A Big Data-Based Method for Voice Line Fault Analysis

Abstract

This invention discloses a voice line fault analysis method based on big data, belonging to the field of line fault analysis technology. The invention includes the following steps: synchronously acquiring raw voice signals and network jitter data from the voice line; preprocessing the voice and network data; extracting amplitude fluctuation features from the processed voice signal; converting voice fluctuations and network jitter into frequency domain signals and calculating the correlation strength between the two within a key frequency range; comparing the calculated fault index with an adaptive dynamic threshold to determine whether a fault has occurred; analyzing the key frequency components leading to the fault and, combined with network topology information, determining the specific physical location of the fault; and generating a structured report containing fault details and location information. This invention, by analyzing the coherence of the voice signal modulation spectrum and network jitter sequence within a key frequency range and constructing a dynamic fault index, can achieve the identification and location of voice line faults.

Description

Technical Field

[0001] This invention belongs to the field of line fault analysis technology, and in particular relates to a voice line fault analysis method based on big data. Background Technology

[0002] In the field of real-time voice communication, especially in demanding applications such as live interactive classrooms on online education platforms, the transmission quality of voice lines directly impacts user experience and teaching effectiveness. These scenarios place extremely high demands on the continuity and clarity of voice communication; even the slightest jitter can cause voice interruptions or stuttering, thereby interfering with the teaching process.

[0003] In existing technologies, fault monitoring of voice lines largely relies on monitoring independent parameter thresholds. A common method is to continuously collect network-level metrics, such as network jitter and packet loss rate; when these metrics exceed preset thresholds, a line anomaly is determined. Another approach is to analyze the quality of the received voice signal, such as detecting silence segments or changes in signal-to-noise ratio. However, both approaches have inherent limitations. They treat network status and voice content as two isolated systems, ignoring the complex interactions between them. For example, the inherent fluctuations in the voice signal itself (such as pauses and emphasis in speech) cause dynamic changes in its amplitude envelope. These changes, when there are slight network fluctuations, can easily be misinterpreted as voice quality degradation caused by network faults. Conversely, when the network experiences jitter at a specific frequency that coincides with the periodic changes in voice content, the system may fail to issue timely alerts due to a lack of correlation analysis.

[0004] The fundamental problem with this isolated analysis paradigm is its inability to effectively distinguish between apparent fluctuations caused by changes in the voice content itself and real faults caused by network issues. This leads to numerous false alarms in the operations and maintenance system, mistaking normal voice fluctuations for faults, or missed alarms, failing to identify hidden network problems coupled with voice rhythm. As a result, operations and maintenance personnel need to spend a significant amount of time manually identifying these issues, hindering the rapid location of the true fault and resulting in low troubleshooting efficiency, making it difficult to meet the real-time system's operational response requirements. Therefore, the following solutions are proposed to address these problems. Summary of the Invention

[0005] The purpose of this invention is to provide a voice line fault analysis method based on big data. By analyzing the coherence of the voice signal modulation spectrum and network jitter sequence within the key frequency range and constructing a dynamic fault index, it is possible to identify and locate voice line faults. This solves the problem of misjudgment and missed judgment caused by the isolated analysis of voice content changes and network transmission problems in the prior art.

[0006] To solve the above-mentioned technical problems, the present invention is achieved through the following technical solution: This invention relates to a voice line fault analysis method based on big data, comprising the following steps: Data acquisition: Simultaneously acquire voice signals and network jitter sequences from the target voice line; Data preprocessing: The speech signal and network jitter sequence are cleaned and standardized to obtain time-aligned speech amplitude envelope sequence and network jitter sequence; Coherence analysis: Calculate the coherence function of the speech amplitude envelope sequence and the network jitter sequence in the frequency domain, and calculate a fault index based on the value of the coherence function within a preset key frequency range; Fault detection: The fault index is compared with a dynamic threshold. If the fault index exceeds the dynamic threshold, the voice line is determined to be faulty.

[0007] Furthermore, the preprocessing of the speech signal in the data preprocessing step includes: The speech signal is subjected to framing, windowing, and filtering. Extract the amplitude envelope of each frame of the speech signal; The amplitude envelope sequence is resampled so that its sampling rate is consistent with that of the network jitter sequence.

[0008] Furthermore, the coherence analysis step specifically includes: Perform a Fourier transform on the speech amplitude envelope sequence to obtain the speech modulation spectrum; Perform a Fourier transform on the network jitter sequence to obtain the network jitter spectrum; Based on the speech modulation spectrum and the network jitter spectrum, calculate the coherence function between them.

[0009] Furthermore, the calculation of the fault index based on the value of the coherence function within a preset key frequency range is achieved in the following way: Calculate the average value of the coherence function within the critical frequency range; define the difference between 1 and the average value as the fault index.

[0010] Furthermore, the preset key frequency range is 4Hz to 8Hz.

[0011] Furthermore, the dynamic threshold is adaptively adjusted based on the statistical characteristics of the historical fault index.

[0012] Furthermore, after determining that there is a fault in the voice line, a fault location step is also included: Analyze the distribution of the coherence function within the critical frequency range to identify abnormal frequency points that cause an increase in the fault index; By combining network topology information, the abnormal frequency points are mapped to physical network nodes, thereby achieving fault location.

[0013] The present invention has the following beneficial effects: 1. This invention identifies voice line faults by analyzing the intrinsic coherence between the modulation spectrum of the voice signal and the network jitter sequence. By calculating the coherence function of key frequency bands, it can effectively distinguish between real faults caused by network infrastructure and apparent fluctuations caused by changes in the voice content itself. This correlation analysis improves the reliability of fault detection, avoids false alarms and missed alarms, and enables maintenance personnel to more accurately judge the true state of the system.

[0014] 2. By continuously monitoring subtle changes in the coherence of the modulation spectrum and jitter sequence, the system can detect abnormal trends in line status before significant degradation of voice quality. This detection mechanism provides a time window for preventive maintenance, allowing the maintenance team to proactively intervene before the impact of the fault expands, reducing the risk of business interruption and improving the continuity of system services.

[0015] 3. By analyzing the abnormal distribution of coherence functions on specific frequency components and combining it with network topology information, this invention can map abstract fault indicators to specific physical devices or link segments. This localization capability shortens the time required for fault diagnosis, allowing maintenance personnel to quickly locate the cause and take remedial measures directly for the problematic nodes, thereby improving troubleshooting efficiency and reducing operation and maintenance costs.

[0016] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating a voice line fault analysis method based on big data according to the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see Figure 1 As shown, this invention is a voice line fault analysis method based on big data, comprising the following steps: Step S1, Data Acquisition Voice and network data are simultaneously collected from the real-time voice live stream of the online education platform. The voice data is acquired in the form of an audio stream with a sampling rate of 16kHz, a frame length of 20 milliseconds, and an overlap rate of 50%. The network data includes a network jitter sequence, with each sampling point being every 10 milliseconds. The jitter value is defined as the variance of the arrival time of data packets. Data acquisition is achieved through the platform's built-in sensors and network monitoring tools to ensure the real-time performance and integrity of the data. The collected data is stored in a distributed database for subsequent processing.

[0021] Step S2: Data Preprocessing The collected voice and network data are cleaned and standardized to remove noise and outliers. The specific steps are as follows: Speech data preprocessing: A Hamming window function is applied to each frame of speech signal to reduce spectral leakage; then a high-pass filter (cutoff frequency 80 Hz) is performed to remove DC offset and low-frequency noise; the amplitude of the filtered speech signal is normalized to the range of [-1,1] to ensure consistency; Network data preprocessing: The network jitter sequence is smoothed by using a moving average filter (window size of 5 sampling points) to reduce random fluctuations; at the same time, outliers (such as data points with jitter values ​​exceeding 100 milliseconds) are checked and removed, and missing values ​​are filled by linear interpolation to ensure sequence continuity. The preprocessed data output consists of a time-aligned audio frame sequence and a network jitter sequence.

[0022] Step S3: Feature Extraction Key features are extracted from the preprocessed data for subsequent analysis; Speech feature extraction: Calculate the amplitude envelope of each frame of speech signal; the amplitude envelope is obtained through Hilbert transform, and is represented as... ,in, The time index is used; then, the amplitude envelope sequence is resampled to match the network jitter sequence (i.e., one point every 10 milliseconds). Network feature extraction: directly using the preprocessed network jitter sequence ,in, In time Jitter value (unit: milliseconds); Extracted feature sequences and This is used as input for step S4.

[0023] Step S4: Modulation spectrum and coherence analysis This step defines the fault index by calculating the modulation spectrum of the speech amplitude envelope and the coherence function of the network jitter sequence. The specific steps are as follows: Modulation spectrum calculation: for speech amplitude envelope sequence Perform a Fourier transform to obtain the modulation spectrum. ,in, Modulation frequency (unit: Hz); Fourier transform uses Fast Fourier Transform (FFT) algorithm, frequency resolution is 0.5Hz; Modulation spectrum Represented as: ; In the formula, For the modulation spectrum, For sequence length, For time indexing, It is a natural constant. The imaginary unit, Pi The sampling rate is 100Hz, corresponding to the amplitude envelope after resampling. Network jitter spectrum calculation: Calculate the network jitter sequence Perform the same Fourier transform to obtain the jitter spectrum. : ; Coherence function calculation: Calculating the modulation spectrum and jitter spectrum coherence function The coherence function measures the correlation between two signals in the frequency domain and is defined as: ; In the formula, for and Cross power spectral density, and They are respectively and The power spectral density was calculated; the power spectral density was estimated using the Welch method with a window length of 256 points and an overlap rate of 50%; coherence function. The value ranges from [0,1], and the higher the value, the stronger the correlation. Fault Index Calculation: Within the critical modulation frequency range (4-8Hz, corresponding to the syllable rate of human speech), calculate the average coherence value and derive the fault index. Failure Index Defined as: ; In the formula, These are discrete frequency points within the critical frequency range (with a step size of 0.5 Hz). The total number of frequency points. For frequency index, For the first A discrete frequency value, For the coherence function at frequency The value at; A value close to 0 indicates normal operation, while a value close to 1 indicates a malfunction.

[0024] Step S5, Fault Detection Based on the failure index Perform fault detection; set dynamic thresholds This threshold is adaptively adjusted based on historical data: In the formula, and Fault index over the past 24 hours The mean and standard deviation; when When the fault occurs, a fault alarm is triggered; at the same time, the timestamp of the fault occurrence and the corresponding voice frame position are recorded for subsequent location.

[0025] Step S6, Fault Location Accurately locate the fault point that triggers the alarm; analyze the coherence function. Distribution within the critical frequency range to identify the specific frequency components causing the fault; when At a specific frequency If there is a significant drop (below 0.3) at a frequency (e.g., 5Hz), then that frequency point is marked as an anomaly. By combining network topology data, the anomaly point is mapped to a physical line node (e.g., a router or switch) to achieve fault location.

[0026] Step S7, Result Output Generate a fault report, including a fault index. The report includes the trigger time, location information, and recommended measures; it is output in a structured format (such as JSON) and integrated into the platform monitoring system to achieve real-time alerts and visualization; at the same time, the data is stored in a historical database for trend analysis and system optimization.

[0027] Working principle: This embodiment identifies voice line faults by comprehensively analyzing voice signals and network jitter data and detecting changes in their coherence within a key frequency range. Specifically, it first synchronously collects voice audio and network jitter data from the real-time voice live streaming line of an online education platform and preprocesses them to eliminate noise and outliers. Then, it extracts the amplitude envelope of the voice signal and the network jitter sequence as features, converts the time-domain signal to a frequency-domain representation using Fourier transform, and calculates the modulation spectrum and jitter spectrum. It introduces coherence function analysis to quantify the correlation between the voice modulation spectrum and network jitter at specific frequencies (e.g., 4-8Hz, corresponding to the syllable rate of human speech) and derives a fault index. This index reflects the degree of abnormality in the line status; a higher value indicates a greater risk of fault. During the analysis, the system automatically triggers fault alarms by comparing the fault index with a dynamic threshold. If the index exceeds the threshold, it indicates that the voice clarity may be impaired due to network jitter. The system will then further locate the fault source by identifying abnormally decreasing frequency points in the coherence function and mapping them to physical devices (such as routers or switches) in conjunction with network topology data. The entire process relies entirely on big data algorithms for processing, requiring no manual intervention, ensuring real-time performance and accuracy. Finally, a structured fault report is output and integrated into the platform monitoring system for rapid response and optimization.

[0028] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0029] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A voice line fault analysis method based on big data, characterized in that, The analytical method includes the following steps: Data acquisition: Simultaneously acquire voice signals and network jitter sequences from the target voice line; Data preprocessing: The speech signal and network jitter sequence are cleaned and standardized to obtain time-aligned speech amplitude envelope sequence and network jitter sequence; Coherence analysis: Calculate the coherence function of the speech amplitude envelope sequence and the network jitter sequence in the frequency domain, and calculate a fault index based on the value of the coherence function within a preset key frequency range; Fault detection: The fault index is compared with a dynamic threshold. When the fault index exceeds the dynamic threshold, it is determined that the voice line has a fault.

2. The method for analyzing voice line faults based on big data according to claim 1, characterized in that, The data preprocessing step includes the preprocessing of the speech signal, which includes: The speech signal is subjected to framing, windowing, and filtering. Extract the amplitude envelope of each frame of the speech signal; The amplitude envelope sequence is resampled so that its sampling rate is consistent with that of the network jitter sequence.

3. The voice line fault analysis method based on big data according to claim 1, characterized in that, The coherence analysis steps specifically include: Perform a Fourier transform on the speech amplitude envelope sequence to obtain the speech modulation spectrum; Perform a Fourier transform on the network jitter sequence to obtain the network jitter spectrum; Based on the speech modulation spectrum and the network jitter spectrum, calculate the coherence function between them.

4. The voice line fault analysis method based on big data according to claim 1, characterized in that, The calculation of the fault index based on the value of the coherence function within a preset critical frequency range is achieved through the following method: Calculate the average value of the coherence function within the critical frequency range; define the difference between 1 and the average value as the fault index.

5. The voice line fault analysis method based on big data according to claim 4, characterized in that, The preset key frequency range is 4Hz to 8Hz.

6. The voice line fault analysis method based on big data according to claim 1, characterized in that, The dynamic threshold is adaptively adjusted based on the statistical characteristics of historical failure indices.

7. The method for analyzing voice line faults based on big data according to claim 1, characterized in that, After determining that there is a fault in the voice line, the method further includes a fault location step: Analyze the distribution of the coherence function within the critical frequency range to identify abnormal frequency points that cause an increase in the fault index; By combining network topology information, the abnormal frequency points are mapped to physical network nodes, thereby achieving fault location.

Images