] Before adjusting the gain of the audio signal to be projected, the method further includes:

[0061] Acquire the evacuation voice command signal and extract the spectral hole region of the target audio signal to be broadcast;

[0062] Using subband modulation technology, the evacuation voice command signal is shifted to the frequency band corresponding to the spectral hole region;

[0063] The modulated evacuation voice command signal is superimposed with the target audio signal to be broadcast in the frequency domain to obtain a mixed target signal containing voice information.

[0064] In one implementation, before adjusting the gain of the audio signal to be projected, this embodiment acquires analog voice input from the command terminal via a microphone input interface, and generates the evacuation voice command signal through analog-to-digital conversion. This embodiment performs a Fast Fourier Transform on the target audio signal to be broadcast, calculating the energy amplitude of each frequency band; extracts the maximum peak value from the energy amplitude sequence, multiplies the maximum peak value by a constant value of 0.1 to obtain the lower limit of the energy tolerance; and extracts the frequency bands with continuous energy amplitudes less than the lower limit of the energy tolerance across the entire spectrum as the spectral hole region.

[0065] It should be noted that the specific process of shifting the evacuation voice command signal to the frequency band corresponding to the spectral hole region using subband modulation technology is as follows: In this embodiment, a sinusoidal carrier signal is calculated based on the center frequency of the spectral hole region; a low-pass filter is used to perform frequency band limiting calculation on the evacuation voice command signal, limiting its highest frequency to no more than half of the physical bandwidth of the spectral hole region; the frequency band-limited evacuation voice command signal and the sinusoidal carrier signal are multiplied in the time domain to achieve linear shift of the signal spectrum, completing the shift to the spectral hole region. In this embodiment, the shifted signal is added to the target audio signal to be broadcast using complex addition to complete frequency domain superposition processing, generating the mixed target signal, and the mixed target signal is used as the input reference for subsequent gain adjustment.

[0066] In one implementation, this embodiment aligns the three-dimensional spatial coordinates of each guiding node with the spatial grid of the environmental noise distribution map generated in the preceding steps, extracts the sound pressure level values ​​recorded at grid points where the coordinates completely coincide or the Euclidean distance is closest, and extracts them as the local noise level; the local noise level is obtained by integrating the power spectral density of the corresponding coordinate point in the environmental noise distribution map within the audible frequency band of 20 Hz to 20 kHz, and then converting it into a decibel scale value after correction by an A-weighted network, thus obtaining a single-value noise level characterizing the subjective loudness of the human ear.

[0067] It is worth noting that the specific logic for determining the required sound pressure level for each guiding node based on the local noise level and the preset target signal-to-noise ratio is as follows: In this embodiment, the extracted local noise level and the preset target signal-to-noise ratio are subjected to arithmetic addition, and the sum is extracted as the required sound pressure level. The preset target signal-to-noise ratio is determined using an engineering calibration method. Different levels of pink noise are injected into the acoustic test chamber to simulate the background sound of a fire, and the speech dictation pass rate data of the test subjects under different signal-to-noise ratio conditions are collected. A continuous correlation curve between the signal-to-noise ratio and the pass rate is plotted, and the first derivative sequence of the continuous correlation curve is calculated. The curve is traversed along the direction of increasing signal-to-noise ratio to identify the first data point where the first derivative sequence shows a monotonically decreasing trend after crossing the maximum value point, and the corresponding pass rate exceeds 85%. The signal-to-noise ratio value corresponding to this point is fixed as the preset target signal-to-noise ratio.

[0068] In one implementation, this embodiment utilizes an adaptive gain control algorithm to perform dynamic range compression and power compensation processing on the generated hybrid target signal based on the sound pressure level (SPL) requirement value. Specifically, this embodiment obtains the electroacoustic conversion reference value of the loudspeaker array. The electroacoustic conversion reference value is determined using an offline acoustic calibration method. A broadband test signal with a full-scale digital level is input to the loudspeaker array in an anechoic chamber, and the output physical SPL is measured at a standard reference distance using a standard sound level meter. The measured physical SPL value is then fixed as the electroacoustic conversion reference value. This embodiment subtracts the electroacoustic conversion reference value from the SPL requirement value and performs an arithmetic subtraction operation to obtain the target digital level value. This embodiment calculates the root mean square (RMS) level value of the hybrid target signal within a sliding time window; the target digital level value is subtracted from the RMS level value, and an arithmetic subtraction operation is performed to obtain the initial gain compensation parameter. This embodiment inputs the initial gain compensation parameter into a smoothing filter containing start-up and release time parameters for iterative calculation to generate a smoothed final gain multiplier. In this embodiment, the time-domain discrete amplitude sequence of the mixed target signal is multiplied by the final gain multiplier, and a hard-limiting truncation operation is performed when the product value exceeds the maximum range limit of the system's digital-to-analog converter, completing dynamic range compression and power compensation processing, and the output is the audio signal to be projected. The start-up time parameter is specifically determined to be ten milliseconds, and the release time parameter is specifically determined to be two hundred milliseconds.

[0069] For example, suppose the local noise level at a guide node in an escape route is extracted as 75 dB from a heatmap, and the preset target signal-to-noise ratio (SNR) determined by optimizing the speech dictation pass rate curve is 15 dB. In this embodiment, 75 dB and 15 dB are added together to calculate the required sound pressure level (SPL) of 90 dB for the guide node. The system determines the electroacoustic conversion reference value of the current speaker array to be 110 dB (i.e., the physical SPL corresponding to a full-scale digital signal output of 110 dB) through offline acoustic calibration. The system subtracts 110 dB from 90 dB to calculate the target digital level as -20 dB full-scale digital level. The system measures the root mean square (RMS) level of the mixed target signal generated by frequency domain hole embedding superposition as -30 dB full-scale digital level. In this embodiment, -20 dB is then subtracted from -30 dB to calculate the initial gain compensation parameter as 10 dB. Using an adaptive gain control algorithm and a 10-millisecond startup time parameter, the system smoothly amplifies the 10 dB gain and applies it to the waveform of the mixed audio, and finally truncates and clips the peaks, outputting the audio signal to be projected after electroacoustic domain logic matching and gain processing.

[0070] In step S15, the phase delay parameter is calculated based on the spatial geometric relationship between each array unit in the speaker array and the guide node, including:

[0071] Calculate the spatial Euclidean distance between each array unit in the loudspeaker array and the currently selected guide node;

[0072] The minimum value among the spatial Euclidean distances is selected as the reference distance value;

[0073] The path difference is obtained by arithmetically subtracting the Euclidean distance of each path from the reference distance value, and the path difference is divided by the preset sound speed constant to obtain the phase delay parameter.

[0074] In one implementation, this embodiment obtains the first physical coordinates of each array unit in the speaker array in the three-dimensional Cartesian coordinate system of the target area, and extracts the second physical coordinates of the currently selected guide node. This embodiment calculates the numerical difference between each of the first and second physical coordinates on the three corresponding coordinate axes; calculates the sum of the squares of these three numerical differences, and performs a square root operation on the sum of squares to obtain the shortest geometric straight-line length of the sound wave emitted by each array unit to reach the target node, which is then determined as the spatial Euclidean distance between each array unit in the speaker array and the currently selected guide node.

[0075] It should be noted that in this embodiment, the calculated spatial Euclidean distances corresponding to all array units are used to form a one-dimensional distance sequence. The distance sequence is sorted in ascending order of numerical value, and the smallest physical distance value at the beginning is extracted and determined as the reference distance value. Subsequently, this embodiment iterates through the distance sequence, subtracting the reference distance value from each spatial Euclidean distance, performing arithmetic subtraction to obtain the spatial length difference, and extracting this as the path difference.

[0076] It is worth noting that the preset sound velocity constant is determined using an environmental calibration method. Specifically, in this embodiment, the real-time ambient temperature value is read by a digital temperature sensor configured within the target area; the real-time ambient temperature value is multiplied by a constant value of 0.6, and the resulting product is arithmetically added to the constant value of 331.4 to calculate the temperature-corrected sound wave propagation speed value, which is then fixed as the preset sound velocity constant. In this embodiment, the path difference corresponding to each array unit is divided by the preset sound velocity constant, and the quotient obtained by performing arithmetic division represents the time difference compensation for the sound wave arriving at the node, which is then uniformly extracted as the phase delay parameter.

[0077] For example, assuming the loudspeaker array contains three array units numbered A, B, and C, the system calculates their spatial Euclidean distances to the currently selected guide node to be 8.5 meters, 9.0 meters, and 9.2 meters, respectively. This embodiment compares this distance sequence and selects the minimum value of 8.5 meters as the reference distance. By performing arithmetic subtraction, the path differences of array units A, B, and C are calculated to be 0 meters, 0.5 meters, and 0.7 meters, respectively. If the digital temperature sensor measures a real-time ambient temperature of 20 degrees Celsius, this embodiment multiplies 20 by 0.6 to obtain 12, adds 331.4, and calculates the preset sound velocity constant to be 343.4 meters per second. Finally, this embodiment divides the path difference of each array unit by 343.4 meters per second, calculating the phase delay parameters corresponding to array units A, B, and C to be approximately 0 seconds, 0.00145 seconds, and 0.00203 seconds, respectively.

[0078] In step S16, phased beamforming is performed on the audio signal to be projected according to the phase delay parameter to obtain a directional projection vector pointing to the guiding node, including:

[0079] Based on the phase delay parameter, the audio signal to be projected is subjected to time delay compensation or phase shift processing to obtain channel signals with phase gradients for each channel.

[0080] The channel signals are weighted to suppress sidelobe energy during the phased beamforming process.

[0081] The weighted signals are combined into the directional projection vector.

[0082] After the directional projection vector pointing to the guiding node is generated, the process further includes:

[0083] Using a preset acoustic simulation model, the predicted sharpness index at the guide node is calculated based on the current directional projection vector;

[0084] When the predicted sharpness index is lower than the preset sharpness threshold, the beamwidth parameter in the directional projection vector is iteratively corrected using a gradient descent algorithm until the predicted sharpness index reaches the preset sharpness threshold.

[0085] In one implementation, this embodiment extracts the phase delay parameter generated in the preceding steps and the audio signal to be projected. For each array unit in the speaker array, this embodiment allocates an independent digital buffer queue in the memory of the digital signal processing module. This embodiment sequentially inputs the discrete sampling sequence of the audio signal to be projected into each of the digital buffer queues, and calculates the required number of sampling points for delay based on the corresponding phase delay parameter. Specifically, the calculation method is to multiply the phase delay parameter by the system sampling frequency and perform a floor operation. Subsequently, this embodiment controls each digital buffer queue to perform a translation delay output operation according to the calculated number of sampling points, generating the channel signals with phase gradients.

[0086] It should be noted that the specific logic for weighting the channel signals is as follows: In this embodiment, a Chebyshev window function is used to calculate the amplitude weighting coefficient of each array unit. In this embodiment, the amplitude weighting coefficient of the unit at the center of the array is set to a value of one, and it decreases symmetrically towards both edges along the physical array. In this embodiment, the instantaneous amplitude of the channel signal corresponding to each array unit is multiplied by its corresponding amplitude weighting coefficient to complete the weighting of the channel signals. This multiplication operation directly suppresses the sidelobe energy of the synthesized sound field at the acoustic interference physics level. Subsequently, in this embodiment, the data channels of the weighted signals are stacked and spliced ​​into a matrix to construct a multidimensional array containing multi-channel amplitude and phase characteristics, which is then extracted as the directional projection vector.

[0087] It is worth noting that the preset acoustic simulation model is an acoustic evolution space pre-established based on a three-dimensional ray tracing algorithm. Its construction process involves importing a three-dimensional building information model of the target area and obtaining a pre-set acoustic reflection coefficient dictionary. The construction process of the pre-set acoustic reflection coefficient dictionary involves obtaining samples of various common building materials, measuring the absorption and reflection coefficients of each material sample in the standard octave band using impedance tube measurement, associating the measured absorption and reflection coefficients with corresponding material name tags, and storing them in a relational database to generate the pre-set acoustic reflection coefficient dictionary. The absorption and reflection coefficients of each building material are stored separately in one-third octave bands, with center frequencies ranging from 100 Hz to 5 kHz, encompassing a total of eighteen bands. In this embodiment, the pre-set acoustic reflection coefficient dictionary is used to assign sound absorption coefficient values ​​to the wall and floor materials in the three-dimensional building information model, generating the preset acoustic simulation model. In this embodiment, the generated directional projection vector is used as the initial emission source. The process of the sound wave ray reaching the guide node position is tracked in the preset acoustic simulation model, and its impulse response sequence is extracted. Based on the impulse response sequence, this embodiment calculates the early effective acoustic energy value within the first 50 milliseconds and the later reverberant acoustic energy value after 50 milliseconds. The early effective acoustic energy value is divided by the later reverberant acoustic energy value to obtain the energy ratio. A base-10 logarithmic operation is performed on this energy ratio, and the logarithmic result is multiplied by a constant value of 10. The calculated value is extracted as the predicted sharpness index.

[0088] It is worth noting that, when calculating the predicted sharpness index, the preset acoustic simulation model divides the audible frequency band into several octave bands or one-third octave bands. Ray tracing is performed on each sub-band, and the early sound energy and late reverberation sound energy within the sub-band are calculated. The early sound energy of all sub-bands is accumulated and divided by the late reverberation sound energy of all sub-bands, and then the common logarithm is multiplied by ten to obtain the predicted sharpness index. In the acoustic simulation model based on the three-dimensional ray tracing algorithm, the energy attenuation of each sound wave ray is tracked until its energy is lower than one-thousandth of the initial energy. The number of rays emitted in each emission direction is not less than one thousand, and the maximum number of reflections is set to ten.

[0089] In one implementation, the preset intelligibility threshold is determined using a subjective-objective calibration method. Specifically, the lowest objective C50 parameter value with a listener subjective speech intelligibility score of 90% or higher is statistically analyzed from historical test databases and determined as the preset intelligibility threshold. This embodiment compares the predicted intelligibility index with the preset intelligibility threshold. When the predicted intelligibility index is lower than the preset intelligibility threshold, this embodiment constructs a loss function with the goal of maximizing the predicted intelligibility index and uses a gradient descent algorithm to iteratively correct the beamwidth parameter in the directional projection vector. The beamwidth parameter is defined as the negative 3 dB width of the main lobe of the sound wave formed by the directional projection vector in spatial angle. By adjusting the sidelobe attenuation level of the Chebyshev window function, the distribution of the amplitude weighting coefficients of each unit in the array is changed, thereby making the main lobe width change monotonically, achieving the mapping from the beamwidth parameter to the directional projection vector.

[0090] Specifically, this embodiment uses the finite difference method to calculate the numerical gradient of the loss function with respect to the beamwidth parameter. This embodiment applies a preset small perturbation increment to the current beamwidth parameter, and then re-inputs the beamwidth parameter with this small perturbation increment into the preset acoustic simulation model for simulation calculation, obtaining the perturbed predicted sharpness index. The arithmetic difference between the perturbed predicted sharpness index and the predicted sharpness index before the perturbation is calculated, and this arithmetic difference is divided by the small perturbation increment; the quotient is extracted as the numerical gradient. The small perturbation increment is fixed at 0.001 degrees, and this value is preset based on the minimum resolvable change of the beamwidth parameter in a typical acoustic simulation scenario, used in the finite difference method to apply a small change to the current beamwidth parameter to calculate the numerical gradient.

[0091] Subsequently, this embodiment performs a numerical correction operation in the inverse gradient direction on the beamwidth parameter according to a preset learning rate step size. The preset learning rate step size is determined using an adaptive attenuation strategy. This embodiment obtains an initial learning rate benchmark value and a preset improvement lower limit threshold. The initial learning rate benchmark value is determined using an offline simulation calibration method. In a historical standard test sound field model, a set of candidate learning rates with a logarithmic distribution is applied to the beamwidth parameter for single-step gradient descent testing; a set of candidate learning rates that do not cause a divergence in the predicted sharpness index value is extracted, and the maximum value in this set is fixed as the initial learning rate benchmark value. The preset improvement lower limit threshold is determined using a statistical analysis method. A statically fixed directional projection vector is input into the preset acoustic simulation model, and multiple independent ray tracing simulation calculations are performed to extract the fluctuation standard deviation of the predicted sharpness index caused by the algorithm's random seed differences; the fluctuation standard deviation is multiplied by a constant value of three, and the resulting product value is fixed as the preset improvement lower limit threshold to define the real physical improvement brought about by parameter optimization and the system simulation noise floor.

[0092] After each iteration of correction, this embodiment calculates the improvement of the current predicted sharpness index compared to the predicted sharpness index of the previous iteration. If the improvement is less than the preset lower limit threshold, the current learning rate step size is multiplied by a constant value of 0.5 to obtain a decayed value, which is then updated to the preset learning rate step size to suppress optimization oscillations during gradient descent. If the improvement is greater than or equal to the preset lower limit threshold, the current learning rate step size remains unchanged, and the next iteration calculation continues. In this embodiment, the corrected beamwidth parameters are re-input into the preset acoustic simulation model for calculation, and the above numerical gradient solution and parameter correction operations are repeated until the recalculated predicted sharpness index is greater than or equal to the preset sharpness threshold. If the predicted sharpness index is greater than or equal to the preset sharpness threshold, the current directional projection vector is directly locked and output to the execution deployment module.

[0093] For example, suppose the system drives 16 array elements to form a beam for a specific guide node through delay control. The system applies a Chebyshev window function to assign a weight of 1.0 to the middle elements and a weight of 0.3 to the edge elements at both ends to generate an initial directional projection vector. The system inputs this vector into a ray tracing model built based on an acoustic reflection coefficient dictionary for sound field simulation, calculating a predicted sharpness index of 1.5 dB based on the acoustic energy comparison between the first 50 milliseconds and the last 50 milliseconds at the guide node. Since 1.5 dB is lower than the preset sharpness threshold of 2.0 dB obtained from historical calibration, the system automatically triggers the gradient descent algorithm. The system uses a small perturbation increment of 0.001 to calculate the numerical gradient using the finite difference method. The system retrieves an initial learning rate baseline of 0.05 determined by offline maximum convergence step size calibration, and a preset improved lower limit threshold of 0.1 dB determined by 3-sigma noise floor measurement. The system narrows the beamwidth by adjusting the edge attenuation coefficient of the window function with a step size of 0.05. In the second iteration, if the improvement in sharpness is found to be only 0.08 dB (less than 0.1 dB), the system halves the learning rate step size to 0.025. After five iterations of differential optimization and step size attenuation, the ineffective reflected sound energy from the sidewalls is significantly reduced, and the predicted sharpness index calculated by resimulation improves to 2.2 dB, reaching the preset sharpness threshold requirement. The system then locks the current beam parameters and ends the iteration operation.

[0094] In step S17, the loudspeaker array is controlled to emit sound waves according to the directional projection vector, forming a spatialized guiding sound field that is dynamically distributed along the escape path with the guiding nodes.

[0095] The spatialized guiding sound field, dynamically distributed along the guide nodes on the escape path, further includes:

[0096] The current guiding node is switched in a polling manner according to a preset scanning frequency;

[0097] The phase delay parameter is updated in real time according to the switched guidance node to drive the direction of the directional projection vector to move sequentially with the escape path.

[0098] After controlling the loudspeaker array to emit sound waves according to the directional projection vector, the method further includes:

[0099] Acquire real-time sound pressure data transmitted from feedback sensors deployed near the guidance node;

[0100] Calculate the deviation between the real-time sound pressure data and the expected sound pressure value;

[0101] The gain compensation parameter is corrected in a closed loop based on the deviation value.

[0102] In one implementation, this embodiment concurrently sends the directional projection vector determined in the preceding steps to each power amplifier module of the speaker array via a high-speed data bus. This embodiment drives the power amplifier modules to amplify the analog signal according to the gain value corresponding to each channel in the directional projection vector, and feeds the amplified signal into the array unit at the corresponding position. At the physical space level, each sound wave forms a high sound pressure level energy focus at the coordinates of the guiding node through constructive interference, resulting in the spatialized guiding sound field.

[0103] It is worth noting that the method for determining the preset scanning frequency is as follows: In this embodiment, the average escape movement speed of trapped personnel in an emergency is obtained, specifically set to 1.5 meters per second. This is based on the average walking speed of personnel rapidly evacuating in a smoke environment according to national emergency evacuation standards, and can be adjusted within the range of 0.8 meters per second to 2.0 meters per second depending on the on-site personnel density and building type. The average step distance between two adjacent guide nodes on the escape path is extracted. In this embodiment, the average escape movement speed is divided by the average step distance, and the quotient obtained is determined as the preset scanning frequency. In this embodiment, an index queue pointing to the multiple consecutive guide nodes is constructed. In this embodiment, according to the preset scanning frequency, a switching command is periodically sent to the index queue to extract the spatial coordinates of the next guide node, and the phase delay calculation and beamforming operation in the preceding steps are repeated. By continuously updating the initial phase of each array unit, the sound beam focus is driven to shift sequentially along the trajectory of the escape path, forming a dynamically moving voice guidance effect in terms of auditory perception.

[0104] It should be noted that the process of determining the expected sound pressure level is as follows: extract the sound pressure level requirement value determined in the preceding step S14; calculate the distance attenuation loss value generated by the sound wave propagating from the speaker position to the corresponding guide node position according to the preset acoustic simulation model; subtract the distance attenuation loss value from the sound pressure level requirement value, and the difference obtained is determined as the expected sound pressure level value. In this embodiment, the feedback sensor (such as a high-sensitivity capacitive microphone) is used to collect the sound wave amplitude at the guide node in real time and convert it into real-time sound pressure data in decibels.

[0105] It is worth noting that the specific implementation of closed-loop correction of the gain compensation parameter based on the deviation value is as follows: the expected sound pressure value is subtracted from the real-time sound pressure data, and the deviation value is obtained by performing an arithmetic subtraction operation. In this embodiment, a proportional-integral (PI) control algorithm is used to process the deviation value. The deviation value is multiplied by a preset proportional coefficient to obtain a proportional term, and the deviation value is accumulated over time and then multiplied by a preset integral coefficient to obtain an integral term. This embodiment sums the proportional term and the integral term to obtain the gain correction amount, and then adds the gain correction amount to the original gain compensation parameter to complete the closed-loop correction.

[0106] It is worth noting that in the closed-loop correction of the gain compensation parameter based on the deviation value, the corrected gain compensation parameter does not change the sound wave that the current guiding node has already emitted. Instead, it is stored and applied to the gain adjustment calculation of the next polling cycle of the same guiding node or subsequent guiding nodes, ensuring that the system converges to the expected sound pressure level point by point during continuous scanning.

[0107] It should be noted that the preset proportional coefficient and the preset integral coefficient are determined by calibration using the Ziegler-Nichols closed-loop oscillation method under offline testing. The specific implementation process is as follows: a loudspeaker array and feedback sensor are deployed in the offline acoustic test field; the integral coefficient is forcibly set to zero in the system control program, retaining only the pure proportional feedback control link; a steady-state broadband test signal is input to the system, and a fixed target sound pressure level is set, while the temporary proportional coefficient is gradually increased with a preset gain step size; after each increase of the temporary proportional coefficient, the temporal fluctuation pattern of the real-time sound pressure data returned by the feedback sensor is continuously monitored; when the real-time sound pressure data first exhibits waveform characteristics of continuous oscillation with constant amplitude around the target sound pressure level, the system stops increasing the proportional coefficient; in this embodiment, the temporary proportional coefficient at this time is extracted as the critical proportional gain parameter, and the time difference between two adjacent peaks in the continuous oscillation waveform with constant amplitude is measured and extracted as the critical oscillation period parameter.

[0108] Subsequently, this embodiment determines the parameters according to the Ziegler-Nichols empirical calculation rule. The critical proportional gain parameter is multiplied by a constant value of 0.45, and the calculated result is fixed as the preset proportional coefficient. The critical oscillation period parameter is multiplied by a constant value of 0.83 to calculate the integral time constant. Finally, this embodiment divides the calculated preset proportional coefficient by the integral time constant, and the quotient obtained by performing arithmetic division is fixed as the preset integral coefficient. This embodiment writes the calibrated preset proportional coefficient and the preset integral coefficient into the non-volatile memory of the execution deployment module for online closed-loop correction in a real-time environment.

[0109] For example, assuming the distance between adjacent guide nodes is 1.5 meters, and based on an average escape movement speed of 1.5 meters per second, the system determines the preset scanning frequency to be 1 Hz. Every second, the system automatically switches to the next guide node and updates the phase of each speaker, causing the sound wave focus to move with the flow of people. If the system's preset sound pressure level requirement is 90 dB and the propagation attenuation is 10 dB, then the expected sound pressure level is 80 dB. If the real-time sound pressure data returned by the feedback sensor deployed at the node is 77.5 dB, then the calculated deviation is -2.5 dB. Using the PI algorithm, the system calculates that an additional 3.2 dB gain correction is needed and adds it to the gain compensation parameters, driving the speakers to increase their output level, thus restoring the actual sound pressure level at the node to 80 dB.

[0110] In summary, this invention constructs a high-precision environmental noise distribution map using a distributed microphone array, and combines discretized node processing of escape paths with phase delay-based phased beamforming technology to achieve spatially precise directional projection of evacuation commands in complex, high-noise environments. By introducing sub-band modulation technology and a closed-loop correction mechanism using feedback sensors, this invention ensures the clarity and intelligibility of voice commands while endowing the system with the ability to adaptively adjust in real time according to changes in the ambient sound field. This completely breaks through the technical bottlenecks of traditional broadcast omnidirectional diffusion and path guidance failure, greatly improving evacuation efficiency and safety in emergency scenarios.

[0111] Reference Figure 2 The second embodiment of the present invention provides a multi-source fusion alarm system for an emergency intelligent broadcasting system, comprising:

[0112] The data acquisition module is used to acquire multi-channel environmental audio data collected by a distributed microphone array and to acquire the target audio signal to be broadcast.

[0113] The noise analysis module is used to perform spatial energy spectrum analysis based on the multi-channel environmental audio data to obtain an environmental noise distribution map characterizing the noise intensity at different coordinate points in space.

[0114] The path discretization module is used to obtain the escape path coordinates of the target area and discretize the escape path coordinates into multiple consecutive guide nodes;

[0115] The strategy adjustment module is used to calculate the gain compensation parameters for each of the guiding nodes based on the environmental noise distribution map, and to adjust the gain of the target audio signal to be broadcast using the gain compensation parameters to obtain the audio signal to be projected.

[0116] The delay calculation module is used to calculate the phase delay parameter based on the spatial geometric relationship between each array unit in the speaker array and the guide node;

[0117] A beamforming module is used to perform phased beamforming processing on the audio signal to be projected according to the phase delay parameter to obtain a directional projection vector pointing to the guiding node;

[0118] The deployment module is used to control the loudspeaker array to emit sound waves according to the directional projection vector, forming a spatialized guiding sound field that is dynamically distributed along the escape path with the guiding nodes.

[0119] It should be noted that the multi-source fusion alarm system for emergency intelligent broadcasting system provided in this embodiment of the invention executes all the process steps of the multi-source fusion alarm method for emergency intelligent broadcasting system in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0120] It should be noted that the system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the system embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0121] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A multi-source fusion alarm method for an emergency intelligent broadcasting system, characterized in that, include: Acquire multi-channel environmental audio data collected by a distributed microphone array, and acquire the target audio signal to be broadcast; Spatial energy spectrum analysis is performed on the multi-channel environmental audio data to obtain an environmental noise distribution map characterizing the noise intensity at different coordinate points in space. Obtain the escape path coordinates of the target area, and discretize the escape path coordinates into multiple consecutive guide nodes; The gain compensation parameters for each of the guiding nodes are calculated based on the environmental noise distribution map, and the gain of the target audio signal to be broadcast is adjusted using the gain compensation parameters to obtain the audio signal to be projected. The phase delay parameter is calculated based on the spatial geometric relationship between each array unit in the speaker array and the guide node; The phased beamforming process is performed on the audio signal to be projected according to the phase delay parameter to obtain a directional projection vector pointing to the guiding node; The speaker array is controlled to emit sound waves according to the directional projection vector, forming a spatialized guiding sound field that is dynamically distributed along the guide node on the escape path. The step of forming a spatialized guiding sound field that is dynamically distributed along the escape path with the guiding nodes further includes: polling and switching the current guiding node according to a preset scanning frequency; and updating the phase delay parameter in real time according to the switched guiding node to drive the direction of the directional projection vector to move sequentially along the escape path. The process of controlling the loudspeaker array to emit sound waves according to the directional projection vector further includes: acquiring real-time sound pressure data transmitted from feedback sensors deployed near the guiding node; calculating the deviation between the real-time sound pressure data and the expected sound pressure value; and performing closed-loop correction on the gain compensation parameter based on the deviation value.

2. The multi-source fusion alarm method for an emergency intelligent broadcasting system according to claim 1, characterized in that, The step of performing spatial energy spectrum analysis based on the multi-channel environmental audio data to obtain an environmental noise distribution map characterizing the noise intensity at different coordinate points in space includes: The local power spectral density of each sampling point is obtained by performing a fast Fourier transform on each of the environmental audio data streams. Obtain the spatial coordinates of each array unit in the microphone array; Spatial interpolation fitting is performed based on the local power spectral density and the spatial location coordinates to generate an environmental noise distribution map covering the target area.

3. The multi-source fusion alarm method for an emergency intelligent broadcasting system according to claim 1, characterized in that, The step of calculating gain compensation parameters for each of the guiding nodes based on the environmental noise distribution map, and adjusting the gain of the target audio signal to be broadcast using the gain compensation parameters to obtain the audio signal to be projected, includes: Extract the local noise level corresponding to the coordinates of each of the guiding nodes in the environmental noise distribution map; The required sound pressure level for each guiding node is determined by calculating the local noise level and the preset target signal-to-noise ratio. Using an adaptive gain control algorithm, the target audio signal to be broadcast is dynamically compressed and power compensated according to the sound pressure level requirement value to obtain the audio signal to be projected.

4. The multi-source fusion alarm method for an emergency intelligent broadcasting system according to claim 1, characterized in that, The calculation of the phase delay parameter based on the spatial geometric relationship between each array unit in the speaker array and the guide node includes: Calculate the spatial Euclidean distance between each array unit in the loudspeaker array and the currently selected guide node; The minimum value among the spatial Euclidean distances is selected as the reference distance value; The path difference is obtained by arithmetically subtracting the Euclidean distance of each path from the reference distance value, and the path difference is divided by the preset sound speed constant to obtain the phase delay parameter.

5. The multi-source fusion alarm method for an emergency intelligent broadcasting system according to claim 1, characterized in that, The step of performing phased beamforming processing on the audio signal to be projected based on the phase delay parameter to obtain a directional projection vector pointing to the guiding node includes: Based on the phase delay parameter, the audio signal to be projected is subjected to time delay compensation or phase shift processing to obtain channel signals with phase gradients for each channel. The channel signals are weighted to suppress sidelobe energy during the phased beamforming process. The weighted signals are combined into the directional projection vector.

6. The multi-source fusion alarm method for an emergency intelligent broadcasting system according to claim 1, characterized in that, Before adjusting the gain of the target audio signal to be broadcast, the method further includes: Acquire the evacuation voice command signal and extract the spectral hole region of the target audio signal to be broadcast; Using subband modulation technology, the evacuation voice command signal is shifted to the frequency band corresponding to the spectral hole region; The modulated evacuation voice command signal is superimposed with the target audio signal to be broadcast in the frequency domain to obtain a mixed target signal containing voice information.

7. The multi-source fusion alarm method for an emergency intelligent broadcasting system according to claim 1, characterized in that, After the directional projection vector pointing to the guiding node is generated, the process further includes: Using a preset acoustic simulation model, the predicted sharpness index at the guide node is calculated based on the current directional projection vector; When the predicted sharpness index is lower than the preset sharpness threshold, the beamwidth parameter in the directional projection vector is iteratively corrected using a gradient descent algorithm until the predicted sharpness index reaches the preset sharpness threshold.

8. A multi-source fusion alarm system for an emergency intelligent broadcasting system, characterized in that, include: The data acquisition module is used to acquire multi-channel environmental audio data collected by a distributed microphone array and to acquire the target audio signal to be broadcast. The noise analysis module is used to perform spatial energy spectrum analysis based on the multi-channel environmental audio data to obtain an environmental noise distribution map characterizing the noise intensity at different coordinate points in space. The path discretization module is used to obtain the escape path coordinates of the target area and discretize the escape path coordinates into multiple consecutive guide nodes; The strategy adjustment module is used to calculate the gain compensation parameters for each of the guiding nodes based on the environmental noise distribution map, and to adjust the gain of the target audio signal to be broadcast using the gain compensation parameters to obtain the audio signal to be projected. The delay calculation module is used to calculate the phase delay parameter based on the spatial geometric relationship between each array unit in the speaker array and the guide node; A beamforming module is used to perform phased beamforming processing on the audio signal to be projected according to the phase delay parameter to obtain a directional projection vector pointing to the guiding node; The execution deployment module is used to control the speaker array to emit sound waves according to the directional projection vector, forming a spatialized guiding sound field that is dynamically distributed along the escape path with the guiding nodes. The step of forming a spatialized guiding sound field that is dynamically distributed along the escape path with the guiding nodes further includes: polling and switching the current guiding node according to a preset scanning frequency; and updating the phase delay parameter in real time according to the switched guiding node to drive the direction of the directional projection vector to move sequentially along the escape path. The process of controlling the loudspeaker array to emit sound waves according to the directional projection vector further includes: acquiring real-time sound pressure data transmitted from feedback sensors deployed near the guiding node; calculating the deviation between the real-time sound pressure data and the expected sound pressure value; and performing closed-loop correction on the gain compensation parameter based on the deviation value.

Images