Rail transit broadcasting method based on sound field analysis

Abstract

The invention discloses a rail transit broadcasting method based on sound field analysis, and the method comprises the steps: collecting real-time sound signal data in a high-speed railway station, and obtaining an energy characteristic parameter matrix of a real-time sound signal; calling sample sound signal data of the sound signal sample library, taking the energy characteristic parameter matrix of the sound sample signal data as input, and training a support vector data description algorithm model to obtain a sound signal recognition model; inputting the energy characteristic parameter matrix of the real-time sound signal data into a sound signal identification model to obtain a real-time sound signal data category, and sending out an early warning signal according to the real-time sound signal data category; according to the method, the energy coefficient distribution matrix of the energy distribution matrix library is obtained, the energy distribution of the broadcast voice signals is adjusted according to the energy coefficient distribution matrix, and the definition of the broadcast voice is improved by adjusting the energy distribution of the broadcast voice signals at different harmonic frequencies on the premise of not increasing the total energy of the voice, so that passengers can hear broadcast contents more clearly.

Description

Technical Field

[0001] The present invention relates to the technical field of rail transit broadcasting, and in particular to a rail transit broadcasting method based on sound field analysis. Background Art

[0002] In recent years, with the large-scale development of urban rail transit, high standards and high requirements for construction, operation and service have received more and more attention. Although all high-speed railway stations have provided visual information prompts for passengers, including LED display screens, LCD display screens and other equipment in high-speed railway stations, during peak hours at stations, when passengers are in a hurry to catch the train, and when important voice information is broadcast in high-speed railway stations, a large number of passengers cannot confirm the location of the high-speed railway train departure platform and departure delays through the LED / LCD display screens. In this case, the best source of information is the station broadcast. When the noise is too loud and the broadcast voice cannot be heard clearly, it will cause passengers to be delayed in catching the high-speed train and unable to effectively receive important voice information broadcast in the high-speed railway station.

[0003] At present, the noise inside high-speed railway stations is still relatively high. Even though the current high-speed railway station projects have basically all enabled noise monitoring and automatic adjustment of broadcast volume, the noise in some sections is still too loud inside the high-speed railway stations. Even if the broadcast system is set to the maximum volume, passengers cannot fully receive the broadcast information, which will affect the effect of the broadcast to varying degrees. On the basis of studying the existing noise reduction measures, we will take a different approach to improve the quality of broadcast audio and enhance the travel experience of passengers. Summary of the invention

[0004] The technical problem solved by the present invention is: when the noise is too loud to hear the broadcast voice clearly, even if noise monitoring is enabled and the broadcast volume is automatically adjusted, the noise in some sections of the high-speed railway station is still too loud. Even if the broadcast system is set to the maximum volume, passengers cannot fully receive the broadcast information, affecting their normal travel experience.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions: a rail transit broadcasting method based on sound field analysis, comprising:

[0006] Step S100, real-time sound signal data in the high-speed railway station is collected, the real-time sound signal data is input into a spectrum display, a spectrum diagram of the real-time sound signal data is obtained, and an energy characteristic parameter matrix of the real-time sound signal is obtained;

[0007] Step S200, retrieving sample sound signal data from a sound signal sample library, taking the energy characteristic parameter matrix of the sound sample signal data as input, training a support vector data description algorithm model, and obtaining a sound signal recognition model based on the support vector data description algorithm;

[0008] Step S300, inputting the energy characteristic parameter matrix of the real-time sound signal data into the sound signal recognition model to obtain the real-time sound signal data category, and issuing a warning signal according to the real-time sound signal data category;

[0009] Step S400, obtaining an energy coefficient distribution matrix corresponding to the real-time sound signal data category from an energy distribution matrix library, and adjusting the distribution of sound energy of the broadcast voice signal according to the energy coefficient distribution matrix.

[0010] As a preferred solution of the rail transit broadcasting method based on sound field analysis described in the present invention, step S100 specifically includes:

[0011] Step S101, obtaining the real-time sound signal data in the high-speed railway station through a sound sensor;

[0012] Step S102, inputting the real-time sound signal data into the spectrum display, and the spectrum display performs short-time Fourier transform on the real-time sound signal data to obtain a spectrum diagram of the real-time sound signal data;

[0013] Step S103, extracting the frequency band energy characteristics of the spectrum using the Mel-frequency cepstrum coefficients to obtain an energy characteristic parameter matrix of the frequency band energy of the real-time sound signal data.

[0014] As a preferred solution of the rail transit broadcasting method based on sound field analysis described in the present invention, step S200 specifically includes:

[0015] Step S201, retrieve all types of sample sound signal data contained in a sound signal sample library, wherein the sample sound signal data category library contains three categories of sound signals, namely, first category sound signal Y, second category sound signal Q and third category sound signal Z;

[0016] The first type of sound signal Y includes air sound signal Y1, wheel-rail sound signal Y2, high-speed railway station equipment sound signal Y3 and passenger conversation sound signal Y4;

[0017] The second type of sound signal Q includes a passenger distress sound signal Q1;

[0018] The third type of sound signal Z includes a landslide sound signal Z1, a tornado sound signal Z2 and an earthquake sound signal Z3;

[0019] Step S202, retrieve the energy characteristic parameter matrices of the first type of sound signal Y, the second type of sound signal Q and the third type of sound signal Z, number the energy characteristic parameter matrices, and obtain the air sound signal Y11, the wheel-rail sound signal Y22, the sound signal of equipment in the high-speed railway station Y33, the passenger conversation sound signal Y44, the passenger calling for help sound signal Q11, the landslide sound signal Z11, the tornado sound signal Z22 and the earthquake sound signal Z33.

[0020] Step S203, input the air sound signal Y11, wheel-rail sound signal Y22, high-speed railway station equipment sound signal Y33, passenger conversation sound signal Y44, passenger call for help sound signal Q11, landslide sound signal Z11, tornado sound signal Z22 and earthquake sound signal Z33 into the support vector data description algorithm model, perform sound signal data recognition, obtain the sample sound signal data category, and obtain a sound signal recognition model based on the support vector data description algorithm.

[0021] As a preferred solution of the rail transit broadcasting method based on sound field analysis described in the present invention, step S300 specifically includes:

[0022] Step S301, inputting the collected real-time sound signal data into the trained sound signal recognition model to obtain the real-time sound signal data category of the sound signal recognition model;

[0023] Step S302, issuing a warning signal according to the real-time sound signal data category, the real-time sound signal data category includes the first category sound signal Y, the second category sound signal Q and the third category sound signal Z, and the warning signal includes a distress warning signal and a natural disaster warning signal; when the real-time sound signal data category belongs to the second category sound signal Q, issuing a distress warning signal;

[0024] When the real-time sound signal data category belongs to the third type of sound signal Z, a natural disaster warning signal is issued.

[0025] As a preferred solution of the rail transit broadcasting method based on sound field analysis described in the present invention, step S400 specifically includes:

[0026] Step S401, the energy distribution matrix library stores the energy coefficient distribution matrix corresponding to each real-time sound signal data category, and searches for a matching energy coefficient distribution matrix according to the real-time sound signal data category output by the sound signal recognition model;

[0027] Step S402, inputting the broadcast voice signal into a spectrum display to obtain a spectrum diagram of the broadcast voice signal;

[0028] Step S403, extracting the frequency band energy characteristics of the spectrogram using the Mel-frequency cepstral coefficients to obtain an energy characteristic parameter matrix of the frequency band energy of the broadcast voice signal;

[0029] Step S404: multiply the energy characteristic parameter matrix of the broadcast voice signal frequency band energy by the energy coefficient distribution matrix to obtain an adjusted energy characteristic parameter matrix.

[0030] As a preferred solution of the rail transit broadcasting method based on sound field analysis described in the present invention, the process of establishing the energy characteristic parameter matrix of the broadcast voice signal frequency band energy is as follows:

[0031] The spectrum of the broadcast voice signal is divided on the frequency axis into n frequency bands, which are denoted as Li to Ln;

[0032] The energy values ​​of all time points in each frequency band are accumulated to obtain the matrix [Li, Li +1 , Li +2 , ...Ln], Li represents the total energy value of the Li band, the matrix [Li, Li +1 , Li +2 , ...Ln] is the energy characteristic parameter matrix of the broadcast voice signal.

[0033] As a preferred solution of the rail transit broadcasting method based on sound field analysis described in the present invention, the process of establishing the energy characteristic parameter matrix of the real-time sound signal is as follows:

[0034] The spectrum of the real-time sound signal is divided on the frequency axis into n frequency bands, which are denoted as Li1 to Ln1;

[0035] The energy values ​​of all time points in each frequency band are accumulated to obtain the matrix [Li1, Li 1+1 , Li 1+2 , ...Ln1], Li1 represents the total energy value of Li1 band, the matrix [Li1, Li 1+1 , Li 1+2 , ...Ln1] is the energy characteristic parameter matrix of the real-time sound signal.

[0036] As a preferred solution of the rail transit broadcasting method based on sound field analysis described in the present invention, the process of establishing the energy coefficient distribution matrix corresponding to the real-time sound signal data category is as follows:

[0037] Obtain the energy characteristic parameter matrix [Li1, Li 1+1 Li 1+2, ...Ln1], Li1 represents the total energy of Li1 Li1+Li 1+1 +Li 1+2 +, ... Ln1 / Li1 band energy;

[0038] Normalize the energy values ​​of the Li1-Ln1 band to a range of 0 to 1, and obtain the matrix [Lg, Lg +1 , Lg +2 , ...Lg n ], the matrix [Lg, Lg +1 , Lg +2 , ...Lg n ] is the energy coefficient distribution matrix of the real-time sound signal.

[0039] As a preferred solution of the rail transit broadcasting method based on sound field analysis described in the present invention, the energy characteristic parameter matrix after adjustment is as follows:

[0040] P=[Lg,Lg +1 , Lg +2 , ...Lg n ]*[Li,Li +1 , Li +2 , ...Ln], the matrix P is the energy characteristic parameter matrix of the broadcast voice signal after adjustment.

[0041] As a preferred solution of the rail transit broadcasting method based on sound field analysis described in the present invention, the specific process of normalizing the energy value to a range of 0 to 1 is as follows:

[0042] S1=Li1+Li 1+1 +Li 1+2 +, ...Ln1 / Li1, S2=Li1+Li 1+1 +Li 1+2 +,...Ln1 / Li 1+1 ;

[0043] D = S1 + S1 + ... Sn;

[0044] Lg=S1 / D,Lg +1 =S2 / D.

[0045] The beneficial effects of the present invention are as follows: by inputting the broadcast voice signal into the spectrum display to obtain a spectrum diagram, using the Mel-frequency cepstrum coefficient to extract the frequency band energy characteristics of the spectrum diagram, and obtaining the energy characteristic parameter matrix of the frequency band energy of the broadcast voice signal, the energy distribution characteristics of the broadcast voice signal in different frequency bands can be predicted, the energy characteristic parameter matrix of the frequency band energy of the real-time sound signal data is negatively normalized to obtain the corresponding energy coefficient distribution matrix, and the energy coefficient distribution matrix is ​​used to allocate the distribution of the frequency band energy of the broadcast voice signal to obtain the frequency distribution segment where the noise energy is concentrated. At this time, the energy coefficient of the frequency distribution segment is the smallest, and the energy of the frequency distribution segment of the broadcast voice signal at the noise energy is concentrated. The energy can be suppressed, thereby improving the clarity of the broadcast voice signal. In a high noise environment, by adjusting the energy distribution at different harmonic frequencies of the broadcast voice signal, the clarity of the broadcast voice is improved without increasing the overall energy of the voice, so that passengers can hear the broadcast content more clearly.

[0046] By collecting real-time sound signal data in the high-speed railway station in real time, the real-time sound signal data is input into the spectrum display to obtain a spectrum diagram, and the frequency band energy characteristics of the spectrum diagram are extracted using the Mel-frequency cepstrum coefficient to obtain the energy characteristic parameter matrix of the frequency band energy of the real-time sound signal data. The energy characteristic parameter matrix is ​​input into the trained sound signal recognition model to obtain the real-time sound signal data category. Since the frequency band energy distributions of different sound signals have their own unique distribution characteristics, the main frequency components and energy proportions in different sound signals vary greatly, and the frequency band energy characteristics are easy to extract, the distribution of frequency band energy is used as the characteristic parameters of different sound signals, which can improve the accuracy of sound signal classification, simplify classification operations, improve calculation efficiency, and can timely monitor the operation of trains to ensure the safety of train operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 A schematic diagram of the basic flow of a rail transit broadcasting method based on sound field analysis is provided for one embodiment of the present invention. DETAILED DESCRIPTION

[0048] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, but not all of the embodiments.

[0049] Example, see Figure 1 , as an embodiment of the present invention, provides a rail transit broadcasting method based on sound field analysis, comprising:

[0050] Step S100, real-time sound signal data in the high-speed railway station is collected, the real-time sound signal data is input into a spectrum display, a spectrum diagram of the real-time sound signal data is obtained, and an energy characteristic parameter matrix of the real-time sound signal is obtained;

[0051] Step S200, retrieving sample sound signal data from a sound signal sample library, taking the energy characteristic parameter matrix of the sound sample signal data as input, training a support vector data description algorithm model, and obtaining a sound signal recognition model based on the support vector data description algorithm;

[0052] Step S300, inputting the energy characteristic parameter matrix of the real-time sound signal data into the sound signal recognition model to obtain the real-time sound signal data category, and issuing a warning signal according to the real-time sound signal data category;

[0053] Step S400, obtaining an energy coefficient distribution matrix corresponding to the real-time sound signal data category from an energy distribution matrix library, and adjusting the distribution of sound energy of the broadcast voice signal according to the energy coefficient distribution matrix.

[0054] Step S100 specifically includes:

[0055] Step S101, obtaining the real-time sound signal data in the high-speed railway station through a sound sensor;

[0056] Step S102, inputting the real-time sound signal data into the spectrum display, and the spectrum display performs short-time Fourier transform on the real-time sound signal data to obtain a spectrum diagram of the real-time sound signal data;

[0057] Step S103, extracting the frequency band energy characteristics of the spectrum using the Mel-frequency cepstrum coefficients to obtain an energy characteristic parameter matrix of the frequency band energy of the real-time sound signal data.

[0058] In this embodiment, a sound sensor is used to obtain real-time sound signal data in a high-speed railway station, and a spectrum display performs a short-time Fourier transform on the real-time sound signal data, so as to convert the time domain signal of the real-time sound signal data into a frequency domain signal, thereby obtaining a spectrum graph of the real-time sound signal data, selecting the frequency band energy of the spectrum graph as a feature, and using Mel-frequency cepstrum coefficients to extract the frequency band energy characteristics of the spectrum graph, thereby obtaining an energy characteristic parameter matrix based on the frequency band energy characteristics.

[0059] Step S200 specifically includes:

[0060] Step S201, retrieve all types of sample sound signal data contained in a sound signal sample library, wherein the sample sound signal data category library contains three categories of sound signals, namely, first category sound signal Y, second category sound signal Q and third category sound signal Z;

[0061] The first type of sound signal Y includes air sound signal Y1, wheel-rail sound signal Y2, high-speed railway station equipment sound signal Y3 and passenger conversation sound signal Y4;

[0062] The second type of sound signal Q includes a passenger distress sound signal Q1;

[0063] The third type of sound signal Z includes a landslide sound signal Z1, a tornado sound signal Z2 and an earthquake sound signal Z3;

[0064] Step S202, retrieve the energy characteristic parameter matrices of the first type of sound signal Y, the second type of sound signal Q and the third type of sound signal Z, number the energy characteristic parameter matrices, and obtain the air sound signal Y11, the wheel-rail sound signal Y22, the sound signal of equipment in the high-speed railway station Y33, the passenger conversation sound signal Y44, the passenger calling for help sound signal Q11, the landslide sound signal Z11, the tornado sound signal Z22 and the earthquake sound signal Z33.

[0065] Step S203, input the air sound signal Y11, wheel-rail sound signal Y22, high-speed railway station equipment sound signal Y33, passenger conversation sound signal Y44, passenger call for help sound signal Q11, landslide sound signal Z11, tornado sound signal Z22 and earthquake sound signal Z33 into the support vector data description algorithm model, perform sound signal data recognition, obtain the sample sound signal data category, and obtain a sound signal recognition model based on the support vector data description algorithm.

[0066] In this embodiment, the sound signal data of the sound signal sample library include air sound signals Y11, wheel-rail sound signals Y22, sound signals of equipment in high-speed railway stations Y33, passenger conversation sound signals Y44, passenger call for help sound signals Q11, landslide sound signals Z11, tornado sound signals Z22 and earthquake sound signals Z33. The sound signal data of the sound signal sample library are pre-input into the spectrum display, and the frequency band energy characteristics of the spectrum graph are extracted using Mel-frequency cepstrum coefficients to obtain an energy characteristic parameter matrix based on the frequency band energy characteristics. The sound signal data are divided into a test set and a training set in a ratio of 3:7, and the energy characteristic parameter matrices of different sound categories are labeled accordingly. The training set is input into a support vector machine to find the optimal hyperplane, maximize the intervals between sound signal samples of different categories, and obtain a sound signal recognition model based on the support vector data description algorithm. The test set can test the accuracy of the sound signal recognition model.

[0067] Step S300 specifically includes:

[0068] Step S301, inputting the collected real-time sound signal data into the trained sound signal recognition model to obtain the real-time sound signal data category of the sound signal recognition model;

[0069] Step S302, issuing a warning signal according to the real-time sound signal data category, the real-time sound signal data category includes the first category sound signal Y, the second category sound signal Q and the third category sound signal Z, and the warning signal includes a distress warning signal and a natural disaster warning signal; when the real-time sound signal data category belongs to the second category sound signal Q, issuing a distress warning signal;

[0070] When the real-time sound signal data category belongs to the first category of sound signal Y, no warning signal is issued;

[0071] When the real-time sound signal data category belongs to the second type of sound signal Q, a distress warning signal is issued;

[0072] When the real-time sound signal data category belongs to the third type of sound signal Z, a natural disaster warning signal is issued.

[0073] In this embodiment, the energy characteristic parameter matrix of the real-time sound signal data frequency band energy is input into the trained sound signal recognition model to obtain the real-time sound signal data category of the sound signal recognition model, and the sound in the train operation is divided into different categories. When the real-time sound signal data belongs to different real-time sound signal data categories, different early warning signals are issued, and emergency situations are monitored according to the real-time sound signal data category. Once the analogy belongs to the second category sound signal Q and the third category sound signal Z, a corresponding early warning signal is issued, which can timely reflect the train operation situation and ensure the safety of the train operation.

[0074] Step S400 specifically includes:

[0075] Step S401, the energy distribution matrix library stores the energy coefficient distribution matrix corresponding to each real-time sound signal data category, and searches for a matching energy coefficient distribution matrix according to the real-time sound signal data category output by the sound signal recognition model;

[0076] Step S402, inputting the broadcast voice signal into a spectrum display to obtain a spectrum diagram of the broadcast voice signal;

[0077] Step S403, extracting the frequency band energy characteristics of the spectrogram using the Mel-frequency cepstral coefficients to obtain an energy characteristic parameter matrix of the frequency band energy of the broadcast voice signal;

[0078] Step S404: multiply the energy characteristic parameter matrix of the broadcast voice signal frequency band energy by the energy coefficient distribution matrix to obtain an adjusted energy characteristic parameter matrix.

[0079] In this embodiment, the energy characteristic parameter matrix of the broadcast voice signal frequency band energy can be used to know in advance the energy distribution of the broadcast voice in different frequency bands. The current real-time sound signal data category can be identified through the sound signal recognition model, and the energy coefficient distribution matrix corresponding to the real-time sound signal data category can be retrieved. Different energy parameter matrices can be obtained according to the changes in the real-time sound signal data category during the operation process, and the energy distribution of the broadcast voice signal frequency band energy in different frequency bands can be adjusted in real time, thereby ensuring the continuity of the clarity of the broadcast voice signal.

[0080] The specific process of establishing the energy characteristic parameter matrix of the broadcast voice signal frequency band energy is as follows:

[0081] The spectrum of the broadcast voice signal is divided on the frequency axis into n frequency bands, which are denoted as Li to Ln;

[0082] The energy values ​​of all time points in each frequency band are accumulated to obtain the matrix [Li, Li +1 , Li +2 , ...Ln], Li represents the total energy value of the Li band, the matrix [Li, Li +1 , Li +2 , ...Ln] is the energy characteristic parameter matrix of the broadcast voice signal.

[0083] The process of establishing the energy characteristic parameter matrix of the real-time sound signal is as follows:

[0084] The spectrum of the real-time sound signal is divided on the frequency axis into n frequency bands, which are denoted as Li1 to Ln1;

[0085] The energy values ​​of all time points in each frequency band are accumulated to obtain the matrix [Li1, Li 1+1 , Li 1+2 , ...Ln1], Li1 represents the total energy value of Li1 band, the matrix [Li1, Li 1+1 , Li 1+2 , ...Ln1] is the energy characteristic parameter matrix of the real-time sound signal.

[0086] The process of establishing the energy coefficient distribution matrix corresponding to the real-time sound signal data category is as follows:

[0087] Obtain the energy characteristic parameter matrix [Li1, Li 1+1 Li1+2 , ...Ln1], Li1 represents the total energy of Li1 Li1+Li 1+1 +Li 1+2 +, ... Ln1 / Li1 band energy;

[0088] Normalize the energy values ​​of the Li1-Ln1 band to a range of 0 to 1, and obtain the matrix [Lg, Lg +1 , Lg +2 , ...Lg n ], the matrix [Lg, Lg +1 , Lg +2 , ...Lg n ] is the energy coefficient distribution matrix of the real-time sound signal.

[0089] The adjusted energy characteristic parameter matrix is ​​as follows:

[0090] P=[Lg,Lg +1 , Lg +2 , ...Lg n ]*[Li,Li +1 , Li +2 , ...Ln], the matrix P is the energy characteristic parameter matrix of the broadcast voice signal after adjustment.

[0091] The specific process of normalizing the energy value to a range of 0 to 1 is as follows:

[0092] S1=Li1+Li 1+1 +Li 1+2 +, ...Ln1 / Li1, S2=Li1+Li 1+1 +Li 1+2 +,...Ln1 / Li 1+1 ;

[0093] D = S1 + S1 + ... Sn;

[0094] Lg=S1 / D,Lg +1 =S2 / D.

[0095] In this embodiment, S1=Li1+Li 1+1 +Li 1+2 +, ...Ln1 / Li1; Li1+Li 1+1 +Li 1+2+, ...Ln1 represents the total energy in each frequency segment of the real-time sound signal data, Li1 represents the total energy of the first frequency segment, the total energy in each frequency segment is divided by the total energy of the first frequency segment to obtain S1, when the total energy of the first frequency segment is larger, the smaller S1 is, the Sn of each frequency segment is obtained, the Sn corresponding to each frequency segment is normalized, Lg = S1 / D, the energy coefficient distribution matrix of the real-time sound signal is obtained, the energy coefficient distribution matrix is ​​multiplied by the energy characteristic parameter matrix of the broadcast voice signal, and P = [Lg, Lg +1 , Lg +2 , ...Lg n ]*[Li,Li +1 , Li +2 , ...Ln], the P matrix reflects the frequency band energy distribution of the adjusted broadcast voice signal, and the energy characteristic parameter matrix P of the adjusted broadcast voice signal is [Lg, Lg +1 , Lg +2 , ...Lg n ] Continuously adjusting according to the changes in the real-time sound signal data category can ensure the reasonable distribution of the frequency band energy of the final broadcast voice signal, improve the clarity of the broadcast audio, and enhance the passenger experience.

[0096] By inputting the broadcast voice signal into the spectrum display to obtain the spectrum graph, the Mel-frequency cepstrum coefficient is used to extract the frequency band energy characteristics of the spectrum graph, and the energy characteristic parameter matrix of the frequency band energy of the broadcast voice signal is obtained. The distribution characteristics of the energy of the broadcast voice signal in different frequency bands can be predicted, and the energy characteristic parameter matrix of the frequency band energy of the real-time sound signal data is negatively normalized to obtain the corresponding energy coefficient distribution matrix. The energy coefficient distribution matrix is ​​used to allocate the distribution of the frequency band energy of the broadcast voice signal to obtain the frequency distribution segment where the noise energy is concentrated. At this time, the energy coefficient of the frequency distribution segment is the smallest, which can suppress the energy of the frequency distribution segment of the broadcast voice signal where the noise energy is concentrated, thereby improving the clarity of the broadcast voice signal. In a high noise environment, by adjusting the energy distribution of the broadcast voice signal at different harmonic frequencies, the clarity of the broadcast voice is improved without increasing the overall energy of the voice, so that passengers can hear the broadcast content more clearly.

[0097] By collecting real-time sound signal data in the high-speed railway station in real time, the real-time sound signal data is input into the spectrum display to obtain a spectrum diagram, and the frequency band energy characteristics of the spectrum diagram are extracted using the Mel-frequency cepstrum coefficient to obtain the energy characteristic parameter matrix of the frequency band energy of the real-time sound signal data. The energy characteristic parameter matrix is ​​input into the trained sound signal recognition model to obtain the real-time sound signal data category. Since the frequency band energy distributions of different sound signals have their own unique distribution characteristics, the main frequency components and energy proportions in different sound signals vary greatly, and the frequency band energy characteristics are easy to extract, the distribution of frequency band energy is used as the characteristic parameters of different sound signals, which can improve the accuracy of sound signal classification, simplify classification operations, improve calculation efficiency, and can timely monitor the operation of trains to ensure the safety of train operation.

[0098] It should be understood by those skilled in the art that the embodiments of the present invention can be provided as methods, systems or computer program products. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program codes. Among them, the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic memory, flash memory, disk or optical disk. These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0099] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A rail transit broadcasting method based on sound field analysis, characterized in that: include: Step S100, real-time sound signal data of the high-speed railway station is collected, the real-time sound signal data is input into a spectrum display, a spectrum diagram of the real-time sound signal data is obtained, and an energy characteristic parameter matrix of the real-time sound signal is obtained; Step S200, retrieving sample sound signal data from a sound signal sample library, taking the energy characteristic parameter matrix of the sound sample signal data as input, training a support vector data description algorithm model, and obtaining a sound signal recognition model based on the support vector data description algorithm; Step S300, inputting the energy characteristic parameter matrix of the real-time sound signal data into the sound signal recognition model to obtain the real-time sound signal data category, and issuing a warning signal according to the real-time sound signal data category; Step S400, obtaining an energy coefficient distribution matrix corresponding to the real-time sound signal data category from an energy distribution matrix library, and adjusting the distribution of sound energy of the broadcast voice signal according to the energy coefficient distribution matrix.

2. A rail transit broadcasting method based on sound field analysis as claimed in claim 1, characterized in that: Step S100 specifically includes: Step S101, obtaining the real-time sound signal data in the high-speed railway station through a sound sensor; Step S102, inputting the real-time sound signal data into the spectrum display, and the spectrum display performs short-time Fourier transform on the real-time sound signal data to obtain a spectrum diagram of the real-time sound signal data; Step S103, extracting the frequency band energy characteristics of the spectrum using the Mel-frequency cepstrum coefficients to obtain an energy characteristic parameter matrix of the frequency band energy of the real-time sound signal data.

3. A rail transit broadcasting method based on sound field analysis as claimed in claim 2, characterized in that: Step S200 specifically includes: Step S201, retrieve all types of sample sound signal data contained in a sound signal sample library, wherein the sample sound signal data category library contains three categories of sound signals, namely, first category sound signal Y, second category sound signal Q and third category sound signal Z; The first type of sound signal Y includes air sound signal Y1, wheel-rail sound signal Y2, high-speed railway station equipment sound signal Y3 and passenger conversation sound signal Y4; The second type of sound signal Q includes a passenger distress sound signal Q1; The third type of sound signal Z includes a landslide sound signal Z1, a tornado sound signal Z2 and an earthquake sound signal Z3; Step S202, retrieve the energy characteristic parameter matrix of the first type of sound signal Y, the second type of sound signal Q and the third type of sound signal Z, number the energy characteristic parameter matrix, and obtain the air sound signal Y11, the wheel-rail sound signal Y22, the high-speed railway station equipment sound signal Y33, the passenger conversation sound signal Y44, the passenger calling for help sound signal Q11, the landslide sound signal Z11, the tornado sound signal Z22 and the earthquake sound signal Z33; Step S203, input the air sound signal Y11, wheel-rail sound signal Y22, high-speed railway station equipment sound signal Y33, passenger conversation sound signal Y44, passenger call for help sound signal Q11, landslide sound signal Z11, tornado sound signal Z22 and earthquake sound signal Z33 into the support vector data description algorithm model, perform sound signal data recognition, obtain the sample sound signal data category, and obtain a sound signal recognition model based on the support vector data description algorithm.

4. A rail transit broadcasting method based on sound field analysis as claimed in claim 3, characterized in that: Step S300 specifically includes: Step S301, inputting the collected real-time sound signal data into the trained sound signal recognition model to obtain the real-time sound signal data category of the sound signal recognition model; Step S302, issuing a warning signal according to the real-time sound signal data category, the real-time sound signal data category includes a first-category sound signal Y, a second-category sound signal Q and a third-category sound signal Z, and the warning signal includes a distress warning signal and a natural disaster warning signal; When the real-time sound signal data category belongs to the first category of sound signal Y, no warning signal is issued; When the real-time sound signal data category belongs to the second type of sound signal Q, a distress warning signal is issued; When the real-time sound signal data category belongs to the third type of sound signal Z, a natural disaster warning signal is issued.

5. A rail transit broadcasting method based on sound field analysis as claimed in claim 4, characterized in that: Step S400 specifically includes: Step S401, the energy distribution matrix library is used to store the energy coefficient distribution matrix corresponding to each real-time sound signal data category, and according to the real-time sound signal data category output by the sound signal recognition model, find the matching energy coefficient distribution matrix; Step S402, inputting the broadcast voice signal into a spectrum display to obtain a spectrum diagram of the broadcast voice signal; Step S403, extracting the frequency band energy characteristics of the spectrogram using the Mel-frequency cepstrum coefficients to obtain an energy characteristic parameter matrix of the frequency band energy of the broadcast voice signal; Step S404: multiply the energy characteristic parameter matrix of the broadcast voice signal frequency band energy by the energy coefficient distribution matrix to obtain an adjusted energy characteristic parameter matrix.

6. A rail transit broadcasting method based on sound field analysis as claimed in claim 5, characterized in that: The specific process of establishing the energy characteristic parameter matrix of the broadcast voice signal frequency band energy is as follows: The spectrum of the broadcast voice signal is divided on the frequency axis into n frequency bands, which are denoted as Li to Ln; The energy values ​​of all time points in each frequency band are accumulated to obtain the matrix [Li, Li +1 , Li +2 , ...Ln], Li represents the total energy value of the Li band, the matrix [Li, Li +1 , Li +2 , ...Ln] is the energy characteristic parameter matrix of the broadcast voice signal.

7. A rail transit broadcasting method based on sound field analysis as claimed in claim 6, characterized in that: The process of establishing the energy characteristic parameter matrix of the real-time sound signal is as follows: The spectrum of the real-time sound signal is divided on the frequency axis into n frequency bands, which are denoted as Li1 to Ln1; The energy values ​​of all time points in each frequency band are accumulated to obtain the matrix [Li1, Li 1+1 , Li 1+2 , ...Ln1], Li1 represents the total energy value of Li1 band, the matrix [Li1, Li 1+1 , Li 1+2 , ...Ln1] is the energy characteristic parameter matrix of the real-time sound signal.

8. A rail transit broadcasting method based on sound field analysis as claimed in claim 7, characterized in that: The process of establishing the energy coefficient distribution matrix corresponding to the real-time sound signal data category is as follows: Obtain the energy characteristic parameter matrix [Li1, Li 1+1 Li 1+2 , ...Ln1], Li1 represents the total energy of Li1 Li1+Li 1+1 +Li 1+2 +, ... Ln1 / Li1 band energy; Normalize the energy values ​​of the Li1-Ln1 band to a range of 0 to 1, and obtain the matrix [Lg, Lg +1 , Lg +2 , ...Lg n ], the matrix [Lg, Lg +1 , Lg +2 , ...Lg n ] is the energy coefficient matrix of the real-time sound signal.

9. A rail transit broadcasting method based on sound field analysis as claimed in claim 8, characterized in that: The adjusted energy characteristic parameter matrix is ​​as follows: P=[Lg,Lg +1 , Lg +2 , ...Lg n ]*[Li,Li +1 , Li +2 , ...Ln], the matrix P is the energy characteristic parameter matrix of the adjusted broadcast voice signal.

10. A rail transit broadcasting method based on sound field analysis as claimed in claim 9, characterized in that: The specific process of normalizing the energy value to a range of 0 to 1 is as follows: S1=Li1+Li 1+1 +Li 1+2 +,...Ln1 / Li1,S2=Li1+Li 1+1 +Li 1+2 +,...Ln1 / Li 1+1 ; D = S1 + S1 + ... Sn; Lg=S1 / D,Lg +1 =S2 / D。

Images