Vehicle-mounted display screen interaction control method and system for railway vehicle

Abstract

The invention relates to the field of data processing, in particular to a rail vehicle-oriented vehicle-mounted display screen interaction control method and system, and the method comprises the steps: collecting and preprocessing original voice data, and obtaining a plurality of components of the original voice data; segmenting the original voice data and the plurality of components according to a corresponding rail vehicle speed time sequence to obtain a plurality of voice local segments; calculating the weights of the voice local segments according to the distribution characteristics of the plurality of voice local segments, the track speed sequential sequence and the distribution characteristic difference of the voice local segments in the time domain and the frequency domain; according to the weight corresponding to each voice local segment, performing weighted reconstruction on all components of the original voice data to obtain reconstructed voice data; and determining a user intention based on the reconstructed voice data, responding to the user intention, and outputting a response result on the vehicle-mounted display screen. Therefore, the accuracy of voice recognition can be effectively improved, and the interaction efficiency of the interaction terminal of the vehicle-mounted display screen of the railway vehicle is improved.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to a method and system for interactively controlling an on-board display screen of a rail vehicle. Background Art

[0002] The on-board display interactive terminal of rail vehicles is an integrated technology that allows passengers to interact with the vehicle's infotainment system through touch screens, gestures, voice or other input methods. During the interaction, it responds to various instructions from passengers to complete corresponding operations or provide information required by passengers, which can bring great convenience to passengers. For the voice interaction function, the control system can collect and analyze the voice data input by passengers, extract effective information to achieve intelligent interaction with passengers.

[0003] During the travel of rail vehicles, it may be inconvenient for passengers to use their hands to operate. The voice interaction function allows passengers to perform various operations by inputting simple voice commands. With the continuous development of voice interaction technology, many rail vehicles have begun to be equipped with on-board display screens with voice interaction functions. For example, a new generation of intercity intelligent EMUs that have recently started operation are equipped with a full-touch screen intelligent interaction terminal in each carriage. Passengers can query various information on the full-touch screen intelligent interaction terminal through simple voice commands, such as platform layout information, bus transfer information, weather information, temperature information inside and outside the car, tourist attractions information, etc., which can greatly improve the travel experience of passengers.

[0004] However, as a public space, rail vehicles can accommodate a large number of passengers. In this environment, passengers may engage in various activities that generate sound (such as chatting, making phone calls, watching videos, etc.). In addition, rail vehicles also generate continuous noise during travel (such as wind noise, noise from the contact between wheels and tracks, etc.). As a result, the on-board display interactive terminal is susceptible to more noise interference when collecting voice data, thereby affecting the recognition accuracy of voice commands, resulting in low accuracy and interaction efficiency of the on-board display interactive terminal, affecting the passenger experience. Summary of the invention

[0005] In order to solve the above technical problems, the object of the present invention is to provide an on-board display screen interactive control method for rail vehicles, comprising:

[0006] Collecting original voice data and preprocessing the original voice data to obtain multiple components of the original voice data;

[0007] Segmenting the original voice data and the multiple components according to the track speed time sequence corresponding to the original voice data to obtain multiple voice local segments of the original voice data and multiple voice local segments of each component;

[0008] For each speech partial segment of each component, a weight of the speech partial segment is calculated according to the distribution characteristics of the plurality of speech partial segments, the track speed time series, and the difference in distribution characteristics of the speech partial segment in the time domain and the frequency domain;

[0009] Performing weighted reconstruction on all components of the original speech data according to the weight corresponding to each speech local segment of each component to obtain reconstructed speech data;

[0010] The user intention is determined based on the reconstructed voice data, the user intention is responded to and the response result is output on the vehicle-mounted display screen.

[0011] Optionally, segmenting the original voice data and the multiple components according to the track speed time series sequence corresponding to the original voice data to obtain multiple voice local segments of the original voice data and multiple voice local segments of each component includes:

[0012] The original voice data and multiple components are segmented according to the track vehicle speed time series sequence corresponding to the original voice data using a preset segmentation method to obtain multiple voice local segments of the original voice data and multiple voice local segments of each component, wherein each voice local segment has a corresponding approximate speed value.

[0013] Optionally, for each speech partial segment of each component, calculating the weight of the speech partial segment according to the distribution characteristics of the plurality of speech partial segments, the track speed time series, and the difference in distribution characteristics of the speech partial segment in the time domain and the frequency domain, comprises:

[0014] Clustering the speech local segments of all components according to the approximate speed value corresponding to each speech local segment to obtain multiple clusters;

[0015] For each speech local segment of each component, the initial noise characteristic value of the speech local segment is calculated according to the overall situation of the cluster to which the speech local segment belongs and the correlation between the change trend of the speech local segment and the change trend of the track speed;

[0016] Calculate the advanced noise characteristic value of the speech local segment according to the initial noise characteristic value of the speech local segment, the possibility that the speech local segment belongs to sudden noise, and the track speed change corresponding to the speech local segment;

[0017] The weight of the speech local segment is calculated according to the advanced noise feature value of the speech local segment and the difference in distribution characteristics of the speech local segment in the time domain and the frequency domain.

[0018] Optionally, the overall situation of the cluster to which the speech partial segment belongs includes the distribution characteristics and quantity of all speech partial segments in the cluster to which the speech partial segment belongs;

[0019] The correlation between the change trend of the speech local segment and the change trend of the track vehicle speed is calculated based on the Pearson correlation coefficients of all speech local segments in the cluster where the speech local segment belongs and the corresponding speed local segments.

[0020] Optionally, the possibility that the speech local segment belongs to burst noise is calculated based on the difference between a first speech local segment and a second speech local segment corresponding to the speech local segment;

[0021] The track vehicle speed change corresponding to the speech partial segment is the difference between the approximate speed value corresponding to the speech partial segment and the approximate speed value corresponding to the previous speech partial segment of its component.

[0022] Optionally, for the d-th speech local segment of a component, its corresponding first speech local segment is the speech local segment obtained by reconstructing the d-th speech local segment of each of the remaining components, and its corresponding second speech local segment is the d-th speech local segment in the original speech data, where d is an integer greater than or equal to 1 and less than or equal to m, and m is the total number of speech local segments in a component.

[0023] Optionally, before calculating the weight of the speech local segment, the method further includes:

[0024] For each speech partial segment, the distribution characteristics of the speech partial segment are calculated according to the distribution disorder degree of all data points in the speech partial segment and the frequency resolution of the speech partial segment in the original speech data change curve.

[0025] Optionally, the preprocessing the original voice data to obtain multiple components of the original voice data includes:

[0026] Performing EMD decomposition on the original speech data to obtain a plurality of intrinsic mode function (IMF) components of the original speech data.

[0027] Optionally, the using a preset segmentation method to segment the original voice data and the multiple components according to the track speed time series sequence corresponding to the original voice data to obtain multiple voice local segments of the original voice data and multiple voice local segments of each component includes:

[0028] Acquire a track speed time sequence sequence corresponding to the original voice data, wherein the original voice data and the track speed time sequence sequence corresponding to the original voice data have the same starting time and the same ending time;

[0029] The track speed time series corresponding to the original voice data is segmented using an adaptive piecewise constant approximation APCA segmentation method to obtain a plurality of speed local segments and an approximate speed value corresponding to each speed local segment;

[0030] According to the starting time and the ending time of each speed partial segment, the original speech data and the multiple components are segmented to obtain multiple speech partial segments of the original speech data and multiple speech partial segments of each component.

[0031] The present invention also provides an on-board display screen interactive control system for rail vehicles, the system comprising:

[0032] A processing module, used for collecting original voice data and preprocessing the original voice data to obtain multiple components of the original voice data;

[0033] A segmentation module, used for segmenting the original voice data and multiple components according to the track speed time sequence corresponding to the original voice data, to obtain multiple voice local segments of the original voice data and multiple voice local segments of each component;

[0034] A calculation module, configured to calculate, for each speech partial segment of each component, a weight of the speech partial segment according to distribution characteristics of the plurality of speech partial segments, a track speed time series, and a difference in distribution characteristics of the speech partial segment in the time domain and the frequency domain;

[0035] A reconstruction module, used for weightedly reconstructing all components of the original speech data according to the weight corresponding to each speech local segment of each component, to obtain reconstructed speech data;

[0036] A response module is used to determine the user intention based on the reconstructed voice data, respond to the user intention and output the response result on the vehicle display screen.

[0037] The present invention has the following beneficial effects: after preprocessing the original voice data to obtain multiple components, the original voice data and the multiple components are further segmented according to the corresponding rail vehicle speed time series sequence to obtain multiple voice local segments, and then the voice local segment of each component is assigned a weight according to the distribution characteristics of the multiple voice local segments, the rail vehicle speed time series sequence, and the difference in the distribution characteristics of the voice local segments in the time domain and the frequency domain, and all components are weightedly reconstructed according to the assigned weight, so that the reconstructed voice data has less noise and higher quality, which can effectively improve the accuracy of voice recognition, improve the interaction efficiency of the rail vehicle on-board display screen interactive terminal, and thus improve the passenger experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0039] Figure 1 A first schematic diagram of an on-board display screen interactive control method for a rail vehicle provided by an embodiment of the present invention;

[0040] Figure 2 A second schematic diagram of the on-board display screen interactive control method for rail vehicles provided by an embodiment of the present invention;

[0041] Figure 3 A schematic diagram of an on-board display screen interactive control system for rail vehicles provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0042] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the on-board display screen interactive control method for rail vehicles proposed by the present invention, its specific implementation method, structure, features and effects, in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0043] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0044] The present invention provides an on-board display screen interactive control method for rail vehicles, and the scheme of the present invention is described in detail below in conjunction with the accompanying drawings.

[0045] See also Figure 1 , which shows a first schematic diagram of an on-board display screen interactive control method for a rail vehicle provided by an embodiment of the present invention, the on-board display screen interactive control method for a rail vehicle (hereinafter referred to as "the method") comprises the following steps:

[0046] Step 101: collect original speech data and pre-process the original speech data to obtain multiple components of the original speech data.

[0047] In this step, the method collects original voice data and preprocesses the collected original voice data to obtain multiple components of the original voice data. The method can collect the voice data of the passenger by using an audio collection device. On a rail vehicle equipped with an on-board display screen with a voice interaction function, the method can collect the voice data of the passenger by an audio collection device set on the on-board display screen. In some embodiments, in order to accurately respond to the needs of passengers, the method can collect voice data after receiving a trigger operation from the passenger. For example, a trigger switch can be displayed on the on-board display screen. When the on-operation of the trigger switch is received, the method starts to collect the voice data of the passenger and preprocess it, so as to avoid being in a state of collecting voice data all the time and avoid waste of resources. The method can also collect voice data all the time, and only when it is recognized that the voice data contains preset words, the subsequent collected voice data is further preprocessed, so as to avoid continuous preprocessing of the collected voice data to reduce the workload of the device.

[0048] The preprocessing of the original voice data includes decomposing the original voice data to obtain multiple components of the original voice data. In some embodiments, the preprocessing can be decomposed using an EMD (Empirical Mode Decomposition) algorithm, and accordingly, the components of the original voice data are IMF (Intrinsic Mode Function) components. The EMD algorithm is an algorithm suitable for nonlinear and non-stationary signal processing. It decomposes a complex signal into multiple IMF components representing different frequencies based on the extreme points of the original signal. The technology of decomposing voice data using the EMD algorithm belongs to the scope of the prior art and will not be described in detail here.

[0049] It is understandable that other noises usually exist when collecting the original voice data of passengers. For the case of collecting the original voice data of passengers during the travel of rail vehicles, some continuous noise and / or sudden noise may appear. Specifically, continuous noise includes wind noise generated by a certain speed during the travel of rail vehicles, noise caused by the contact between wheels and rails, etc. Continuous noise is usually wide-band, and the noise intensity is easy to change with the changes in rail vehicle speed and other environmental conditions. Sudden noise includes the sound generated by strong friction and impact between components of the braking system during emergency braking and / or various sounds that passengers may make when they are on rail vehicles (such as door closing, passenger conversation, luggage rolling, etc.). Sudden noise usually appears suddenly, the noise intensity may be very high, and its energy on the spectrum may be very concentrated (it may be particularly strong in certain specific frequency ranges). Various noises mixed with voice signals are likely to cause greater interference to the recognition of voice signals. Therefore, in this embodiment, by performing EMD decomposition on the collected original speech data, the original speech data mixed with various noises and speech signals is decomposed into multiple components with different frequency characteristics, so that the components with different frequency characteristics can be processed separately to improve the recognition accuracy of the speech signal.

[0050] Step 102: segment the original voice data and multiple components according to the track speed time series sequence corresponding to the original voice data to obtain multiple voice local segments of the original voice data and multiple voice local segments of each component.

[0051] In this step, the method segments the original voice data and multiple components according to the track speed time series sequence corresponding to the original voice data, and obtains multiple voice local segments of the original voice data and multiple voice local segments of each component.

[0052] The wind noise generated by the fluid dynamics interaction between the rail vehicle and the air during the running process will easily change with the change of the rail vehicle speed. The higher the rail vehicle speed, the greater the relative movement speed between the rail vehicle and the air, and accordingly, the wind noise intensity will also increase. It can be seen that the wind noise generated during the running of the rail vehicle will have different intensities at different rail vehicle speeds. When the speed remains unchanged, the continuous noise generated during the running of the rail vehicle will have some similar characteristics. For example, when the rail vehicle speed is constant, the frequency distribution of the continuous noise will also remain in a relatively stable state, and in the absence of other interference, the noise intensity will also remain in a relatively stable range; when the track conditions and the structure of the rail vehicle remain unchanged, the noise caused by factors such as rail joints and wheel out-of-roundness may recur periodically. When the human body speaks, the frequency and amplitude of the speech signals of different words are different at different time periods, because the speech signal is composed of a series of different phonemes, each of which has its own unique frequency and amplitude characteristics.

[0053] It can be seen that within the time period with the same track vehicle speed, the sounds with similar distribution characteristics in the original voice data are more likely to belong to continuous noise. Therefore, in this embodiment, the original voice data and multiple components are segmented according to the track vehicle speed time series sequence corresponding to the original voice data to obtain multiple voice local segments of the original voice data and multiple voice local segments of each component, so that the rail vehicles corresponding to each voice local segment are the same or similar, which is convenient for subsequent further processing.

[0054] The method can use a preset segmentation method to segment the original voice data and multiple components according to the track speed time sequence corresponding to the original voice data, and obtain multiple voice local segments of the original voice data and multiple voice local segments of each component, wherein each voice local segment has a corresponding approximate speed value. In some embodiments, the method can obtain the track speed time sequence corresponding to the original voice data, wherein the original voice data and the track speed time sequence corresponding to it have the same starting time and the same ending time; use the APCA (Adaptive Piecewise Constant Approximation) segmentation method to segment the track speed time sequence corresponding to the original voice data, and obtain multiple speed local segments and the approximate speed value corresponding to each speed local segment; according to the starting time and ending time of each speed local segment, the original voice data and multiple components are segmented to obtain multiple voice local segments of the original voice data and multiple voice local segments of each component. The technology of segmenting using the APCA segmentation method belongs to the scope of the prior art and will not be described in detail here.

[0055] The track speed time sequence corresponding to the original voice data is a track speed time sequence with the same starting time and ending time as the original voice data. The method can synchronously acquire the track speed time sequence corresponding to the original voice data when acquiring the original voice data, or further acquire the track speed time sequence corresponding to the original voice data according to the starting time and ending time of the original voice data after acquiring the original voice data.

[0056] Step 103: For each speech local segment of each component, the weight of the speech local segment is calculated according to the distribution characteristics of the multiple speech local segments, the track speed time series, and the difference in distribution characteristics of the speech local segment in the time domain and the frequency domain.

[0057] In this step, for each speech local segment of each component, the method calculates the weight of the speech local segment according to the distribution characteristics of the multiple speech local segments, the track speed time series, and the difference in the distribution characteristics of the speech local segment in the time domain and the frequency domain. Specifically, the method can cluster the speech local segments of all components according to the approximate speed value corresponding to each speech local segment to obtain multiple clusters; for each speech local segment of each component, the initial noise characteristic value of the speech local segment is calculated according to the overall situation of the cluster to which the speech local segment belongs and the correlation between the change trend of the speech local segment and the change trend of the track speed; the advanced noise characteristic value of the speech local segment is calculated according to the initial noise characteristic value of the speech local segment, the possibility that the speech local segment belongs to sudden noise, and the change of the track speed corresponding to the speech local segment; the weight of the speech local segment is calculated according to the advanced noise characteristic value of the speech local segment and the difference in the distribution characteristics of the speech local segment in the time domain and the frequency domain.

[0058] In this embodiment, the speech local segments of all components are clustered according to the approximate speed value corresponding to each speech local segment to obtain multiple clusters, and then the weights of the speech local segments are further calculated based on the distribution characteristics of the speech local segments themselves and the situation of the clusters they belong to, so as to assign weights to the noise local segments of different categories respectively, thereby improving the speech recognition effect.

[0059] Step 104: weightedly reconstruct all components of the original speech data according to the weight corresponding to each speech local segment of each component to obtain reconstructed speech data.

[0060] In this step, the method performs weighted reconstruction on all components of the original voice data according to the weight corresponding to each speech local segment of each component to obtain reconstructed voice data. Specifically, the method can perform weighted reconstruction on each component according to the weight corresponding to each speech local segment of each component to obtain multiple reconstructed components. Then, all reconstructed components are smoothed and superimposed to obtain reconstructed voice data.

[0061] Step 105: determine the user intention based on the reconstructed voice data, respond to the user intention and output the response result on the vehicle display screen.

[0062] In this step, the method determines the user intention based on the reconstructed voice data, responds to the user intention and outputs the response result on the on-board display screen. Specifically, the method can identify and analyze the reconstructed voice data to determine the user intention, respond to the user intention after determining the user intention and output the response result on the on-board display screen of the rail vehicle. The technology of identifying and analyzing voice data and responding to user intention belongs to the scope of the prior art and will not be described in detail here. In some embodiments, when the method outputs the response result on the on-board display screen, it can also simultaneously broadcast the response result through an audio output device.

[0063] In an embodiment of the present invention, after preprocessing the original voice data to obtain multiple components, the method further performs segmentation processing on the original voice data and the multiple components according to the corresponding rail vehicle speed time series sequence to obtain multiple voice local segments, and then assigns weights to the voice local segments of each component according to the distribution characteristics of the multiple voice local segments, the rail vehicle speed time series sequence, and the distribution characteristics of the voice local segments in the time domain and the frequency domain, and performs weighted reconstruction on all components according to the assigned weights, so that the reconstructed voice data has less noise and higher quality, which can effectively improve the accuracy of voice recognition, improve the interaction efficiency of the rail vehicle on-board display screen interactive terminal, and thus improve the passenger experience.

[0064] Optionally, segmenting the original voice data and multiple components according to the track speed timing sequence corresponding to the original voice data to obtain multiple voice local segments of the original voice data and multiple voice local segments of each component includes: using a preset segmentation method to segment the original voice data and multiple components according to the track speed timing sequence corresponding to the original voice data to obtain multiple voice local segments of the original voice data and multiple voice local segments of each component, wherein each voice local segment has a corresponding approximate speed value.

[0065] In this embodiment, the method uses a preset segmentation method to segment the original voice data and multiple components according to the rail vehicle speed time series sequence corresponding to the original voice data, to obtain multiple voice local segments of the original voice data and multiple voice local segments of each component, so that the rail vehicle corresponding to each voice local segment is the same or similar, which is convenient for subsequent further processing.

[0066] Specifically, the method can obtain the rail vehicle speed time series sequence corresponding to the original voice data, wherein the original voice data and the rail vehicle speed time series sequence corresponding to it have the same starting time and the same ending time; use the adaptive piecewise constant approximation APCA segmentation method to segment the rail vehicle speed time series sequence corresponding to the original voice data to obtain multiple speed local segments and the approximate speed value corresponding to each speed local segment; according to the starting time and the ending time of each speed local segment, the original voice data and the multiple components are segmented to obtain multiple voice local segments of the original voice data and multiple voice local segments of each component.

[0067] Optionally, before calculating the weight of the speech local segment, the method also includes: for each speech local segment, calculating the distribution characteristics of the speech local segment according to the distribution disorder of all data points in the speech local segment and the frequency resolution of the speech local segment in the original speech data change curve.

[0068] In this embodiment, for each speech local segment, the method calculates the distribution characteristics of the speech local segment according to the distribution disorder degree of all data points in the speech local segment and the frequency resolution of the speech local segment in the original speech data change curve.

[0069] Specifically, for the dth speech local segment of the ith component, the method may calculate its distribution feature C using the following formula: i,d :

[0070]

[0071] Among them, C i,d represents the distribution characteristics of the dth speech local segment of the ith component, j represents the jth type of data in the dth speech local segment of the ith component, F represents the number of data types in the dth speech local segment of the ith component, and g i,d,j represents the number of the jth type of data in the dth local segment of speech of the ith component, G represents the total number of data in the dth local segment of speech of the ith component, ln() represents the natural logarithm function; t o Indicates the time node where the data of the oth extreme point in the dth speech local segment of the ith component is located, t o+1 represents the time node where the data of the o+1th extreme value point in the dth speech local segment of the ith component is located, n represents the number of groups of two adjacent extreme value points in the dth speech local segment of the ith component, and w represents the wth group of adjacent extreme value points in the dth speech local segment of the ith component.

[0072] In the above formula, It represents the proportion of the jth data point in the dth speech local segment of the ith component to all data points in the dth speech local segment of the ith component. represents the distribution disorder of all data points in the dth speech local segment of the ith component, The larger the value, The larger the value, the more chaotic the distribution of all data points in the dth speech local segment of the ith component.

[0073] Indicates the frequency resolution of the d-th speech local segment of the ith component in the original speech data change curve. The larger the value, the higher the frequency resolution of the d-th speech local segment of the ith component in the original speech data change curve.

[0074] Optionally, the preprocessing the original speech data to obtain multiple components of the original speech data includes: performing empirical mode decomposition (EMD) decomposition on the original speech data to obtain multiple intrinsic mode function (IMF) components of the original speech data.

[0075] Optionally, the using of a preset segmentation method to segment the original voice data and multiple components according to the track speed time series sequence corresponding to the original voice data to obtain multiple voice local segments of the original voice data and multiple voice local segments of each component includes: acquiring the track speed time series sequence corresponding to the original voice data, wherein the original voice data and the track speed time series sequence corresponding to it have the same starting time and the same ending time; using an adaptive piecewise constant approximation APCA segmentation method to segment the track speed time series sequence corresponding to the original voice data to obtain multiple speed local segments and an approximate speed value corresponding to each speed local segment; segmenting the original voice data and multiple components according to the starting time and the ending time of each speed local segment to obtain multiple voice local segments of the original voice data and multiple voice local segments of each component.

[0076] Please read further Figure 2 , Figure 2 is a second schematic diagram of the on-board display screen interactive control method for rail vehicles provided by an embodiment of the present invention, such as Figure 2 As shown, the method comprises the following steps:

[0077] Step 201: collect original speech data and pre-process the original speech data to obtain multiple components of the original speech data.

[0078] The step 201 and Figure 1 The step 101 shown is the same and will not be repeated here.

[0079] Step 202: Use a preset segmentation method to segment the original voice data and multiple components according to the track vehicle speed time series sequence corresponding to the original voice data, to obtain multiple voice local segments of the original voice data and multiple voice local segments of each component, wherein each voice local segment has a corresponding approximate speed value.

[0080] In this step, the method uses a preset segmentation method to segment the original voice data and multiple components according to the track speed time sequence corresponding to the original voice data, and obtains multiple voice local segments of the original voice data and multiple voice local segments of each component, wherein each voice local segment has a corresponding approximate speed value. Specifically, the method can obtain the track speed time sequence corresponding to the original voice data, wherein the original voice data and the corresponding track speed time sequence have the same starting time and the same ending time; use the adaptive piecewise constant approximation APCA segmentation method to segment the track speed time sequence corresponding to the original voice data, and obtain multiple speed local segments and the approximate speed value corresponding to each speed local segment; according to the starting time and ending time of each speed local segment, the original voice data and multiple components are segmented to obtain multiple voice local segments of the original voice data and multiple voice local segments of each component. In this way, the original voice data and multiple components are segmented according to the track vehicle speed time series sequence corresponding to the original voice data, so as to obtain multiple voice local segments of the original voice data and multiple voice local segments of each component, so that the rail vehicle corresponding to each voice local segment is the same or similar, which is convenient for subsequent further processing.

[0081] Step 203: clustering the speech partial segments of all components according to the approximate speed value corresponding to each speech partial segment to obtain a plurality of clusters.

[0082] In this step, the method clusters the speech local segments of all components according to the approximate speed value corresponding to each speech local segment to obtain multiple clusters.

[0083] Since the continuous noise generated during the running of rail vehicles is related to the rail speed, the greater the rail speed, the greater the continuous noise intensity generated. In this way, among all the speech local segments of each component, the speech local segments with the corresponding rail speeds closer to each other will contain more similar distribution characteristics of continuous noise data. Therefore, in this embodiment, the method clusters the speech local segments of all components according to the approximate speed value corresponding to each speech local segment to obtain multiple clusters. Specifically, the difference between the approximate speed values ​​corresponding to the speech local segments (i.e., the absolute value of the difference) can be used as the clustering distance to cluster the speech local segments of all components. In some embodiments, the K-means clustering algorithm can be used for clustering.

[0084] Step 204: For each speech local segment of each component, the initial noise characteristic value of the speech local segment is calculated according to the overall situation of the cluster to which the speech local segment belongs and the correlation between the change trend of the speech local segment and the change trend of the track vehicle speed.

[0085] In this step, for each speech local segment of each component, the method calculates the initial noise characteristic value of the speech local segment according to the overall situation of the cluster to which the speech local segment belongs and the correlation between the change trend of the speech local segment and the change trend of the track vehicle speed.

[0086] Specifically, for the dth speech local segment of the ith component, its initial noise characteristic value PC can be calculated using the following formula: i,d :

[0087]

[0088] Among them, PC i,d represents the initial noise characteristic value of the dth speech local segment of the ith component; R i,d represents the correlation between the changing trend of all speech local segments in the cluster where the d-th speech local segment of the ith component belongs and the changing trend of the track speed; exp() represents the natural exponential function, l represents the l-th speech local segment in the cluster where the d-th speech local segment of the ith component belongs except the d-th speech local segment, L i,d represents the total number of speech local segments included in the cluster where the dth speech local segment of the ith component belongs, C i.d represents the distribution characteristics of the dth local speech segment of the ith component, C i.d,l Represents the distribution characteristics of the lth speech local segment in the cluster where the dth speech local segment of the i-th component belongs, except the dth speech local segment.

[0089] In the above formula, Indicates the difference in distribution characteristics between the dth speech local segment of the ith component and other speech local segments in its cluster. The larger the value, the greater the difference in distribution characteristics between the dth speech local segment of the ith component and other speech local segments in its cluster, and the lower the distribution characteristic similarity. i,d represents the correlation between the changing trend of all speech local segments in the cluster where the d-th speech local segment of the ith component belongs and the changing trend of the track speed, R i,d The larger the value is, the more consistent the change trend of all speech local segments in the cluster where the d-th speech local segment of the ith component belongs is with the change trend of the track speed, and the higher the possibility that the ith component is continuous noise.

[0090] Step 205: Calculate the advanced noise characteristic value of the speech partial segment according to the initial noise characteristic value of the speech partial segment, the possibility that the speech partial segment belongs to sudden noise, and the track speed change corresponding to the speech partial segment.

[0091] In this step, the method calculates the advanced noise characteristic value of the speech local segment according to the initial noise value of the speech local segment, the possibility that the speech local segment belongs to sudden noise, and the change of the track speed corresponding to the speech local segment. It can be understood that when the rail vehicle performs emergency braking, strong friction and impact will be generated between the components of the braking system, which is easy to generate sudden noise. Therefore, the method calculates the advanced noise characteristic value of the speech local segment in combination with the change of the track speed corresponding to the speech local segment.

[0092] Specifically, for the dth speech local segment of the ith component, its advanced noise feature value G can be calculated using the following formula: i,d :

[0093]

[0094] Among them, G i,d represents the advanced noise feature value of the dth speech local segment of the ith component, PC i,d represents the initial noise feature value of the dth speech local segment of the ith component, C′ i,d represents the distribution characteristics of the speech local segment obtained by reconstructing the d-th speech local segment of each of the remaining components, represents the distribution characteristics of the dth local segment of speech in the original speech data; v i,d,d―1 Indicates the track speed change corresponding to the d-th speech local segment of the ith component.

[0095] In the above formula, It indicates the difference in distribution characteristics (i.e., the absolute value of the difference) between the speech local segment obtained by reconstructing the dth speech local segment of each component and the dth speech local segment of the original speech data at the time period corresponding to the dth speech local segment. The smaller the difference, the lower the influence of the dth speech local segment of the ith component in the original speech data, and the lower the possibility that it is a burst noise. i,d,d―1 for The weighting coefficient, v i,d,d―1 When the value is larger, the track speed changes more, the possibility of emergency braking of the track vehicle increases, and the possibility of sudden noise generation increases. The higher the credibility.

[0096] Step 206: Calculate the weight of the speech local segment according to the advanced noise feature value of the speech local segment and the difference in distribution characteristics of the speech local segment in the time domain and the frequency domain.

[0097] In this step, the method calculates the weight of the local segment of speech according to the advanced noise characteristic value of the local segment of speech and the difference in the distribution characteristics of the local segment of speech in the time domain and the frequency domain. Since the noise generated by the vehicle during driving and the passenger's command voice have different generation mechanisms and propagation paths, they will appear as sound data of different frequencies in many cases. However, rail vehicles are a public space with many passengers. In this environment, many passengers' conversations, phone calls and other activities may generate sounds with the same or similar frequencies as the command voice. In this case, it is necessary to distinguish the command voice from the voices of other passengers to reduce the interference of other passengers' explanations on the command voice.

[0098] Normally, when passengers issue voice commands, they usually use relatively short sentences to express clear needs, and the expressions are relatively clear and continuous. The distribution characteristics of clear and continuous voice signals in the time domain and frequency domain are relatively small, because such voice signals have a certain rhythm and high consistency in sound intensity, and their time domain waveforms and frequency domain spectrograms are relatively standardized. Unlike the clear purpose of voice commands, the content of other passengers' conversations is often more casual, and the corresponding voice signals may be more vague and intermittent, and the distribution characteristics in the time domain and frequency domain may be quite different. Therefore, in this step, the method further determines the possibility that the local segment of voice includes command voice data based on the difference in the distribution characteristics of the local segment of voice in the time domain and frequency domain, and further calculates the weight of the local segment of voice accordingly.

[0099] Specifically, for the dth speech local segment of the ith component, the method can calculate its weight G′ according to the following formula: i,d :

[0100]

[0101] Among them, G′ i,d represents the weight of the dth local segment of the ith component, exp() represents the natural exponential function; G i,d represents the advanced noise feature value of the dth speech local segment of the ith component, C i,d represents the distribution characteristics of the dth local segment of the i-th component, s represents the sth data point of the dth local segment of the i-th component in the frequency domain, J represents the number of data points of the dth local segment of the i-th component in the frequency domain, h′ i,s represents the size of the sth data point of the dth speech local segment of the ith component in the frequency domain, σ′ i,srepresents the variance of all data points of the d-th local speech segment of the ith component in the frequency domain, and norm() represents the normalization function.

[0102] In the above formula, It represents the difference in distribution characteristics of data points in the time domain and frequency domain in the d-th local speech segment of the ith component (i.e., the absolute value of the difference). The larger the value is, the greater the difference in distribution characteristics of the d-th local speech segment of the ith component in the time domain and frequency domain, and the greater the possibility that the local speech segment contains interference factors. Represents the final noise characteristic value of the dth speech local segment of the ith component. The larger the value, the greater the possibility of noise in the speech local segment, and the lower its importance. The corresponding weight G′ i,d The smaller it is.

[0103] Step 207: weightedly reconstruct all components of the original speech data according to the weight corresponding to each speech local segment of each component to obtain reconstructed speech data.

[0104] Step 208: Determine the user intention based on the reconstructed voice data, respond to the user intention and output the response result on the vehicle display screen.

[0105] Steps 207 and 208 are Figure 1 Steps 104 and 105 in the illustrated embodiment are the same and will not be described in detail herein.

[0106] In this embodiment, after preprocessing the original voice data to obtain multiple components, the method further performs segmentation processing on the original voice data and the multiple components according to the corresponding rail vehicle speed time series sequence to obtain multiple voice local segments, and clusters the voice local segments of all components according to the approximate speed value corresponding to each voice local segment to obtain multiple clusters; then, according to the overall situation of the cluster where the voice local segment is located, the rail vehicle speed time series sequence, and the distribution characteristic differences of the voice local segment in the time domain and the frequency domain, a weight is assigned to the voice local segment of each component, and all components are weightedly reconstructed according to the assigned weight, so that the reconstructed voice data has less noise and higher quality, which can effectively improve the accuracy of voice recognition, improve the interaction efficiency of the rail vehicle on-board display screen interactive terminal, and thus improve the passenger experience.

[0107] Optionally, the overall situation of the cluster where the speech local segment is located includes the distribution characteristics and quantity of all speech local segments in the cluster where the speech local segment is located; the correlation between the change trend of the speech local segment and the change trend of the track vehicle speed is calculated based on the Pearson correlation coefficients of all speech local segments in the cluster where the speech local segment is located and the corresponding speed local segments.

[0108] Specifically, the average of the Pearson correlation coefficients of all speech local segments in the cluster where the speech local segment belongs and the corresponding speed local segments can be calculated as the correlation between the change trend of the speech local segment and the change trend of the track vehicle speed.

[0109] Optionally, the possibility that the speech local segment belongs to burst noise is calculated based on the difference between a first speech local segment and a second speech local segment corresponding to the speech local segment; the track vehicle speed change corresponding to the speech local segment is the difference between the approximate speed value corresponding to the speech local segment and the approximate speed value corresponding to the previous speech local segment of its component.

[0110] In this embodiment, the method calculates the possibility that the speech local segment belongs to burst noise based on the difference between the first speech local segment and the second speech local segment corresponding to the speech local segment. For the d-th speech local segment of a component, its corresponding first speech local segment is the speech local segment obtained by reconstructing the d-th speech local segment of each of the remaining components, and its corresponding second speech local segment is the d-th speech local segment in the original speech data, wherein d is an integer greater than or equal to 1 and less than or equal to m, and m is the total number of speech local segments in a component.

[0111] The track speed change corresponding to the speech local segment can be the difference between the approximate speed value corresponding to the speech local segment and the approximate speed value corresponding to the previous speech local segment of the component in which it is located. For example, for the dth speech local segment of the i-th component, the track speed change v corresponding to it is i,d,d―1 is the absolute value of the difference between the approximate speed value corresponding to the dth speech local segment of the ith component and the approximate speed value corresponding to the d-1th speech local segment of the ith component. It should be noted that, for the first speech local segment in the component (i.e., d=1), since there is no previous speech local segment in the component, the average track speed in the preset time period before the time period corresponding to the first speech local segment can be obtained, and the absolute value of the difference between the average track speed and the approximate speed value corresponding to the first speech local segment is calculated as the track speed change corresponding to the first speech local segment.

[0112] Optionally, for the d-th speech local segment of a component, its corresponding first speech local segment is the speech local segment obtained by reconstructing the d-th speech local segment of each of the remaining components, and its corresponding second speech local segment is the d-th speech local segment in the original speech data, where d is an integer greater than or equal to 1 and less than or equal to m, and m is the total number of speech local segments in a component.

[0113] The embodiment of the present invention also provides an on-board display screen interactive control system for rail vehicles, please refer to Figure 3 The on-board display screen interactive control system 300 for rail vehicles includes:

[0114] The processing module 301 is used to collect original voice data and pre-process the original voice data to obtain multiple components of the original voice data;

[0115] A segmentation module 302 is used to segment the original voice data and multiple components according to the track speed time sequence corresponding to the original voice data, so as to obtain multiple voice local segments of the original voice data and multiple voice local segments of each component;

[0116] A calculation module 303 is used to calculate the weight of each speech partial segment of each component according to the distribution characteristics of the plurality of speech partial segments, the track speed time series, and the difference in distribution characteristics of the speech partial segment in the time domain and the frequency domain;

[0117] A reconstruction module 304, configured to perform weighted reconstruction on all components of the original speech data according to the weight corresponding to each speech local segment of each component, to obtain reconstructed speech data;

[0118] The response module 305 is used to determine the user intention based on the reconstructed voice data, respond to the user intention and output the response result on the vehicle display screen.

[0119] Optionally, the segmentation module 302 segments the original voice data and multiple components according to the track speed time series sequence corresponding to the original voice data to obtain multiple voice local segments of the original voice data and multiple voice local segments of each component, including: using a preset segmentation method to segment the original voice data and multiple components according to the track speed time series sequence corresponding to the original voice data to obtain multiple voice local segments of the original voice data and multiple voice local segments of each component, wherein each voice local segment has a corresponding approximate speed value.

[0120] Optionally, for each speech local segment of each component, the calculation module 303 calculates the weight of the speech local segment according to the distribution characteristics of the multiple speech local segments, the track speed time series, and the difference in distribution characteristics of the speech local segment in the time domain and the frequency domain, including: clustering the speech local segments of all components according to the approximate speed value corresponding to each speech local segment to obtain multiple clusters; for each speech local segment of each component, calculating the initial noise characteristic value of the speech local segment according to the overall situation of the cluster to which the speech local segment belongs and the correlation between the change trend of the speech local segment and the change trend of the track speed; calculating the advanced noise characteristic value of the speech local segment according to the initial noise characteristic value of the speech local segment, the possibility that the speech local segment belongs to sudden noise, and the change of the track speed corresponding to the speech local segment; calculating the weight of the speech local segment according to the advanced noise characteristic value of the speech local segment and the difference in distribution characteristics of the speech local segment in the time domain and the frequency domain.

[0121] Optionally, the overall situation of the cluster where the speech local segment is located includes the distribution characteristics and quantity of all speech local segments in the cluster where the speech local segment is located; the correlation between the change trend of the speech local segment and the change trend of the track vehicle speed is calculated based on the Pearson correlation coefficients of all speech local segments in the cluster where the speech local segment is located and the corresponding speed local segments.

[0122] Optionally, the possibility that the speech local segment belongs to burst noise is calculated based on the difference between a first speech local segment and a second speech local segment corresponding to the speech local segment; the track vehicle speed change corresponding to the speech local segment is the difference between the approximate speed value corresponding to the speech local segment and the approximate speed value corresponding to the previous speech local segment of its component.

[0123] Optionally, for the d-th speech local segment of a component, its corresponding first speech local segment is the speech local segment obtained by reconstructing the d-th speech local segment of each of the remaining components, and its corresponding second speech local segment is the d-th speech local segment in the original speech data, where d is an integer greater than or equal to 1 and less than or equal to m, and m is the total number of speech local segments in a component.

[0124] Optionally, before calculating the weight of the speech local segment, the calculation module 303 is also used to: for each speech local segment, calculate the distribution characteristics of the speech local segment according to the distribution disorder of all data points in the speech local segment and the frequency resolution of the speech local segment in the original speech data change curve.

[0125] Optionally, the processing module 301 preprocesses the original speech data to obtain multiple components of the original speech data, including: performing empirical mode decomposition (EMD) decomposition on the original speech data to obtain multiple intrinsic mode function (IMF) components of the original speech data.

[0126] Optionally, the segmentation module 302 uses a preset segmentation method to segment the original voice data and multiple components according to the rail speed time series sequence corresponding to the original voice data to obtain multiple voice local segments of the original voice data and multiple voice local segments of each component, including: obtaining the rail speed time series sequence corresponding to the original voice data, wherein the original voice data and the rail speed time series sequence corresponding to it have the same starting time and the same ending time; using the adaptive piecewise constant approximation APCA segmentation method to segment the rail speed time series sequence corresponding to the original voice data to obtain multiple speed local segments and an approximate speed value corresponding to each speed local segment; segmenting the original voice data and multiple components according to the starting time and the ending time of each speed local segment to obtain multiple voice local segments of the original voice data and multiple voice local segments of each component.

[0127] In this embodiment, after preprocessing the original voice data to obtain multiple components, the on-board display screen interactive control system for rail vehicles further performs segmentation processing on the original voice data and the multiple components according to the corresponding rail vehicle speed time series sequence to obtain multiple voice local segments, and then assigns weights to the voice local segments of each component according to the distribution characteristics of the multiple voice local segments, the rail vehicle speed time series sequence, and the distribution characteristics of the voice local segments in the time domain and the frequency domain, and performs weighted reconstruction on all components according to the assigned weights, so that the reconstructed voice data has less noise and higher quality, which can effectively improve the accuracy of voice recognition, improve the interaction efficiency of the rail vehicle on-board display screen interactive terminal, and thus enhance the user experience of passengers.

[0128] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0129] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A method for interactive control of a vehicle-mounted display screen for a rail vehicle, characterized in that: The method comprises: Collecting original voice data and preprocessing the original voice data to obtain multiple components of the original voice data; Segmenting the original voice data and the multiple components according to the track speed time sequence corresponding to the original voice data to obtain multiple voice local segments of the original voice data and multiple voice local segments of each component; For each speech partial segment of each component, a weight of the speech partial segment is calculated according to the distribution characteristics of the plurality of speech partial segments, the track speed time series, and the difference in distribution characteristics of the speech partial segment in the time domain and the frequency domain; Performing weighted reconstruction on all components of the original speech data according to the weight corresponding to each speech local segment of each component to obtain reconstructed speech data; The user intention is determined based on the reconstructed voice data, the user intention is responded to and the response result is output on the vehicle-mounted display screen.

2. The method for interactive control of a vehicle-mounted display screen for a rail vehicle according to claim 1, characterized in that: The step of segmenting the original voice data and the multiple components according to the track speed time sequence corresponding to the original voice data to obtain multiple voice local segments of the original voice data and multiple voice local segments of each component includes: The original voice data and multiple components are segmented according to the track vehicle speed time series sequence corresponding to the original voice data using a preset segmentation method to obtain multiple voice local segments of the original voice data and multiple voice local segments of each component, wherein each voice local segment has a corresponding approximate speed value.

3. The method for interactive control of a vehicle-mounted display screen for a rail vehicle according to claim 2, characterized in that: The step of calculating the weight of each speech partial segment of each component according to the distribution characteristics of the plurality of speech partial segments, the track speed time series, and the difference in distribution characteristics of the speech partial segment in the time domain and the frequency domain includes: Clustering the speech local segments of all components according to the approximate speed value corresponding to each speech local segment to obtain multiple clusters; For each speech local segment of each component, the initial noise characteristic value of the speech local segment is calculated according to the overall situation of the cluster to which the speech local segment belongs and the correlation between the change trend of the speech local segment and the change trend of the track speed; Calculate the advanced noise characteristic value of the speech local segment according to the initial noise characteristic value of the speech local segment, the possibility that the speech local segment belongs to sudden noise, and the track speed change corresponding to the speech local segment; The weight of the speech local segment is calculated according to the advanced noise feature value of the speech local segment and the difference in distribution characteristics of the speech local segment in the time domain and the frequency domain.

4. The method for interactive control of a vehicle-mounted display screen for a rail vehicle according to claim 3, characterized in that: The overall situation of the cluster where the speech local segment belongs to includes the distribution characteristics and quantity of all speech local segments in the cluster where the speech local segment belongs to; The correlation between the change trend of the speech local segment and the change trend of the track vehicle speed is calculated based on the Pearson correlation coefficients of all speech local segments in the cluster where the speech local segment belongs and the corresponding speed local segments.

5. The method for interactive control of a vehicle-mounted display screen for a rail vehicle according to claim 3, characterized in that: The possibility that the speech local segment belongs to the burst noise is calculated based on the difference between the first speech local segment and the second speech local segment corresponding to the speech local segment; The track vehicle speed change corresponding to the speech partial segment is the difference between the approximate speed value corresponding to the speech partial segment and the approximate speed value corresponding to the previous speech partial segment of its component.

6. The method for interactive control of a vehicle-mounted display screen for a rail vehicle according to claim 5, characterized in that: For the d-th speech local segment of a component, its corresponding first speech local segment is the speech local segment obtained by reconstructing the d-th speech local segment of each of the remaining components, and its corresponding second speech local segment is the d-th speech local segment in the original speech data, where d is an integer greater than or equal to 1 and less than or equal to m, and m is the total number of speech local segments in a component.

7. The method for interactive control of an on-board display screen for a rail vehicle according to any one of claims 1 to 6, characterized in that: Before calculating the weight of the speech local segment, the method further includes: For each speech partial segment, the distribution characteristics of the speech partial segment are calculated according to the distribution disorder degree of all data points in the speech partial segment and the frequency resolution of the speech partial segment in the original speech data change curve.

8. The method for interactive control of an on-board display screen for a rail vehicle according to any one of claims 1 to 6, characterized in that: The preprocessing of the original voice data to obtain multiple components of the original voice data includes: Performing EMD decomposition on the original speech data to obtain a plurality of intrinsic mode function (IMF) components of the original speech data.

9. The method for interactive control of a vehicle-mounted display screen for a rail vehicle according to claim 2, characterized in that: The method of segmenting the original voice data and the multiple components according to the track speed time sequence corresponding to the original voice data using a preset segmentation method to obtain multiple voice local segments of the original voice data and multiple voice local segments of each component includes: Acquire a track speed time sequence sequence corresponding to the original voice data, wherein the original voice data and the track speed time sequence sequence corresponding to the original voice data have the same starting time and the same ending time; The track speed time series corresponding to the original voice data is segmented using an adaptive piecewise constant approximation APCA segmentation method to obtain a plurality of speed local segments and an approximate speed value corresponding to each speed local segment; According to the starting time and the ending time of each speed partial segment, the original speech data and the multiple components are segmented to obtain multiple speech partial segments of the original speech data and multiple speech partial segments of each component.

10. An on-board display screen interactive control system for rail vehicles, characterized in that: The system comprises: A processing module, used for collecting original voice data and preprocessing the original voice data to obtain multiple components of the original voice data; A segmentation module, used for segmenting the original voice data and multiple components according to the track speed time sequence corresponding to the original voice data, to obtain multiple voice local segments of the original voice data and multiple voice local segments of each component; A calculation module, configured to calculate, for each speech partial segment of each component, a weight of the speech partial segment according to distribution characteristics of the plurality of speech partial segments, a track speed time series, and a difference in distribution characteristics of the speech partial segment in the time domain and the frequency domain; A reconstruction module, used for weightedly reconstructing all components of the original speech data according to the weight corresponding to each speech local segment of each component, to obtain reconstructed speech data; A response module is used to determine the user intention based on the reconstructed voice data, respond to the user intention and output the response result on the vehicle display screen.

Images