A method for determining the correctness of a mobile communication voice decoding

Abstract

The present application belongs to the technical field of mobile communication voice processing, and particularly relates to a mobile communication voice decoding correctness discrimination method, comprising the following steps: S1, constructing a fixed time length sliding window and a loop buffer LoopBuf; S2, adopting an incremental mean optimization mechanism; S3, based on the natural distribution characteristics of the silence frames of normal conversation voice; S4, traversing a plurality of groups of preset decoding parameters, respectively calculating the voice amplitude mean under the sliding window corresponding to each group of parameters, extracting the minimum value and the second minimum value from all amplitude means and calculating the difference value, when the difference value is greater than a preset difference threshold, determining that the decoding parameter corresponding to the minimum value is the optimal decoding parameter, and simultaneously confirming that the decoding result under the parameter is correct, the present application relies on the natural distribution characteristics of the silence frames of normal conversation voice through discrimination logic, and can complete decoding effectiveness discrimination only through the amplitude statistics of the decoded voice frames, without increasing the transmission layer overhead, and can still work stably in a weak network and packet loss scenario.

Description

Technical Field

[0001] This invention belongs to the field of mobile communication voice processing technology, and specifically relates to a method for judging the correctness of mobile communication voice decoding. Background Technology

[0002] In mobile voice communication services, the correctness of voice frame decoding directly determines call quality and user experience. Current technologies typically rely on transport layer checksums, network-side quality reporting signaling, or complete voice semantic recognition to assess decoding performance. However, these methods have the following drawbacks in practical applications: Relying on additional signaling or verification fields increases transmission overhead, and verification information is prone to failure in weak network or packet loss scenarios, making it impossible to complete the judgment solely based on the natural characteristics of the voice itself; embedded mobile communication terminals have limited computing power and memory resources, and the full frame-by-frame decoding and comparison scheme has too much computational load, making it difficult to run in real time on low-power terminals; voice frame synchronization offset is prone to decoding errors, and existing solutions cannot accurately locate the starting synchronization position of voice frames with low computational overhead, and cannot simultaneously complete synchronization calibration and decoding correctness judgment.

[0003] In existing technologies, silence frame detection technology is mostly used in speech activity detection (VAD) and speech endpoint detection scenarios. There is no solution that combines the natural proportion feature of silence frames with sliding window loop caching and incremental calculation for rapid judgment of the correctness of mobile communication voice decoding and optimal parameter matching. Summary of the Invention

[0004] The purpose of this invention is to provide a method for judging the correctness of mobile communication voice decoding. Based on the natural distribution characteristics of voice silence frames, combined with a sliding window loop caching mechanism and an incremental mean optimization algorithm, it can quickly judge the correctness of decoding without the need for additional signaling, while reducing memory and computing power overhead, adapting to embedded terminals, and simultaneously completing the accurate matching of voice frame synchronization positions.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a method for judging the correctness of mobile communication voice decoding, comprising the following steps: S1. Construct a fixed-duration sliding window and a loop buffer (LoopBuf). The loop buffer stores historical voice frame data in a loop. The sliding window traverses the candidate synchronization points of the voice frames in units of step frames. S2. An incremental mean optimization mechanism is adopted. Each time the sliding window slides, the decoding operation is only performed on the newly added speech frames. The decoded calculation results of the historical speech frames in the window are reused to calculate the amplitude mean of all speech frames in the current window. S3. Based on the natural distribution characteristics of silent frames in normal call speech, the proportion of silent frames in the current sliding window is counted, and the proportion is compared with a preset discrimination threshold to preliminarily determine the decoding effectiveness under the current decoding parameters. S4. Iterate through multiple preset sets of decoding parameters, calculate the average speech amplitude under the sliding window corresponding to each set of parameters, extract the minimum and second minimum values ​​among all amplitude averages and calculate the difference between them. When the difference is greater than the preset difference threshold, determine that the decoding parameter corresponding to the minimum value is the optimal decoding parameter, and confirm that the decoding result under this parameter is correct.

[0006] As a mobile communication voice decoding correctness judgment method of the present invention, preferably, in step S1, the loop buffer LoopBuf adopts a circular queue structure, the buffer capacity is matched with the duration of the sliding window, and when a new voice frame is written to the buffer, it overwrites the earliest stored historical voice frame, so as to realize the circular storage with fixed memory occupation.

[0007] As a mobile communication voice decoding correctness judgment method of the present invention, preferably, the step frame length of the sliding window in S1 is the duration of a single frame of voice. When the sliding window traverses the candidate synchronization points, each slide corresponds to a voice frame offset, and the starting synchronization position of the voice frame is matched one by one.

[0008] As a mobile communication voice decoding correctness judgment method of the present invention, preferably, the incremental mean optimization mechanism in S2 is as follows: the decoding status and amplitude calculation result of each frame of voice in the window are marked in advance. When the sliding window is updated, the historical frame data of the moved-out window is removed, and only the newly moved-in voice frame is decoded and the amplitude calculation is performed. The overall amplitude mean of the current window is obtained by weighting the calculation results of the remaining historical frames.

[0009] As a mobile communication voice decoding correctness judgment method of the present invention, preferably, the determination of the silence frame in S3 is based on the amplitude value of a single frame of voice being lower than a preset silence amplitude threshold; the preset threshold for the proportion of silence frames in a normal call scenario is 30%~60%, and when the proportion of silence frames in the decoded window is lower than the preset threshold, it is determined that the decoding has failed under the current decoding parameters.

[0010] As a mobile communication voice decoding correctness judgment method of the present invention, preferably, the multiple sets of decoding parameters in S4 include at least two combinations of different encoding rates, synchronization offsets, and decoding gain coefficients, and each set of parameters corresponds to an independent decoding process and amplitude mean statistical results.

[0011] As a method for judging the correctness of mobile communication voice decoding according to the present invention, preferably, the preset difference threshold in S4 is 10% to 30% of the minimum value; when the difference between the minimum value and the second smallest value is less than or equal to the preset difference threshold, the range of decoding parameter traversal is expanded, and parameter traversal and amplitude mean calculation are re-executed.

[0012] As a method for judging the correctness of mobile communication voice decoding according to the present invention, preferably, the method is adapted to AMR, EVS and FR voice coding formats in mobile communication, and the decoding operation is directly performed based on the encoded voice frames received by the physical layer of mobile communication.

[0013] As a mobile communication voice decoding correctness judgment method of the present invention, preferably, the method is deployed in an embedded mobile communication terminal, and the operation of loopback buffer and sliding window is performed on the local MCU / DSP of the terminal, without relying on additional signaling interaction on the network side.

[0014] As a method for judging the correctness of mobile communication voice decoding according to the present invention, preferably, after the sliding window traverses all candidate synchronization points, the synchronization position with the closest proportion of silent frames to the normal call interval is selected as the precise synchronization start point of the voice frame, and the frame synchronization calibration is completed.

[0015] Compared with the prior art, the beneficial effects of the present invention are: This invention relies on the natural distribution characteristics of silence frames in normal voice calls to determine the validity of decoding simply by statistically analyzing the amplitude of the decoded voice frames. It does not increase transmission layer overhead and can still work stably in weak network and packet loss scenarios.

[0016] This invention employs an optimization mechanism combining loopback caching and incremental mean calculation. When updating the sliding window, only newly added voice frames are decoded, and the calculation results of historical frames are reused, which greatly reduces the amount of repetitive calculations. At the same time, the loopback caching adopts a ring structure with fixed memory usage, which significantly reduces memory and computing power overhead, making it suitable for embedded mobile communication terminals with limited computing power and memory.

[0017] This invention, through a sliding window that iterates through candidate synchronization points frame by frame, can simultaneously match the optimal starting synchronization position of the speech frame. While completing the decoding correctness judgment, it also achieves frame synchronization calibration, thereby improving the synchronization accuracy and stability of speech decoding.

[0018] This invention determines the optimal parameter by the difference between the minimum and second minimum values ​​of the mean amplitude under multiple sets of decoding parameters. It can automatically match the best decoding configuration, avoid the problem of poor adaptability of a single parameter, and improve the decoding success rate in complex channel environments. Attached Figure Description

[0019] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart illustrating the steps involved in determining the correctness of mobile communication voice decoding. Detailed Implementation

[0020] 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.

[0021] Please see Figure 1 The present invention provides the following technical solution: a method for judging the correctness of mobile communication voice decoding, applied to terminal-side voice decoding verification in VoLTE mobile communication scenarios, the specific steps of which are as follows: Step 1: Cache and Window Initialization The sliding window duration is configured to be 200ms, corresponding to 10 standard 20ms audio frames; a loop buffer LoopBuf matching the window capacity is constructed, using a circular queue structure with a buffer depth of 10 frames. Data writing and reading are controlled by write pointers and read pointers respectively. When a new frame is written, the oldest historical frame data is automatically overwritten. The memory usage is fixed throughout the process, with no dynamic memory allocation.

[0022] Step 2: Sliding window traversal and incremental decoding The sliding window traverses candidate synchronization points with a single frame as the step size. Each time it slides, one historical speech frame is moved out of the window and one new speech frame is moved into the window. Decoding operations are only performed on the newly moved speech frame and its amplitude value is calculated. The amplitude results of the remaining 9 historical speech frames in the window are directly reused. The overall average speech amplitude of the current window is obtained by weighted summation and divided by the total number of frames.

[0023] Step 3: Preliminary validity assessment based on the proportion of silent frames The preset silence amplitude threshold is -50dBm. A single frame of speech with an amplitude below this threshold is judged as a silent frame. In normal call scenarios, there are naturally many silent segments during the pauses in human speech, and the proportion of silent frames is usually stable at around 40%. Therefore, the discrimination threshold is set to 35%. When the proportion of silent frames in the window is ≥35%, it is initially determined that the decoding is effective under the current decoding parameters. If the proportion is <35%, it means that the decoding output is irregular noise, which does not conform to the characteristics of real speech, and the current decoding parameters are judged to be incorrect and the decoding is invalid.

[0024] Step 4: Multi-parameter traversal and determination of optimal decoding parameters Five sets of decoding parameters with different synchronization offsets are preset, corresponding to synchronization offsets of 0 to 4 bits respectively. The sliding window decoding and amplitude mean calculation are performed sequentially using each set of parameters to obtain the five corresponding window amplitude mean values. The minimum and second minimum values ​​are extracted from the five mean values, and their difference is calculated. A difference threshold of 20% of the minimum value is set. If the difference is greater than this threshold, it indicates that the parameter corresponding to the minimum value best matches the speech energy distribution of the real speech characteristics after decoding, and this parameter is determined to be the optimal decoding parameter, and the corresponding decoding result is correct. If the difference is less than or equal to the threshold, it indicates that the parameter's discrimination is insufficient, and three new sets of offset parameters are added and the process is repeated until the difference meets the threshold requirement.

[0025] Step 5: Synchronize Position Calibration After traversing all candidate synchronization offsets through the sliding window, the offset position with the closest proportion of silent frames to the normal range of 40% is selected as the precise synchronization start point of the speech frame, and the frame synchronization calibration is completed. Subsequent speech decoding is performed based on this calibration position.

[0026] In this embodiment, the method can be deployed on a mobile communication terminal with a DSP chip. All operations are completed locally on the terminal, without the need for interaction with the network side to verify signaling. The single-window discrimination operation takes less than 1ms and the memory usage is less than 2KB, making it suitable for low-power embedded devices.

[0027] The overall process of this method revolves around three core logics: natural feature discrimination, low-overhead computation, and integrated optimization of synchronization parameters. This achieves signaling-free, low-resource, and highly accurate decoding correctness verification. Normal human conversations naturally involve pauses, resulting in a consistent proportion of silent frames in the speech data. This proportion remains stable between 30% and 60% in normal conversations, representing an inherent statistical property of speech itself, requiring no additional verification information. When decoding parameters are incorrect or decoding fails, the output is an irregular noise signal with uniform energy distribution and almost no silent frames meeting the amplitude threshold. The proportion of silent frames is significantly lower than in the normal range. This method utilizes this difference, using the proportion of silent frames as the core criterion for decoding correctness, achieving effectiveness verification with zero additional transmission overhead. To adapt to the limited computing power and memory of embedded terminals, this method reduces resource consumption through two levels of optimization: LoopBuf: It uses a circular queue to store historical voice frames in a fixed manner. New data overwrites old data, and the memory usage remains constant throughout the process, avoiding the problem of dynamic memory growth in streaming processing. Incremental mean calculation: The operation mode of full re-decoding of sliding window is optimized to "decoding only new frames + reuse of historical frame results". Each window slide only needs to complete the decoding operation of 1 frame. The amount of operation is reduced from O(N) (N is the total number of frames in the window) to O(1), which greatly reduces repeated calculations and meets the real-time requirements. The process of traversing the sliding window frame by frame is essentially to verify the decoding effect under different synchronization start positions and different decoding parameters, thus solving the decoding error problem caused by the synchronization offset of the speech frame. In the comparison of multiple sets of parameters, the characteristics that "when decoding correctly, the speech energy conforms to the true distribution and the overall amplitude mean is lower, while when decoding incorrectly, the noise energy is higher and the results of different incorrect parameters are small" are utilized. The difference between the minimum and the second minimum amplitude mean is used as a confidence index: the larger the difference, the more outstanding the decoding effect of the parameter corresponding to the minimum value, and the higher the confidence of the result; when the difference is insufficient, the parameter range is expanded to ensure that the optimal decoding parameters are finally selected, and the decoding correctness judgment and frame synchronization calibration are completed simultaneously.

[0028] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for judging the correctness of mobile communication voice decoding, characterized in that, Includes the following steps: S1. Construct a fixed-duration sliding window and a loop buffer (LoopBuf). The loop buffer stores historical voice frame data in a loop. The sliding window traverses the candidate synchronization points of the voice frames in units of step frames. S2. An incremental mean optimization mechanism is adopted. Each time the sliding window slides, the decoding operation is only performed on the newly added speech frames. The decoded calculation results of the historical speech frames in the window are reused to calculate the amplitude mean of all speech frames in the current window. S3. Based on the natural distribution characteristics of silent frames in normal call speech, the proportion of silent frames in the current sliding window is counted, and the proportion is compared with a preset discrimination threshold to preliminarily determine the decoding effectiveness under the current decoding parameters. S4. Iterate through multiple preset sets of decoding parameters, calculate the average speech amplitude under the sliding window corresponding to each set of parameters, extract the minimum and second minimum values ​​among all amplitude averages and calculate the difference between them. When the difference is greater than the preset difference threshold, determine that the decoding parameter corresponding to the minimum value is the optimal decoding parameter, and confirm that the decoding result under this parameter is correct.

2. The method for judging the correctness of mobile communication voice decoding according to claim 1, characterized in that: In step S1, the loop buffer LoopBuf adopts a circular queue structure, and the buffer capacity matches the duration of the sliding window. When a new voice frame is written to the buffer, it overwrites the earliest stored historical voice frame, thus achieving circular storage with fixed memory usage.

3. The method for judging the correctness of mobile communication voice decoding according to claim 1, characterized in that: In S1, the step frame length of the sliding window is the duration of a single audio frame. When the sliding window traverses the candidate synchronization points, each slide corresponds to an audio frame offset, and the starting synchronization position of the audio frames is matched one by one.

4. The method for judging the correctness of mobile communication voice decoding according to claim 1, characterized in that: The incremental mean optimization mechanism in S2 is as follows: the decoding status and amplitude calculation results of each frame of speech within the window are marked in advance. When the sliding window is updated, the historical frame data that is moved out of the window is removed, and only the newly moved speech frame is decoded and the amplitude is calculated. The overall amplitude mean of the current window is obtained by weighting the calculation results of the remaining historical frames.

5. The method for judging the correctness of mobile communication voice decoding according to claim 1, characterized in that: The determination of silent frames in S3 is based on the amplitude value of a single frame of speech being lower than a preset silent amplitude threshold. In a normal call scenario, the preset threshold for the proportion of silent frames is 30% to 60%. When the proportion of silent frames in the decoded window is lower than the preset threshold, the decoding is determined to be invalid under the current decoding parameters.

6. The method for judging the correctness of mobile communication voice decoding according to claim 1, characterized in that: The S4 contains multiple sets of decoding parameters, including at least two combinations of different encoding rates, synchronization offsets, and decoding gain coefficients. Each set of parameters corresponds to an independent decoding process and amplitude mean statistical results.

7. The method for judging the correctness of mobile communication voice decoding according to claim 1, characterized in that: In S4, the preset difference threshold is 10% to 30% of the minimum value; when the difference between the minimum value and the second smallest value is less than or equal to the preset difference threshold, the range of decoding parameter traversal is expanded, and parameter traversal and amplitude mean calculation are re-executed.

8. The method for judging the correctness of mobile communication voice decoding according to claim 1, characterized in that: The method is adapted to AMR, EVS, and FR voice coding formats in mobile communications, and the decoding operation is performed directly based on the encoded voice frames received by the physical layer of mobile communications.

9. The method for judging the correctness of mobile communication voice decoding according to claim 1, characterized in that: The method is deployed on an embedded mobile communication terminal. The operations of loopback buffering and sliding window are performed on the local MCU / DSP of the terminal, without relying on additional signaling interaction on the network side.

10. The method for judging the correctness of mobile communication voice decoding according to claim 3, characterized in that: After the sliding window traverses all candidate synchronization points, the synchronization position with the closest proportion of silent frames to the normal call interval is selected as the precise synchronization start point of the voice frame, thus completing the frame synchronization calibration.

Images