Flight state information judgment method and system based on cockpit audio data analysis

Abstract

The invention discloses a flight state information judgment method and system based on cockpit audio data analysis, and the method comprises the steps: 1, obtaining the audio data in a cockpit through a cockpit recorder or / and a protection recorder; step 2, performing data preprocessing and noise suppression processing on the audio data in the cockpit so as to convert the non-language information into sampling points with specified sampling speeds, extracting the audio data corresponding to each sampling point, and performing noise suppression processing on the extracted audio data; and step 3, detecting the category and the accurate time of each alarm sound by adopting an alarm sound detection operation law, and evaluating the engine torque by utilizing an engine torque evaluation operation law. According to the technical scheme, the problem that accurate sound extraction is difficult to realize due to the fact that audios in the cockpit are various in category, different in feature and doped with various noises in an existing analysis mode for audio data in the helicopter cockpit is solved.

Description

Technical Field

[0001] The present invention relates to the field of maintenance and support technology, but is not limited to it, and in particular to a method and system for determining flight status information based on cockpit audio data analysis. Background Art

[0002] Safety is an extremely important indicator requirement for aviation equipment. The aviation industry at home and abroad has been committed to improving the safety of helicopter use, but the accident rate of helicopters is still higher than that of fixed-wing aircraft. For example, in 2010, the number of helicopters accounted for 12% of the total aircraft fleet, but its accident rate accounted for 71.26% of the total aircraft accident rate. What is even more shocking is that 75% of helicopter accidents are caused by human error.

[0003] Due to the high accident rate of helicopters, a large number of studies have been carried out at home and abroad to understand the specific state of the cockpit when a helicopter incident / accident occurs, especially to analyze the cockpit audio data to supplement the work of the flight data recorder. The cockpit audio data includes various sounds, such as the conversations of the flight crew, the alarm sounds, the movement of the control surfaces or the operating surfaces, the triggering sounds of switches and the sounds of the engine, etc. The various types of sounds in the helicopter cockpit have relatively unique characteristics, such as frequency and duration, etc. In addition, these sounds are also mixed with various noises, such as wind noise. The above factors have become the reasons for restricting the analysis of cockpit audio data.

[0004] At present, the industry has been using traditional methods such as subjective "identification and hearing" and single signal analysis to monitor status. However, traditional methods are not easy to accurately extract sound. Therefore, how to accurately identify flight status information is extremely important to improving equipment safety and is also an urgent technical breakthrough. Summary of the invention

[0005] Purpose of the present invention: In order to solve the above-mentioned problems, the embodiments of the present invention provide a method and system for determining flight status information based on cockpit audio data analysis, so as to solve the problem that it is difficult to accurately extract sound due to the large number of audio categories, different characteristics, and various noises in the cockpit.

[0006] Technical solution of the present invention: An embodiment of the present invention provides a method for determining flight status information based on cockpit audio data analysis, comprising:

[0007] Step 1: Use the cockpit voice recorder and / or protection recorder to obtain the audio data in the cockpit;

[0008] Step 2, performing data preprocessing and noise suppression processing on the audio data in the cockpit to convert the non-language information into sampling points with a specified sampling rate, extracting the audio data corresponding to each sampling point, and performing noise suppression processing on the extracted audio data;

[0009] Step 3: Use the warning sound detection operation law to detect the type and precise time of each warning sound, and use the engine torque evaluation operation law to perform engine torque evaluation.

[0010] Optionally, in the flight status information determination method based on cockpit audio data analysis as described above, the method of obtaining the audio data in the cockpit in step 1 includes:

[0011] Method 1, using a cockpit voice recorder to obtain audio data in the cockpit, including: four channels in the cockpit voice recorder work simultaneously, channel 1 records non-verbal information, that is, the warning sound in the cockpit area, channel 2 and channel 3 record the voice of the pilot and co-pilot respectively, and channel 4 records the sound information of the radio panel;

[0012] Method 2, using a protection recorder to obtain audio data in the cockpit, includes: the protection recorder has one channel working to record all sounds in the cockpit, and the recorded sounds include alarms and engine sounds in the audio data.

[0013] Optionally, in the flight status information determination method based on cockpit audio data analysis as described above, the data preprocessing method in step 2 includes:

[0014] Step 21, dividing the audio data into language information and non-language information: the language information includes the voices of the positive and negative drivers, and the non-language information includes other noise information and background sound;

[0015] Step 22, information conversion preprocessing: convert the non-language information into a set of sampling points with a specified sampling rate, and extract the audio data corresponding to each sampling point.

[0016] Optionally, in the flight status information determination method based on cockpit audio data analysis as described above, the noise suppression processing method in step 2 includes:

[0017] Step 23, using high-pass and low-pass filters to suppress noise on the audio data of each sampling point extracted to retain the audio data within a specified frequency range;

[0018] Among them, filtered data is used for warning sound detection, and unfiltered audio data is used for engine torque estimation.

[0019] Optionally, in the flight status information determination method based on cockpit audio data analysis as described above, detecting each warning sound in step 3 includes:

[0020] For warning sound detection, the category of the warning sound is identified by the highest correlation between the warning sound in the acquired cockpit audio data and the sound of the warning sound database generated using the warning characteristics.

[0021] Optionally, in the flight status information determination method based on cockpit audio data analysis as described above, the engine torque evaluation in step 3 includes:

[0022] For engine torque evaluation, the correlation between the engine sound main frequency range and the known engine parameters in the noise and uncertainty range is obtained to determine the main frequency of the engine.

[0023] Optionally, in the flight status information determination method based on cockpit audio data analysis as described above, the manner of detecting each warning sound in step 3 includes:

[0024] S31, identifying features of the alarm sound, including: using Fourier transform and high power spectral density to identify features of each alarm sound in an ideal environment, and detecting the occurrence time of each alarm sound in the ideal environment; before analyzing the alarm sound data, analyzing features of each alarm sound in the ideal environment to generate an alarm sound database;

[0025] S32, performing frequency domain analysis on the alarm sound, obtaining the types of the collected alarm sounds and the detection time of their occurrence through STFT analysis, and converting the audio data into a matrix of audio signals through calculation, wherein each element in the matrix represents the time and frequency of the audio signal;

[0026] S33, comparing the similarity between the warning sounds recorded in the audio data in the cockpit and the warning sounds in the warning sound database containing all possible sounds of the relevant helicopter; the correlation between the Pth warning in the warning sound database and all the recorded warning sounds in the audio data corresponding to the sampling time t is expressed as:

[0027]

[0028] Where n is the number of frequencies in the high power spectral density, P(p,t) j,数据库 is the PSD value of the pth alarm at frequency j in the alarm sound data at time t, and P(t) j,试验 is the PSD value of the alarm frequency j recorded in the audio data at sampling time t.

[0029] Optionally, in the flight status information determination method based on cockpit audio data analysis as described above, the engine torque evaluation method in step 3 includes:

[0030] S41, within each segment, discarding frequencies with low power spectral density, and distinguishing the remaining frequencies by using clusters, each cluster representing a main sustained frequency within the audio data;

[0031] S42, the flight data and each main continuous frequency cluster are correlated by Pearson correlation to determine the frequency corresponding to the known engine parameters; the correlation is expressed as:

[0032]

[0033] Where ρ is the Pearson correlation coefficient, x represents the engine parameter value, and y represents the frequency from the signal processing technology. and Represents the mean values ​​corresponding to x and y respectively; the ρ value is between 0 and 1, and the higher the ρ value, the higher the correlation between the engine parameter value and the given frequency cluster;

[0034] S43, based on the audio samples, the correlation between the engine parameters and the audio data is established, and the statistical model is expressed as follows using piecewise regression analysis:

[0035] y = βx + α;

[0036]

[0037] Wherein, α and β are estimated quantities, and the statistical model describes the engine parameters and noise at the corresponding main frequencies;

[0038] S44, establishing a statistical model to represent the relationship between the main frequency (x) and the engine parameter (y), and estimating the flight parameter by the following formula based on the statistical model when the audio characteristics of the cockpit are known;

[0039]

[0040] In the formula, is the evaluation parameter, and x is the dominant frequency of the recorded in-cockpit audio data.

[0041] The embodiment of the present invention also provides a flight status information judgment system based on cockpit audio data analysis, comprising: a memory and a processor, and a cockpit voice recorder and / or a protection recorder;

[0042] Wherein, the memory is configured to store executable instructions;

[0043] The processor is specifically configured to use a cockpit voice recorder and / or a protection recorder to implement the flight status information judgment method based on cockpit audio data analysis as described in any one of the above when executing the executable instructions stored in the memory.

[0044] Beneficial effects of the present invention: The embodiment of the present invention provides a flight status information judgment method and system based on cockpit audio data analysis, specifically a method for facilitating detection of warning sounds and inference of engine status information. In order to achieve two sub-goals in this flight status information judgment method, namely: detection of warning sounds and inference of engine-related parameters; based on the above two sub-goals, the cockpit audio analysis operation is divided into two sub-audio processing operation law analysis. On the one hand, for warning detection, the warning detection operation scheme is a combination of signal processing technology and control chart mutation detection technology to accurately identify or determine the type of warning, and detect their same start and end time; on the other hand, for engine torque information inference, a single processing technology is combined with a data mining method to determine the main frequency in the noise audio data, and then a statistical model is established to link engine-related parameters (related flight parameters) with audio information. Based on the detection of the warning sound type and the time of occurrence, as well as the inference of engine-related parameters, the audio data in the cockpit can supplement the data recorded by the flight data recorder, so as to better understand the status in the cockpit during flight, especially when an accident / incident occurs; thereby facilitating research and analysis to understand the cause of the event / accident, thereby improving the safety of the equipment and having certain engineering application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The accompanying drawings are used to provide a further understanding of the technical solution of the present invention and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the technical solution of the present invention and do not constitute a limitation on the technical solution of the present invention.

[0046] Figure 1 A flowchart of a method for determining flight status information based on cockpit audio data analysis provided by an embodiment of the present invention;

[0047] Figure 2 A schematic diagram of frequency clustering recognition in a flight status information determination method based on cockpit audio data analysis provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0048] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] The above background technology has explained the current situation of high accident rate of helicopters, and it is considered to supplement the work of flight data recorders by analyzing the audio data of the cockpit. However, based on the relatively unique characteristics of different types of sounds in the audio data in the cockpit, and the characteristics that various sounds are mixed with noise; the existing analysis methods for audio data in the helicopter cockpit are difficult to achieve accurate sound extraction.

[0050] In response to the above problems, an embodiment of the present invention provides a flight status information judgment method and system based on cockpit audio data analysis, which detects the sound type and precise time of each alarm triggering through multiple cockpit audio analysis operations, and uses cockpit audio data to evaluate engine-related parameters.

[0051] The present invention provides the following specific embodiments which can be combined with each other. The same or similar concepts or processes may not be described in detail in some embodiments.

[0052] Figure 1 The present invention provides a flowchart of a method for determining flight status information based on cockpit audio data analysis. Figure 1 As shown, the flight status information determination method provided by the embodiment of the present invention includes:

[0053] Step 1: Use the cockpit voice recorder and / or protection recorder to obtain the audio data in the cockpit;

[0054] Step 2, performing data preprocessing and noise suppression processing on the audio data in the cockpit to convert the non-language information into sampling points with a specified sampling rate, extracting the audio data corresponding to each sampling point, and performing noise suppression processing on the extracted audio data;

[0055] Step 3: Use the warning sound detection operation law to detect the type and precise time of each warning sound, and use the engine torque evaluation operation law to perform engine torque evaluation.

[0056] In one implementation of the embodiment of the present invention, in step 1, the following two methods may be used to obtain the audio data in the cockpit, including:

[0057] Method 1, using a cockpit voice recorder to obtain audio data in the cockpit, including: four channels in the cockpit voice recorder work simultaneously, channel 1 records non-verbal information, the non-verbal information is the warning sound in the cockpit area, channel 2 and channel 3 record the voice of the pilot and co-pilot respectively, and channel 4 records the sound information of the radio panel;

[0058] Method 2, using a protection recorder to obtain audio data in the cockpit, includes: the protection recorder has one channel working to record all sounds in the cockpit, and the recorded sounds include alarms and engine sounds in the audio data.

[0059] In one implementation of the embodiment of the present invention, the data preprocessing method in step 2 may include:

[0060] Step 21, dividing the audio data into language information and non-language information: the language information includes the voices of the positive and negative drivers, and the non-language information includes other noise information and background sound;

[0061] Step 22, information conversion preprocessing: convert the non-language information into a set of sampling points with a specified sampling rate, and extract the audio data corresponding to each sampling point.

[0062] Furthermore, the noise suppression processing method in step 2 of the implementation may include:

[0063] Step 23, using high-pass and low-pass filters to suppress noise on the audio data of each sampling point extracted to retain the audio data within a specified frequency range;

[0064] Among them, filtered data is used for warning sound detection, and unfiltered audio data is used for engine torque estimation.

[0065] Step 3 of the embodiment of the present invention includes two parts:

[0066] The first part, step 3, detects each alarm sound, including:

[0067] For warning sound detection, the category of the warning sound is identified by the highest correlation between the warning sound in the obtained cockpit audio data and the sound of the warning sound database generated using the warning characteristics;

[0068] The second part, engine torque evaluation in step 3, includes:

[0069] For engine torque evaluation, the correlation between the engine sound main frequency range and the known engine parameters in the noise and uncertainty range is obtained to determine the main frequency of the engine.

[0070] In an implementation of the embodiment of the present invention, the method of detecting each alarm sound in step 3 includes:

[0071] S31, identifying features of the alarm sound, including: using Fourier transform and high power spectral density to identify features of each alarm sound in an ideal environment, and detecting the occurrence time of each alarm sound in the ideal environment; before analyzing the alarm sound data, analyzing features of each alarm sound in the ideal environment to generate an alarm sound database;

[0072] S32, performing frequency domain analysis on the alarm sound, obtaining the types of the collected alarm sounds and the detection time of their occurrence through STFT analysis, and converting the audio data into a matrix of audio signals through calculation, wherein each element in the matrix represents the time and frequency of the audio signal;

[0073] S33, comparing the similarity between the warning sounds recorded in the audio data in the cockpit and the warning sounds in the warning sound database containing all possible sounds of the relevant helicopter; the correlation between the Pth warning in the warning sound database and all the recorded warning sounds in the audio data corresponding to the sampling time t is expressed as:

[0074]

[0075] Where n is the number of frequencies in the high power spectral density, P(p,t) j,数据库 is the PSD value of the pth alarm at frequency j in the alarm sound data at time t, and P(t) j,试验 is the PSD value of the alarm frequency j recorded in the audio data at sampling time t.

[0076] In one implementation of the embodiment of the present invention, the method of evaluating the engine torque in step 3 includes:

[0077] S41, within each segment, discarding frequencies with low power spectral density, and distinguishing the remaining frequencies by using clusters, each cluster representing a main sustained frequency within the audio data;

[0078] S42, the flight data and each main continuous frequency cluster are correlated by Pearson correlation to determine the frequency corresponding to the known engine parameters; the correlation is expressed as:

[0079]

[0080] Where ρ is the Pearson correlation coefficient, x represents the engine parameter value, and y represents the frequency from the signal processing technology. and represents the mean of x and y respectively; the ρ value is between 0 and 1, and the higher the ρ value, the higher the correlation between the engine parameter value and the given frequency cluster;

[0081] S43, based on the audio samples, the correlation between the engine parameters and the audio data is established, and the statistical model is expressed as follows using piecewise regression analysis:

[0082] y=βx+α ; (6)

[0083]

[0084] Wherein, α and β are estimated quantities, and the statistical model describes the engine parameters and noise at the corresponding main frequencies;

[0085] S44, establishing a statistical model to represent the relationship between the main frequency (x) and the engine parameter (y), and estimating the flight parameter by the following formula based on the statistical model when the audio characteristics of the cockpit are known;

[0086]

[0087] In the formula, is the evaluation parameter, and x is the dominant frequency of the recorded in-cockpit audio data.

[0088] The flight status information judgment method based on cockpit audio data analysis provided by the embodiment of the present invention is specifically a method for facilitating the detection of warning sounds and inferring engine status information. In order to achieve two sub-goals in this flight status information judgment method, namely: detection of warning sounds and inference of engine-related parameters; based on the above two sub-goals, the cockpit audio analysis operation is divided into two sub-audio processing operation law analysis. On the one hand, for warning detection, the warning detection operation scheme is a combination of signal processing technology and control chart mutation detection technology to accurately identify or determine the warning type, and detect their same start and end time; on the other hand, for engine torque information inference, a single processing technology is combined with a data mining method to determine the main frequency in the noise audio data, and then a statistical model is established to link the engine-related parameters (related flight parameters) with the audio information. Based on the detection of the warning sound type and the time of occurrence, as well as the inference of engine-related parameters, the audio data in the cockpit can supplement the data recorded by the flight data recorder, so as to better understand the status in the cockpit during flight, especially when an accident / incident occurs; thereby facilitating research and analysis to understand the cause of the event / accident, thereby improving the safety of the equipment and having certain engineering application value.

[0089] Based on the flight status information determination method based on cockpit audio data analysis provided by the above embodiment of the present invention, the embodiment of the present invention further provides a flight status information determination system based on cockpit audio data analysis, comprising: a memory and a processor, and a cockpit voice recorder and / or a protection recorder;

[0090] Wherein, the memory is configured to store executable instructions;

[0091] The processor is specifically configured to use a cockpit voice recorder and / or a protection recorder to implement a flight status information determination method based on cockpit audio data analysis as provided in any of the above embodiments when executing executable instructions stored in a memory.

[0092] The following is a schematic illustration of the implementation of the flight status information determination method based on cockpit audio data analysis provided by an embodiment of the present invention through a specific embodiment.

[0093] Example

[0094] This embodiment provides a flight status information judgment method based on cockpit audio data analysis, which detects the sound type and precise time of each alarm triggering through multiple audio analysis operations, and uses the cockpit audio data to evaluate engine related parameters. The main method flow is as follows: Figure 1 The process shown:

[0095] Step 1: Formulate audio analysis operation rules;

[0096] S1.1, determine the input of the operation law: use the cockpit audio data of the warning and engine sound as the input of the operation law;

[0097] S1.2, audio data preprocessing: converting audio data information into sampling points with a specified sampling rate, and obtaining audio data of the sampling points;

[0098] S1.3, noise suppression: Use high and low bandpass filters to suppress noise and background noise, such as wind noise and human voices. The warning sound detection algorithm uses filtered data, while the engine torque evaluation algorithm uses non-filtered audio data.

[0099] S1.4, find audio data correlation: For warning sound detection, identify the type of warning sound by obtaining the highest correlation between the cockpit audio data warning sound and the sound of the warning sound database generated using the warning characteristics; for engine torque evaluation, find the correlation between the main frequency range of the engine sound and the known engine parameters in the noise and uncertainty range to determine the main frequency of the engine;

[0100] S1.5, Statistical model: Use regression method to generate statistical model;

[0101] S1.6, detect the type and occurrence time of the alarm, and establish relevant flight parameters.

[0102] Step 2: Obtaining audio data in the cockpit;

[0103] S2.1, obtain cockpit audio data; this step can be done in two ways, namely using a cockpit voice recorder and a protection recorder; wherein:

[0104] Type 1: The cockpit voice recorder generally has four channels working simultaneously to record cockpit sounds; channel 1 records non-verbal information, i.e. the warning sounds in the cockpit area, channel 2 and channel 3 record the voices of the pilot and co-pilot respectively, and channel 4 records the sound information of the radio panel;

[0105] Type 2: The protection recorder generally has only one channel working, which records all sounds. Different from the audio data of the cockpit voice recorder, the protection recorder obtains the warning sound and engine sound in the audio data.

[0106] Step 3: preprocess and reduce noise of cockpit audio data;

[0107] 3.1, the audio data is divided into language information and non-language information: the language information includes the voices of positive and negative drivers, and the non-language information includes other noise information and background sound;

[0108] 3.2, Signal conversion preprocessing: convert non-language information into a set of sampling points with a specified sampling rate, and extract the audio data corresponding to each sampling point for further analysis;

[0109] 3.3. Noise suppression: Use high-pass and low-pass filters to suppress the noise of the audio data extracted from each sampling point. The bandpass filter can effectively suppress unwanted frequencies and only retain the audio data within the relevant specified frequency range; for the evaluation of engine torque, use unfiltered sound.

[0110] Step 4: Analyze cockpit audio data;

[0111] The cockpit audio analysis algorithm consists of two sub-algorithms: one is the warning sound detection, and the other is the engine torque evaluation.

[0112] The first type, alarm sound detection, includes the following steps:

[0113] a) First, create an alarm sound database.

[0114] Use Fourier transform and high power spectral density to identify the characteristics of each alarm sound in an ideal environment, and then detect the occurrence time of each alarm sound in the ideal environment. The alarm sound in an ideal environment refers to the alarm sound in an environment without background sound and noise. Before analyzing the alarm sound data, analyze the characteristics of each alarm sound in an ideal environment in order to generate an alarm sound database. The database is generated in the following way: the main frequency of the alarm sound can be used to identify the characteristics of a specific sound, and the main frequency signal can be represented as a distribution in a frequency range. Based on the characteristic that the alarm sound is a periodic signal, a discrete Fourier transform is used to discretize it based on the specified sampling rate in step 2, and the alarm sound data of each discrete point constitutes the alarm sound database.

[0115] This is given by the following formula.

[0116]

[0117] Where N is the number of samples, n is the sampling point, X(k) is the amplitude of the alarm sound signal at frequency k, x[n] represents the frequency value corresponding to sampling point n; P(k) is the high power spectral density (PSD) at frequency k; PSD is expressed as:

[0118] P(k)=|x(k)| 2 ; (2)

[0119] b) Then, perform frequency domain analysis on the alarm sound.

[0120] The categories of the alarm sounds corresponding to the adopted points and their occurrence times are obtained through STFT analysis. The audio data is converted into a matrix of audio signals through calculation, and each element in the matrix represents the time and frequency of the audio signal.

[0121]

[0122] Where i is the time index, k is the frequency, and w(ni) is the window function used to modify the value of x(n).

[0123] c) Finally, the statistical value is calculated. The correlation using the Euclidean distance can compare the similarity between the warning sounds recorded in the cockpit audio data and the warning sound database containing all possible sounds of the relevant helicopter. The correlation between the Pth warning in the warning sound database and all the recorded warning sounds in the audio data at sampling time t can be provided by the following formula.

[0124]

[0125] Where n is the number of frequencies in the high power spectral density, P(p,t) j,数据库is the PSD value of the pth alarm at frequency j in the alarm sound data at time t, and P(t) j,试验 is the PSD value of the alarm frequency j recorded in the audio data at sampling time t. It should be noted that the alarm analysis operation law proposed here calculates the mean of all values ​​of the correlation between the alarm in the database and the alarm sound recorded in the audio data, and uses the mean to calculate the statistical value.

[0126] The second type, engine torque evaluation, includes the following steps:

[0127] a) First, the audio data is processed using the signal processing technique applied to 60s segments (i.e., Equation 3 above). Within each segment, frequencies with low power spectral density are discarded and the remaining frequencies are distinguished by using clustering. This step eliminates indirect source noise. This density-based clustering method has the advantage of identifying irregularly shaped clusters and the flexibility of controlling data of different sizes. These clusters can be represented as the main continuous frequencies in the audio data; Figure 2 FIG. 1 is a schematic diagram of frequency clustering recognition in a flight status information determination method based on cockpit audio data analysis provided by an embodiment of the present invention. Figure 2 The blue, cyan, and purple colors in the figure represent clusters, respectively.

[0128] b) The flight data is then correlated with each cluster of dominant frequencies to determine the frequencies (if any) corresponding to known engine parameters. This is done using the Pearson correlation given by the following equation:

[0129]

[0130] Where ρ is the Pearson correlation coefficient, x represents the engine parameter value, and y represents the frequency from the signal processing technology. and Represents the mean of x and y respectively; the ρ value is between 0 and 1, and the higher the ρ value, the higher the correlation between the engine parameter value and the given frequency cluster.

[0131] c) Then, the correlation (relationship) between engine parameters and audio data is established based on a large number of audio samples, and the statistical model is expressed using segmented regression analysis. The engine parameter is treated as an independent variable (i.e., y), and the corresponding frequency is treated as a dependent variable (i.e., x). In this way, the linear regression is defined as:

[0132] y=βx+α ; (6)

[0133]

[0134] Where α and β are estimated quantities. This statistical model describes the engine parameters and noise at the corresponding main frequency;

[0135] d) Finally, a statistical model is established to provide the relationship between the main frequency (i.e. x) and the engine parameters (i.e. y). Based on this, given the audio characteristics of the cockpit, the flight parameters can be estimated by the following formula.

[0136]

[0137] In the formula, is the evaluation parameter and x is the relevant dominant frequency in the recorded cockpit sound.

[0138] Although the embodiments disclosed in the present invention are as above, the contents are only embodiments adopted to facilitate understanding of the present invention and are not intended to limit the present invention. Any technician in the field to which the present invention belongs can make any modifications and changes in the form and details of implementation without departing from the spirit and scope disclosed in the present invention, but the patent protection scope of the present invention shall still be subject to the scope defined in the attached claims.

Claims

1. A flight status information judgment method based on cockpit audio data analysis, characterized in that: include: Step 1: Use the cockpit voice recorder and / or protection recorder to obtain the audio data in the cockpit; Step 2, performing data preprocessing and noise suppression processing on the audio data in the cockpit to convert the non-language information into sampling points with a specified sampling rate, extracting the audio data corresponding to each sampling point, and performing noise suppression processing on the extracted audio data; Step 3: Use the warning sound detection operation law to detect the type and precise time of each warning sound, and use the engine torque evaluation operation law to perform engine torque evaluation.

2. The flight status information determination method based on cockpit audio data analysis according to claim 1, characterized in that: The method of obtaining the audio data in the cockpit in step 1 includes: Method 1, using a cockpit voice recorder to obtain audio data in the cockpit, including: four channels in the cockpit voice recorder work simultaneously, channel 1 records non-verbal information, that is, the warning sound in the cockpit area, channel 2 and channel 3 record the voice of the pilot and co-pilot respectively, and channel 4 records the sound information of the radio panel; Method 2, using a protection recorder to obtain audio data in the cockpit, includes: the protection recorder has one channel working to record all sounds in the cockpit, and the recorded sounds include alarms and engine sounds in the audio data.

3. The flight status information determination method based on cockpit audio data analysis according to claim 2, characterized in that: The data preprocessing method in step 2 includes: Step 21, dividing the audio data into language information and non-language information: the language information includes the voices of the positive and negative drivers, and the non-language information includes other noise information and background sound; Step 22, information conversion preprocessing: convert the non-language information into a set of sampling points with a specified sampling rate, and extract the audio data corresponding to each sampling point.

4. The flight status information determination method based on cockpit audio data analysis according to claim 3 is characterized in that: The noise suppression processing method in step 2 includes: Step 23, using high-pass and low-pass filters to suppress noise on the audio data of each sampling point extracted to retain the audio data within a specified frequency range; Among them, filtered data is used for warning sound detection, and unfiltered audio data is used for engine torque estimation.

5. The flight status information determination method based on cockpit audio data analysis according to claim 3, characterized in that: Detecting each alarm sound in step 3 includes: For warning sound detection, the category of the warning sound is identified by the highest correlation between the warning sound in the acquired cockpit audio data and the sound of the warning sound database generated using the warning characteristics.

6. The flight status information determination method based on cockpit audio data analysis according to claim 3, characterized in that: The engine torque evaluation in step 3 includes: For engine torque evaluation, the correlation between the engine sound main frequency range and the known engine parameters in the noise and uncertainty range is obtained to determine the main frequency of the engine.

7. The flight status information determination method based on cockpit audio data analysis according to claim 5, characterized in that: The method of detecting each alarm sound in step 3 includes: S31, identifying features of the alarm sound, including: using Fourier transform and high power spectral density to identify features of each alarm sound in an ideal environment, and detecting the occurrence time of each alarm sound in the ideal environment; before analyzing the alarm sound data, analyzing features of each alarm sound in the ideal environment to generate an alarm sound database; S32, performing frequency domain analysis on the alarm sound, obtaining the types of the collected alarm sounds and the detection time of their occurrence through STFT analysis, and converting the audio data into a matrix of audio signals through calculation, wherein each element in the matrix represents the time and frequency of the audio signal; S33, comparing the similarity between the warning sounds recorded in the audio data in the cockpit and the warning sounds in the warning sound database containing all possible sounds of the relevant helicopter; the correlation between the Pth warning in the warning sound database and all the recorded warning sounds in the audio data corresponding to the sampling time t is expressed as: Where n is the number of frequencies in the high power spectral density, P(p,t) j,数据库 is the PSD value of the pth alarm at frequency j in the alarm sound data at time t, and P(t) j,试验 is the PSD value of the alarm frequency j recorded in the audio data at sampling time t.

8. The flight status information determination method based on cockpit audio data analysis according to claim 6, characterized in that: The method of evaluating the engine torque in step 3 includes: S41, within each segment, discarding frequencies with low power spectral density, and distinguishing the remaining frequencies by using clusters, each cluster representing a main sustained frequency within the audio data; S42, the flight data and each main continuous frequency cluster are correlated by Pearson correlation to determine the frequency corresponding to the known engine parameters; the correlation is expressed as: Where ρ is the Pearson correlation coefficient, x represents the engine parameter value, and y represents the frequency from the signal processing technology. and Represents the mean values ​​corresponding to x and y respectively; the ρ value is between 0 and 1, and the higher the ρ value, the higher the correlation between the engine parameter value and the given frequency cluster; S43, based on the audio samples, the correlation between the engine parameters and the audio data is established, and the statistical model is expressed as follows using piecewise regression analysis: y = βx + α; Wherein, α and β are estimated quantities, and the statistical model describes the engine parameters and noise at the corresponding main frequencies; S44, establishing a statistical model to represent the relationship between the main frequency (x) and the engine parameter (y), and estimating the flight parameter by the following formula based on the statistical model when the audio characteristics of the cockpit are known; In the formula, is the evaluation parameter, and x is the dominant frequency of the recorded in-cockpit audio data.

9. A flight status information judgment system based on cockpit audio data analysis, characterized in that: include: Memory and processor, and cockpit voice recorder and / or protection recorder; Wherein, the memory is configured to store executable instructions; The processor is specifically configured to use a cockpit voice recorder and / or a protection recorder to implement the flight status information determination method based on cockpit audio data analysis as described in any one of claims 1 to 8 when executing the executable instructions stored in the memory.

Images