HK40022645B - Method and device for processing voice, computer readable storage medium and computer apparatus - Google Patents

Description

Speech processing methods, apparatus, computer-readable storage media, and computer devices

Technical Field

This application relates to the field of speech processing technology, and in particular to a speech processing method, apparatus, computer-readable storage medium, and computer device.

Background Technology

During a voice call, voice can be transmitted from the sender to the receiver over the network. Due to network quality issues, voice data packets may be lost during transmission, causing the receiver to experience stuttering and discontinuity, thus affecting the quality of the call.

In traditional packet loss mitigation schemes, voice data packets are encoded using Forward Error Correction (FEC) to obtain redundant packets. These redundant packets are then sent together to the receiver. If packet loss occurs, the receiver can recover the complete voice at the point of loss based on the redundant packets, thus achieving packet loss mitigation. A higher FEC redundancy (the ratio of redundant packets to voice data packets) provides stronger packet loss mitigation, but it consumes significantly more bandwidth. Conversely, a lower FEC redundancy will not achieve the desired error correction.

Summary of the Invention

Therefore, it is necessary to provide a voice processing method, apparatus, computer-readable storage medium, and computer device to address the technical problem of how to ensure effective error correction of voice data packets while consuming low bandwidth.

A speech processing method, comprising:

The acquired speech is subjected to speech rate detection to obtain the speech rate value;

Obtain the forward error correction redundancy;

The forward error correction redundancy is adjusted based on the speech rate value to obtain the target redundancy.

The speech is encoded to obtain a speech encoding packet;

The speech coding packet is forward-corrected according to the target redundancy to obtain a redundant packet;

The redundant packet and the voice encoding packet are sent to the receiving end.

A voice processing device, the device comprising:

The detection module is used to detect the speech rate of the acquired speech and obtain the speech rate value;

The acquisition module is used to obtain the forward error correction redundancy.

An adjustment module is used to adjust the forward error correction redundancy based on the speech rate value to obtain the target redundancy.

The first encoding module is used to encode the speech to obtain a speech encoding packet;

The second encoding module is used to perform forward error correction encoding on the speech encoding packet according to the target redundancy to obtain a redundant packet;

The sending module is used to send the redundant packet and the voice encoding packet to the receiving end.

A computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to perform the steps of the speech processing method.

A computer device includes a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the steps of the speech processing method.

The aforementioned speech processing method, apparatus, computer-readable storage medium, and computer device, by detecting speech rate and adjusting the forward error correction redundancy using the detected speech rate value, can use the adjusted target redundancy to perform forward error correction encoding on speech coding packets, thereby obtaining redundant packets. When the speech rate is slow, the speech data packet contains less speech content, while when the speech rate is fast, the speech data packet contains more speech content. Dynamically adjusting the forward error correction redundancy according to the speech rate value can ensure that lost speech data packets can be effectively recovered, thereby achieving effective error correction of speech data packets and avoiding the consumption of a large amount of bandwidth.

Attached Figure Description

Figure 1 is an application environment diagram of a speech processing method in one embodiment;

Figure 2 is a flowchart illustrating a speech processing method in one embodiment;

Figure 3 is a schematic diagram of speech framing in one embodiment;

Figure 4 is a schematic diagram of forward error correction coding performed at the transmitting end in one embodiment;

Figure 5 is a flowchart illustrating the steps for calculating speech rate in one embodiment;

Figure 6 is a flowchart illustrating the forward error correction redundancy adjustment process in one embodiment, which involves comparing the obtained target redundancy with the upper and lower limits of redundancy, and performing forward error correction coding based on the comparison results.

Figure 7 is a schematic diagram of the process in one embodiment where the transmitting end adjusts the forward error correction redundancy and performs forward error correction coding, and the receiving end performs forward error correction decoding to recover the voice coding packet.

Figure 8 is a structural block diagram of a speech processing device in one embodiment;

Figure 9 is a structural block diagram of the voice processing device in another embodiment;

Figure 10 is a structural block diagram of a computer device in one embodiment.

Detailed Implementation

To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

Figure 1 illustrates an application environment of the speech processing method in one embodiment. Referring to Figure 1, the speech processing method is applied to a speech processing system. This system includes a terminal 110, a transmission node 120, and a terminal 130. Terminal 110, transmission node 120, and terminal 130 are connected via a network. Terminal 110 can function as a transmitter (or receiver), specifically a desktop terminal or a mobile terminal, where the mobile terminal can be at least one of a mobile phone, tablet computer, or laptop computer. Correspondingly, terminal 130 can function as a receiver (or transmitter). When terminal 110 is the transmitter, terminal 130 can be the receiver, specifically a desktop terminal or a mobile terminal, where the mobile terminal can be at least one of a mobile phone, tablet computer, or laptop computer. Transmission node 120 may include switches (or routers) in the network and other transmission equipment, such as SDH (Synchronous Digital Hierarchy) or PTN (Packet Transport Network) equipment. In addition, transmission node 120 may also be a communication base station, such as 3G, 4G and 5G and later versions of communication base stations.

As shown in Figure 2, in one embodiment, a voice processing method is provided. This embodiment mainly uses the application of this method to the terminal 110 in Figure 1 as an example. Referring to Figure 2, the voice processing method specifically includes the following steps:

S202, Perform speech rate detection on the acquired speech to obtain the speech rate value.

The language in question can be the voice uttered by the user during a voice or video call, or during a live voice or video broadcast. The speech rate value indicates the speed at which a speaker speaks; different speakers may have different speech rate values. This speech rate value can be an average speech rate or a value at a specific instant.

In one embodiment, when making a voice or video call, the terminal captures the user's voice through a microphone. For example, when a user makes a voice or video call with another person using an instant messaging application, the terminal captures the user's voice through a built-in microphone. The instant messaging application may include social networking applications and other applications used for instant messaging.

In one embodiment, when conducting live audio or video streaming, the terminal captures the user's voice through a microphone. For example, when a user uses live streaming software to conduct live audio or video streaming, the terminal captures the user's voice through a built-in microphone.

In one embodiment, the terminal performs phoneme detection on the collected speech to obtain a phoneme sequence; then, based on the obtained phoneme sequence, it calculates the number of phonemes per unit time, and determines the speech rate value based on the number of phonemes per unit time. The number of phonemes can be determined by changes in the fundamental frequency or pitch. For example, if there are 20 changes in the fundamental frequency or pitch per unit time, then there are 20 phonemes per unit time. Phonemes are divided into two main categories: vowels and consonants. They are the smallest units of speech defined based on the natural attributes of speech, analyzed according to the articulation actions within a syllable; one action constitutes one phoneme. For example, in Chinese, the syllable 'a' has only one phoneme, 'ai' has two phonemes, and 'dai' has three phonemes, etc.

In one embodiment, the terminal performs phoneme detection on the collected speech to obtain a phoneme sequence; then, it converts the phoneme sequence into a corresponding character sequence, calculates the number of characters per unit time based on the converted character sequence, and determines the speech rate value based on the number of characters per unit time.

In one embodiment, the terminal uses a window function to frame the acquired speech, thus obtaining framed speech. Specifically, overlapping segmentation can be used to frame the acquired speech, allowing for smooth transitions between frames. The overlapping portion between the previous and next frames is called the frame shift, and the ratio between the frame shift length and the speech frame length is typically in the range of 0 to 0.5. The window function can be a rectangular window function, Hanning window function, Hamming window function, or Blackman window function, etc.

For example, assuming the speech is represented by s(n), a window function ω(n) is used to multiply s(n) to form the windowed speech Sω(n) = s(n) × ω(n), as shown in Figure 3. The length of the speech frame is N, and the length of the shifted frame is M.

In one embodiment, the terminal detects each speech frame to determine whether it contains speech content. The terminal can then perform speech rate detection on speech frames containing speech content to obtain a speech sequence. Based on the obtained phoneme sequence, the terminal calculates the number of phonemes per unit time and determines the speech rate value accordingly. Alternatively, the terminal converts the phoneme sequence into a corresponding word sequence, calculates the number of words per unit time based on the converted word sequence, and determines the speech rate value based on the number of words per unit time.

S204, obtain the forward error correction redundancy.

The aforementioned forward error correction redundancy is configured based on the packet loss rate. Forward error correction is an error control method that transmits redundant packets simultaneously with voice transmission. When packet loss or errors occur during transmission, the receiver can reconstruct the lost or erroneous portion of the voice from the redundant packets. For example, before voice is sent to the transmission channel, the corresponding voice coding packet undergoes forward error correction coding to obtain a redundant packet with the characteristics of the voice itself. Then, the voice coding packet and the redundant packet are transmitted together to the receiver. The receiver decodes the received voice coding packet and redundant packet to identify and correct any erroneous or lost voice coding packets generated during transmission. Forward error correction redundancy can be defined as the ratio of the number of redundant packets formed during forward error correction coding to the number of voice coding packets. The forward error correction redundancy can be configured based on the voice coding packet loss rate.

In one embodiment, when a terminal receives a voice-coded packet sent by another party, it determines the forward error correction redundancy based on the packet loss rate of the received voice-coded packet. A higher packet loss rate results in a higher configured forward error correction redundancy; conversely, a lower packet loss rate results in a lower configured forward error correction redundancy.

In one embodiment, the terminal can also predict the packet loss rate based on network quality and configure the corresponding forward error correction redundancy based on the predicted packet loss rate. Alternatively, the terminal can also configure the corresponding forward error correction redundancy based on network quality.

For example, when network quality is poor, the packet loss rate is usually high, so a larger forward error correction redundancy can be configured. When network quality is good, the packet loss rate is usually low, so a smaller forward error correction redundancy can be configured.

S206, adjust the forward error correction redundancy based on the speech rate value to obtain the target redundancy.

Different speakers, due to differences in language and speaking habits, exhibit varying speech rates. When a speaker speaks quickly, the amount of information contained in the speech collected per unit time is greater, meaning it includes many different phonemes within a given timeframe. Therefore, even the loss of a small number of speech packets can result in the absence of many phonemes, leading to incomplete information received by the receiver. Conversely, when a speaker speaks slowly, the amount of information contained in the speech collected per unit time is less, meaning it contains fewer phonemes within a given timeframe, mostly consisting of phonemes with similar characteristics. In this case, even if a small number of speech packets are lost, the receiver can still understand the content expressed by the sender through the remaining phonemes heard.

In one embodiment, when the speech rate is high, the terminal can increase the forward error correction redundancy; when the speech rate is low, the terminal can decrease the forward error correction redundancy to obtain the target redundancy.

In one embodiment, when the speech rate is high, the terminal can obtain a corresponding first adjustment coefficient and use the product of the first adjustment coefficient and the forward error correction redundancy as the target redundancy. When the speech rate is low, the terminal can obtain a corresponding second adjustment coefficient and use the product of the second adjustment coefficient and the forward error correction redundancy as the target redundancy.

S208, perform speech encoding on the speech to obtain a speech encoding packet.

In one embodiment, the terminal samples the acquired speech, wherein the sampling frequency is greater than twice the highest frequency of the speech signal. Then, the terminal quantizes the sampled speech, which can be uniform quantization or non-uniform quantization. Non-uniform quantization can employ μ-law compression or A-law compression algorithms. Finally, the terminal encodes the quantized speech and packages the resulting encoded speech data into multiple speech code packets. The encoding methods include waveform coding (such as pulse code modulation), parametric coding, and hybrid coding.

When uniform quantization is used to quantize sampled speech, the same quantization interval is applied to both large and small amplitude speech, thus adapting to large amplitude speech while ensuring quantization accuracy. When non-uniform quantization is used, a larger quantization interval is applied to large amplitude speech, and a smaller quantization interval is applied to small amplitude speech, thus allowing for a smaller quantization bit depth while maintaining accuracy.

S210, perform forward error correction coding on the speech coding packet according to the target redundancy to obtain the redundant packet.

In one embodiment, the terminal performs forward error correction coding on the speech coding packets according to the target redundancy to obtain redundant packets. The number of redundant packets is the product of the target redundancy and the number of speech coding packets.

For example, suppose the number of speech coding packets is k, and the word length is w bits, where w can be 8, 16, or 32. The terminal performs forward error correction coding on the k speech coding packets according to the target redundancy, generating m redundant packets corresponding to the speech coding packets.

S212, send redundant packets and voice coding packets to the receiving end.

In one embodiment, the terminal uses a real-time transmission protocol to encapsulate voice encoding packets and redundancy packets to obtain encapsulated voice data packets, and then sends the encapsulated voice data packets obtained from the voice encoding packets and redundancy packets to the receiving end.

The Real-Time Transport Protocol (RTP) provides end-to-end voice transmission services with real-time characteristics. RTP implements ordered transmission, allowing the receiver to reassemble the packet sequence from the sender, and the sequence number is also used to determine the appropriate packet position. The voice data packet is an RTP message format voice data packet, consisting of two parts: a header and a payload. The payload consists of voice-coded packets and redundant packets.

For example, let the forward error correction redundancy of k speech coding packets be r/k. Based on the target redundancy and the speech coding packets, the number of redundant packets is calculated to be r. Let the r redundant data packets be C = ( _C1 , _C2 , ..., _Cr ), then the speech data packet can be represented as follows: where _Yi = _Di (0 ≤ i ≤ k-1), _Yj = _Cj (k ≤ j ≤ n-1). B is an n×k dimensional forward error correction generator matrix, composed of an identity matrix I and a matrix G. The speech data packet can then be represented as follows:

As an example, as shown in Figure 4, the terminal encodes the collected speech to obtain speech encoding packets p1-p8. When the forward error correction redundancy is adjusted to obtain the target redundancy, forward error correction encoding is performed on speech encoding packets p1-p8 according to the target redundancy requirements to obtain redundant packets r1, r2, and r3. Then, RTP encapsulation is used to obtain a voice data packet containing speech encoding packets p1-p8 and redundant packets r1-r3. Finally, the voice data packet is sent to the receiving end through the network.

In one embodiment, the terminal can also receive voice data packets sent from the receiving end. The voice data packets contain voice encoded packets and redundant packets. If packet loss is found when parsing the voice data packets, the lost voice encoded packets can be reconstructed based on the remaining voice encoded packets and redundant packets to obtain the complete voice encoded packets. The voice encoded packets can then be decoded to obtain the corresponding voice.

For example, at the receiving end, if the receiving end receives any k data packets from the voice data packet, it can extract the corresponding rows from the forward error correction generator matrix based on the position information of the received data packets in the voice data packet, and form a new k×k dimensional matrix B’, then:

If matrix B’ is a non-singular matrix, the original speech code packet is obtained through the following inverse transformation, thus completing the recovery: The transformation formula is as follows:

In the above embodiments, by detecting the speech rate, the forward error correction redundancy is adjusted using the detected speech rate value. The adjusted target redundancy can then be used to perform forward error correction encoding on the speech coding packet, thereby obtaining a redundant packet. When the speech rate is slow, the speech data packet contains less speech content, while when the speech rate is fast, the speech data packet contains more speech content. Dynamically adjusting the forward error correction redundancy according to the speech rate value can ensure that lost speech data packets can be effectively recovered, thereby achieving effective error correction of speech data packets and avoiding the consumption of a large amount of bandwidth.

In one embodiment, as shown in FIG5, S202 may specifically include:

S502, for voice recording.

The voice recording can be taken through a microphone during the initial stage of a user's voice or video call; or it can be taken through a microphone during the initial stage of a user's voice or video live stream.

In one embodiment, when making a voice or video call through an instant messaging application, the terminal captures the user's voice through a microphone. The instant messaging application may include social networking applications and other applications used for instant messaging.

In one embodiment, when conducting voice or video live streaming through live streaming software, the terminal captures the user's voice through a microphone.

In one embodiment, the terminal uses a window function to frame the acquired speech, thereby obtaining framed speech. Specifically, overlapping segmentation can be used to frame the acquired speech, thus enabling smooth transitions between frames. The overlapping portion between the previous and next frames is called frame shift, and the ratio between the frame shift length and the speech frame length is typically in the range of 0 to 0.5.

In one embodiment, the terminal detects each speech frame to determine whether each speech frame contains speech content, thereby enabling the terminal to detect the speech rate of speech frames containing speech content.

In one embodiment, detecting whether each speech frame contains speech content can specifically include: the terminal performing pulse code modulation on each speech frame to obtain PCM speech data, inputting the PCM speech data into a VAD speech detection function, and then outputting a speech identifier. For example, if the output speech identifier is 0, then it does not contain speech content; if the output speech identifier is 1, then it contains speech content.

S504 identifies phoneme sequences from speech.

Phonemes are divided into two main categories: vowels and consonants. They are the smallest units of speech determined by the natural attributes of speech sounds. They are analyzed based on the articulation actions within a syllable, with each action constituting a phoneme. For example, the Chinese syllable 'a' has only one phoneme, 'ai' has two phonemes, and 'dai' has three phonemes, etc.

In one embodiment, S504 may specifically include: performing pulse code modulation on the speech to obtain speech code data; identifying speech segments containing speech content from the speech code data; and identifying phoneme sequences from the speech segments of the speech code data.

In one embodiment, the step of performing pulse code modulation on speech to obtain speech coded data may specifically include: the terminal sampling the acquired speech, wherein the sampling frequency is greater than twice the highest frequency of the speech signal; then, the terminal quantizing the sampled speech, which may be uniform quantization or non-uniform quantization, and non-uniform quantization may employ a μ-law compression algorithm or an A-law compression algorithm; finally, the terminal encoding the quantized speech, and then packaging the encoded speech data into multiple speech coded packets, wherein the encoding methods include waveform encoding, parametric encoding, and hybrid encoding.

In another embodiment, S504 may specifically include: the terminal extracting speech features from the speech; decoding the speech features to obtain decoded speech features; and identifying a phoneme sequence from the decoded speech features.

Among them, speech features can be logarithmic power spectrum or Mel frequency cepstral coefficients with respect to speech.

In one embodiment, the terminal performs a Fourier transform on the collected speech, converting the speech in the time domain into a spectrum in the frequency domain. The terminal obtains the amplitude corresponding to the spectrum and calculates the power spectrum using the power density function based on the amplitude.

For example, assuming the signal expression of speech is f(t), we can obtain the spectrum by performing a Fourier transform on f(t). Let the expression of the spectrum be _FT (w). By substituting the amplitude corresponding to the spectrum into the following power spectral density function, we can obtain the power spectrum of speech.

Specifically, the terminal performs Viterbi decoding on the extracted speech features based on an adaptive acoustic model, and identifies the phoneme sequence from the decoded speech features. Furthermore, the terminal can determine the start and end times of each phoneme in the phoneme sequence.

S506, determine the speech rate value based on the frequency of phoneme transitions in the phoneme sequence.

In one embodiment, S506 may specifically include: detecting the number of jumps in the fundamental period or fundamental frequency of a phoneme per unit time in a phoneme sequence; and determining the speech rate value based on the number of jumps per unit time.

In one embodiment, the terminal determines whether the number of fundamental frequency or fundamental period transitions exceeds a preset fundamental frequency transition threshold. If so, it is determined that the pitch of the speech has changed significantly; otherwise, it is determined that the pitch of the speech has not changed significantly. The fundamental frequency and fundamental period are reciprocals of each other and can be converted between each other.

In one embodiment, the terminal performs pulse code modulation on each speech frame to obtain PCM speech data. This PCM speech data is then input into a fundamental frequency estimation function to obtain the fundamental frequency corresponding to each speech frame. The fundamental frequency estimation function can be based on a time-domain autocorrelation function.

In the above embodiments, by identifying phoneme sequences from the collected speech and determining the speech rate value based on the frequency of phoneme transitions in the phoneme sequence, the forward error correction redundancy can be dynamically adjusted according to the speech rate value. This ensures that lost speech data packets can be effectively recovered, thereby achieving effective error correction of speech data packets and avoiding the consumption of a large amount of bandwidth.

In one embodiment, as shown in FIG6, S206 may specifically include:

S602, when the speech rate value is greater than the speech rate upper limit value but less than the speech rate upper limit value, calculate the adjustment parameters based on the speech rate value; adjust the forward error correction redundancy according to the adjustment parameters to obtain the target redundancy.

In one embodiment, when the speech rate value is greater than the upper limit of speech rate but less than the upper limit of speech rate, if the speech rate value is larger, the terminal will increase the forward error correction redundancy; if the speech rate value is smaller, the terminal will decrease the forward error correction redundancy.

In one embodiment, the terminal inputs the speech rate value into a formula for adjusting the forward error correction redundancy. When calculating the adjustment parameters, the forward error correction redundancy is also adjusted to obtain the target redundancy.

For example, the formula for adjusting the forward error correction redundancy can be _V1 ≤ v ≤ _V2 , where r' is the adjusted target redundancy, _r0 is the forward error correction redundancy, and is the adjustment parameter, where c is a constant, v is the speech rate value, and _V1 and _V2 are the lower limit and upper limit of speech rate, respectively.

S604 compares the target redundancy with the upper limit and lower limit of redundancy, respectively.

For example, referring to the function _V1 ≤ v ≤ _V2 , the target redundancy is compared with the upper limit of redundancy _Rmax and the lower limit of redundancy _Rmin , respectively, and the final target redundancy is determined based on the comparison results. When the target redundancy is less than the upper limit of redundancy but greater than the lower limit of redundancy, the target redundancy is taken as the final target redundancy, and S606 is executed. When the target redundancy is less than the lower limit of redundancy, the lower limit of redundancy is taken as the final target redundancy, and S608 is executed. When the target redundancy is greater than the upper limit of redundancy, the upper limit of redundancy is taken as the final target redundancy, and S610 is executed.

S606: When the target redundancy is less than the upper limit of redundancy and greater than the lower limit of redundancy, forward error correction coding is performed on the speech coding packet according to the target redundancy to obtain the redundant packet.

In one embodiment, the terminal performs forward error correction coding on the speech coding packets according to the target redundancy to obtain redundant packets. The number of redundant packets is the product of the target redundancy and the number of speech coding packets.

For example, suppose the number of speech coding packets is k, and the word length is w bits, where w can be 8, 16, or 32. The terminal performs forward error correction coding on the k speech coding packets according to the target redundancy, generating m redundant packets corresponding to the speech coding packets.

S608: When the target redundancy is less than the redundancy lower limit, the speech coding packet is forward-corrected according to the redundancy lower limit to obtain the redundant packet.

In one embodiment, when the target redundancy is less than the redundancy lower limit, the terminal performs forward error correction coding on the speech coding packet according to the redundancy lower limit to obtain a redundant packet. The number of the redundant packets is the product of the redundancy lower limit and the number of speech coding packets.

S610: When the target redundancy is greater than the upper limit of redundancy, the speech coding packet is forward-corrected according to the upper limit of redundancy to obtain the redundant packet.

In one embodiment, when the target redundancy is greater than the redundancy limit, the terminal performs forward error correction coding on the speech coding packets according to the redundancy limit to obtain redundant packets. The number of redundant packets is the product of the redundancy limit and the number of speech coding packets.

Referring to the following function, if the speech rate value is less than the upper limit of speech rate, select the maximum value from the forward error correction redundancy and the lower limit of redundancy, and perform forward error correction coding on the speech code packet according to the maximum value to obtain a redundant packet; if the speech rate value is greater than the upper limit of speech rate but less than the upper limit of speech rate, perform forward error correction coding on the speech code packet according to the target redundancy to obtain a redundant packet; if the speech rate value is greater than the upper limit of speech rate, select the minimum value from the forward error correction redundancy and the upper limit of redundancy, and perform forward error correction coding on the speech code packet according to the minimum value to obtain a redundant packet.

In the above embodiments, the forward error correction redundancy is adjusted according to the speech rate value to obtain the adjusted target redundancy. Thus, the speech coding packet can be forward error corrected according to the target redundancy to obtain a redundant packet. The redundant packet and the speech coding packet are encapsulated into a speech data packet and sent to the receiving end. This can ensure that the speech data packets lost during transmission can be effectively recovered, thereby achieving effective error correction of the speech data packets and avoiding the consumption of a large amount of bandwidth.

As an example, this embodiment first performs speech rate detection on the user's speech to obtain an average speech rate value v. Assuming the forward error correction redundancy obtained based on the traditional forward error correction scheme is r _₀ , and the target redundancy after adjustment in this embodiment is r′, the calculation of the average speech rate value, the method of adjusting the forward error correction redundancy, and the forward error correction coding using the adjusted target redundancy are detailed below:

1) Calculate the average speech rate

In actual conversations, since the content of speech is unrestricted, this embodiment employs a no-reference detection method to measure the speaker's speech rate. The no-reference speech rate detection calculation is based on statistics of the vad and the rate of change of the pitch period. Because the pitch period (or pitch frequency) of the same phoneme is continuous with small jumps, while the pitch period (or pitch frequency) of different phonemes changes significantly between frames, the speech rate v is equivalently described by analyzing the number of pitch period (or pitch frequency) abrupt changes and the number of speech frame jumps per unit time. The pseudocode is as follows:

//initial

Changecnt = 0; // Changecnt represents the sum of the number of fundamental frequency transitions per unit time and the number of transitions from non-speech frames to speech frames.

Totcnt = 0; //Totcnt represents the current number of detected frames.

vad_cur = 0; //vad_cur represents the speech identifier of the current frame. A value of 0 indicates a non-speech frame, and a value of 1 indicates a speech frame.

vad_pre = 0; // vad_pre represents the voice identifier of the previous frame. A value of 0 indicates a non-voice frame, and a value of 1 indicates a voice frame.

pitchfreq_cur = 0; // pitchfreq_cur represents the fundamental frequency of the current frame.

pitchfreq_pre = 0; // pitchfreq_pre represents the fundamental frequency of the previous frame.

pitchfreq_cur = PitchFreqEst(pcmdata); // PitchFreqEst() is the fundamental frequency estimation function. The input is PCM speech data, and the output is the estimated fundamental frequency. The fundamental frequency can be estimated based on time-domain autocorrelation or cepstral methods.

vad_cur = VadDet(pcmdata); // VadDet() is the VAD speech detection function. The input is PCM speech data, and the output is a speech identifier. An output of 0 indicates a non-speech frame without speech content, and an output of 1 indicates a speech frame containing speech content.

If Totcnt <= T Then //T represents the speech rate detection period

Totcnt++;

If vad_cur = 1, then

if abs(pitchfreq_pre-pitchfreq_cur)>threshold1 Then //threshold1 represents the set fundamental frequency hopping threshold. If the change value exceeds this threshold, it indicates that the pitch has changed significantly.

Changecnt++;

End

End

pitchfreq_pre=pitchfreq_cur;

If vad_cur＝1and vad_pre＝0Then

Changecnt++;

End

Else

v = Changecnt;

Totcnt = 0;

Changecnt = 0;

End

Through the above process, the splicing speech rate value v can be obtained, which can be used for subsequent FEC redundancy calculation.

2) FEC redundancy calculation

The average speech rate value v obtained above, and the final target redundancy r′ are calculated using the following formula:

The above formula pre-sets the following constant values: upper limit of speech rate _V2 and lower limit of speech rate _V1 , lower limit of FEC redundancy _Rmin and upper limit of redundancy _Rmax , and c is a constant. The target redundancy r′ can be calculated using the above formula.

3) Acquire the user's voice and encode it to obtain multiple voice encoding packets; then perform forward error correction encoding on the voice encoding packets according to the target redundancy r′ to obtain the corresponding redundant packets; then package the redundant packets and voice encoding packets together using RTP to obtain RTP voice data packets, and then send the RTP voice data packets to the receiving end through the network, as shown in Figure 7.

As shown in Figure 7, after receiving the RTP voice data packets, the receiving end calculates the packet loss rate and performs forward error correction decoding to recover the lost voice encoded packets. Then, it decodes all the voice encoded packets to obtain the user's original voice.

Adjusting FEC redundancy based on speaker rate detection results ensures more effective protection of transmitted voice content, improves end-to-end call voice quality, and enables highly reliable real-time voice data transmission for services such as VoIP (Voice over Internet Protocol), broadcasting, and live voice and video streaming.

Figures 2, 5, and 6 are schematic flowcharts of a speech processing method in one embodiment. It should be understood that although the steps in the flowcharts of Figures 2, 5, and 6 are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Furthermore, at least some of the steps in Figures 2, 5, and 6 may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps.

As shown in Figure 8, this embodiment of the invention provides a voice processing device, which includes: a detection module 802, an acquisition module 804, an adjustment module 806, a first encoding module 808, a second encoding module 810, and a transmission module 812; wherein:

The detection module 802 is used to detect the speech rate of the acquired speech and obtain the speech rate value;

Module 804 is used to obtain the forward error correction redundancy.

The adjustment module 806 is used to adjust the forward error correction redundancy based on the speech rate value to obtain the target redundancy.

The first encoding module 808 is used to encode the speech to obtain a speech encoding packet;

The second encoding module 810 is used to perform forward error correction encoding on the speech encoding packet according to the target redundancy to obtain a redundant packet;

The transmitting module 812 is used to send redundant packets and voice coding packets to the receiving end.

In one embodiment, as shown in FIG9, the device further includes: an encapsulation module 814; wherein:

The encapsulation module 814 is used to encapsulate voice coding packets and redundant packets using a real-time transmission protocol to obtain encapsulated voice data packets.

The transmitting module 812 is also used to transmit voice data packets obtained by encapsulating voice coding packets and redundancy packets to the receiving end.

In the above embodiments, by detecting the speech rate, the forward error correction redundancy is adjusted using the detected speech rate value. The adjusted target redundancy can then be used to perform forward error correction encoding on the speech coding packet, thereby obtaining a redundant packet. When the speech rate is slow, the speech data packet contains less speech content, while when the speech rate is fast, the speech data packet contains more speech content. Dynamically adjusting the forward error correction redundancy according to the speech rate value can ensure that lost speech data packets can be effectively recovered, thereby achieving effective error correction of speech data packets and avoiding the consumption of a large amount of bandwidth.

In one embodiment, the detection module 802 is further configured to: acquire speech; identify phoneme sequences from the speech; and determine speech rate values based on the frequency of phoneme transitions in the phoneme sequence.

In one embodiment, the detection module 802 is further configured to: perform pulse code modulation on the speech to obtain speech code data; identify speech segments from the speech code data; and identify phoneme sequences from the speech segments of the speech code data.

In one embodiment, the detection module 802 is further configured to: extract speech features from speech; decode the speech features to obtain decoded speech features; and identify phoneme sequences from the decoded speech features.

In one embodiment, the detection module 802 is further configured to: detect the number of jumps in the fundamental period or fundamental frequency of a phoneme within a unit time in the phoneme sequence; and determine the speech rate value based on the number of jumps within a unit time.

In the above embodiments, by identifying phoneme sequences from the collected speech and determining the speech rate value based on the frequency of phoneme transitions in the phoneme sequence, the forward error correction redundancy can be dynamically adjusted according to the speech rate value. This ensures that lost speech data packets can be effectively recovered, thereby achieving effective error correction of speech data packets and avoiding the consumption of a large amount of bandwidth.

In one embodiment, the adjustment module 806 is further configured to: calculate adjustment parameters based on the speech rate value when the speech rate value is greater than the speech rate upper limit value but less than the speech rate upper limit value; and adjust the forward error correction redundancy according to the adjustment parameters to obtain the target redundancy.

In one embodiment, as shown in FIG9, the device further includes: a comparison module 816; wherein:

The comparison module 816 is used to compare the target redundancy with the upper limit and lower limit of redundancy, respectively. When the target redundancy is less than the upper limit and greater than the lower limit, the second encoding module 810 performs forward error correction encoding on the speech encoding packet according to the target redundancy to obtain the redundant packet.

In one embodiment, the second encoding module 810 is further configured to: when the target redundancy is less than the lower limit of redundancy, perform forward error correction encoding on the speech encoding packet according to the lower limit of redundancy to obtain a redundant packet; when the target redundancy is greater than the upper limit of redundancy, perform forward error correction encoding on the speech encoding packet according to the upper limit of redundancy to obtain a redundant packet.

In one embodiment, the second encoding module 810 is further configured to:

If the speech rate value is less than the upper limit of speech rate, the maximum value is selected from the forward error correction redundancy and the lower limit of redundancy. The speech coding packet is then forward error correction encoded according to the maximum value to obtain the redundant packet.

If the speech rate value is greater than the speech rate limit but less than the speech rate limit, perform forward error correction coding on the speech code packet according to the target redundancy to obtain a redundant packet.

If the speech rate value is greater than the upper limit of speech rate, the minimum value is selected from the forward error correction redundancy and the upper limit of redundancy. Forward error correction coding is performed on the speech coding packet according to the minimum value to obtain the redundant packet.

In the above embodiments, the forward error correction redundancy is adjusted according to the speech rate value to obtain the adjusted target redundancy. Thus, the speech coding packet can be forward error corrected according to the target redundancy to obtain a redundant packet. The redundant packet and the speech coding packet are encapsulated into a speech data packet and sent to the receiving end. This can ensure that the speech data packets lost during transmission can be effectively recovered, thereby achieving effective error correction of the speech data packets and avoiding the consumption of a large amount of bandwidth.

Figure 10 shows an internal structural diagram of a computer device in one embodiment. Specifically, this computer device may be the terminal 110 in Figure 1. As shown in Figure 10, the computer device includes a processor, memory, network interface, input device, and display screen connected via a system bus. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and may also store a computer program. When executed by the processor, this computer program enables the processor to implement a voice processing method. The internal memory may also store a computer program, which, when executed by the processor, enables the processor to perform a voice processing method. The display screen of the computer device may be a liquid crystal display (LCD) or an e-ink display. The input device may be a touch layer covering the display screen, or buttons, a trackball, or a touchpad mounted on the computer device casing, or an external keyboard, touchpad, or mouse, etc.

Those skilled in the art will understand that the structure shown in Figure 10 is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or may combine certain components, or may have different component arrangements.

In one embodiment, the voice processing apparatus provided in this application can be implemented as a computer program, which can run on the computer device shown in FIG10. The memory of the computer device can store various program modules constituting the voice processing apparatus, such as the detection module 802, acquisition module 804, adjustment module 806, first encoding module 808, second encoding module 810, and transmission module 812 shown in FIG8. The computer program composed of these program modules causes the processor to execute the steps in the voice processing methods of the various embodiments of this application described in this specification.

For example, the computer device shown in Figure 10 can execute S202 via the detection module 802 in the voice processing device shown in Figure 8. The computer device can execute S204 via the acquisition module 804. The computer device can execute S206 via the adjustment module 806. The first encoding module 808 executes S208. The computer device can execute S210 via the second encoding module 810. The computer device can execute S212 via the transmission module 812.

In one embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, which, when executed by the processor, causes the processor to perform the following steps: performing speech rate detection on acquired speech to obtain a speech rate value; obtaining forward error correction redundancy; adjusting the forward error correction redundancy according to the speech rate value to obtain a target redundancy; performing speech encoding on the speech to obtain a speech encoding packet; performing forward error correction encoding on the speech encoding packet according to the target redundancy to obtain a redundant packet; and sending the redundant packet and the speech encoding packet to a receiving end.

In one embodiment, when a computer program is executed by a processor to detect the speech rate of acquired speech and obtain a speech rate value, the processor specifically performs the following steps: acquiring speech; identifying a phoneme sequence from the speech; and determining the speech rate value based on the frequency of phoneme transitions in the phoneme sequence.

In one embodiment, when a computer program is executed by a processor to identify a phoneme sequence from speech, the processor specifically performs the following steps: pulse code modulation of the speech to obtain speech code data; identifying speech segments containing speech content from the speech code data; and identifying a phoneme sequence from the speech segments of the speech code data.

In one embodiment, when a computer program is executed by a processor to identify a phoneme sequence from speech, the processor specifically performs the following steps: extracting speech features from the speech; decoding the speech features to obtain decoded speech features; and identifying a phoneme sequence from the decoded speech features.

In one embodiment, when a computer program is executed by a processor to determine a speech rate value based on the frequency of phoneme transitions in a phoneme sequence, the processor specifically performs the following steps: detecting the number of transitions of the fundamental period or fundamental frequency of phonemes per unit time in the phoneme sequence; and determining the speech rate value based on the number of transitions per unit time.

In one embodiment, when the computer program is executed by the processor to adjust the forward error correction redundancy based on the speech rate value to obtain the target redundancy, the processor specifically performs the following steps: when the speech rate value is greater than the speech rate upper limit value but less than the speech rate upper limit value, calculate the adjustment parameters based on the speech rate value; adjust the forward error correction redundancy according to the adjustment parameters to obtain the target redundancy.

In one embodiment, when the computer program is executed by the processor, the processor further performs the following steps: comparing the target redundancy with the upper limit of redundancy and the lower limit of redundancy respectively; when the target redundancy is less than the upper limit of redundancy and greater than the lower limit of redundancy, performing forward error correction coding on the speech coding packet according to the target redundancy to obtain a redundant packet.

In one embodiment, when the computer program is executed by the processor, the processor also performs the following steps: when the target redundancy is less than the redundancy lower limit, the speech coding packet is forward-corrected according to the redundancy lower limit to obtain a redundant packet;

When the target redundancy exceeds the upper limit of redundancy, forward error correction coding is performed on the speech code packet according to the upper limit of redundancy to obtain the redundant packet.

In one embodiment, when the computer program is executed by the processor, the processor also performs the following steps:

If the speech rate value is less than the upper limit of speech rate, the maximum value is selected from the forward error correction redundancy and the lower limit of redundancy. The speech coding packet is then forward error correction encoded according to the maximum value to obtain the redundant packet.

If the speech rate value is greater than the speech rate limit but less than the speech rate limit, perform forward error correction coding on the speech code packet according to the target redundancy to obtain a redundant packet.

If the speech rate value is greater than the upper limit of speech rate, the minimum value is selected from the forward error correction redundancy and the upper limit of redundancy. Forward error correction coding is performed on the speech coding packet according to the minimum value to obtain the redundant packet.

In one embodiment, when the computer program is executed by the processor, the processor also performs the following steps: encapsulating the voice coding packet and the redundant packet using a real-time transport protocol to obtain the encapsulated voice data packet;

When a computer program is executed by a processor to send redundant packets and voice-coded packets to the receiving end, the processor specifically performs the following steps:

Send voice data packets, which are encapsulated with voice encoding packets and redundant packets, to the receiving end.

In one embodiment, a computer-readable storage medium is provided, storing a computer program that, when executed by a processor, causes the processor to perform the following steps: performing speech rate detection on the acquired speech to obtain a speech rate value; obtaining forward error correction redundancy; adjusting the forward error correction redundancy according to the speech rate value to obtain a target redundancy; performing speech encoding on the speech to obtain a speech encoding packet; performing forward error correction encoding on the speech encoding packet according to the target redundancy to obtain a redundant packet; and sending the redundant packet and the speech encoding packet to a receiving end.

In one embodiment, when a computer program is executed by a processor to detect the speech rate of acquired speech and obtain a speech rate value, the processor specifically performs the following steps: acquiring speech; identifying a phoneme sequence from the speech; and determining the speech rate value based on the frequency of phoneme transitions in the phoneme sequence.

In one embodiment, when a computer program is executed by a processor to identify a phoneme sequence from speech, the processor specifically performs the following steps: pulse code modulation of the speech to obtain speech code data; identifying speech segments containing speech content from the speech code data; and identifying a phoneme sequence from the speech segments of the speech code data.

In one embodiment, when a computer program is executed by a processor to identify a phoneme sequence from speech, the processor specifically performs the following steps: extracting speech features from the speech; decoding the speech features to obtain decoded speech features; and identifying a phoneme sequence from the decoded speech features.

In one embodiment, when a computer program is executed by a processor to determine a speech rate value based on the frequency of phoneme transitions in a phoneme sequence, the processor specifically performs the following steps: detecting the number of transitions of the fundamental period or fundamental frequency of phonemes per unit time in the phoneme sequence; and determining the speech rate value based on the number of transitions per unit time.

In one embodiment, when the computer program is executed by the processor to adjust the forward error correction redundancy based on the speech rate value to obtain the target redundancy, the processor specifically performs the following steps: when the speech rate value is greater than the speech rate upper limit value but less than the speech rate upper limit value, calculate the adjustment parameters based on the speech rate value; adjust the forward error correction redundancy according to the adjustment parameters to obtain the target redundancy.

In one embodiment, when the computer program is executed by the processor, the processor further performs the following steps: comparing the target redundancy with the upper limit of redundancy and the lower limit of redundancy respectively; when the target redundancy is less than the upper limit of redundancy and greater than the lower limit of redundancy, performing forward error correction coding on the speech coding packet according to the target redundancy to obtain a redundant packet.

In one embodiment, when the computer program is executed by the processor, the processor also performs the following steps: when the target redundancy is less than the redundancy lower limit, the speech coding packet is forward-corrected according to the redundancy lower limit to obtain a redundant packet;

When the target redundancy exceeds the upper limit of redundancy, forward error correction coding is performed on the speech code packet according to the upper limit of redundancy to obtain the redundant packet.

In one embodiment, when the computer program is executed by the processor, the processor also performs the following steps:

If the speech rate value is less than the upper limit of speech rate, the maximum value is selected from the forward error correction redundancy and the lower limit of redundancy. The speech coding packet is then forward error correction encoded according to the maximum value to obtain the redundant packet.

If the speech rate value is greater than the speech rate limit but less than the speech rate limit, perform forward error correction coding on the speech code packet according to the target redundancy to obtain a redundant packet.

If the speech rate value is greater than the upper limit of speech rate, the minimum value is selected from the forward error correction redundancy and the upper limit of redundancy. Forward error correction coding is performed on the speech coding packet according to the minimum value to obtain the redundant packet.

In one embodiment, when the computer program is executed by the processor, the processor also performs the following steps: encapsulating the voice coding packet and the redundant packet using a real-time transport protocol to obtain the encapsulated voice data packet;

When a computer program is executed by a processor to send redundant packets and voice-coded packets to the receiving end, the processor specifically performs the following steps:

Send voice data packets, which are encapsulated with voice encoding packets and redundant packets, to the receiving end.

Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and/or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM, etc.

The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims (22)

1. A speech processing method, comprising: The acquired speech is subjected to speech rate detection to obtain the speech rate value; Obtain the forward error correction redundancy; When the speech rate value is greater than the lower limit of speech rate and less than the upper limit of speech rate, the adjustment parameters are calculated based on the speech rate value; the forward error correction redundancy is adjusted according to the adjustment parameters to obtain the target redundancy. The speech is encoded to obtain a speech encoding packet; The speech coding packet is forward-corrected according to the target redundancy to obtain a redundant packet; The redundant packet and the voice encoding packet are sent to the receiving end. 2. The method according to claim 1, characterized in that, the step of performing speech rate detection on the acquired speech to obtain a speech rate value includes: Voice recording; Identify the phoneme sequence from the speech; The speech rate value is determined based on the frequency of phoneme transitions in the phoneme sequence. 3. The method according to claim 2, wherein identifying the phoneme sequence from the speech comprises: The speech is subjected to pulse code modulation to obtain speech coded data; Identify speech segments containing speech content from the speech coding data; Phoneme sequences are identified from the speech segments of the speech encoded data. 4. The method according to claim 2, wherein identifying the phoneme sequence from the speech comprises: Extract speech features from the speech; The speech features are decoded to obtain the decoded speech features; Phoneme sequences are identified from the decoded speech features. 5. The method according to claim 2, wherein determining the speech rate value based on the frequency of phoneme transitions in the phoneme sequence comprises: The number of fundamental frequency jumps of a phoneme per unit time is detected in the phoneme sequence. The speech rate value is determined based on the number of jumps per unit time. 6. The method according to any one of claims 1 to 5, wherein the forward error correction redundancy is configured according to network quality; or, The forward error correction redundancy is determined based on the packet loss rate, which is obtained based on the network quality prediction. 7. The method according to claim 1, characterized in that the method further comprises: The target redundancy is compared with the upper limit and lower limit of redundancy, respectively. When the target redundancy is less than the upper limit of redundancy and greater than the lower limit of redundancy, the step of performing forward error correction coding on the speech code packet according to the target redundancy to obtain a redundant packet is performed. 8. The method according to claim 7, characterized in that the method further comprises: When the target redundancy is less than the redundancy lower limit, the speech coding packet is forward error correction encoded according to the redundancy lower limit to obtain a redundant packet; When the target redundancy is greater than the upper limit of redundancy, the speech coding packet is forward-corrected according to the upper limit of redundancy to obtain a redundant packet. 9. The method according to any one of claims 1 to 5, characterized in that the method further comprises: If the speech rate value is less than the upper limit of speech rate, the maximum value is selected from the forward error correction redundancy and the lower limit of redundancy, and the speech coding packet is forward error correction encoded according to the maximum value to obtain a redundant packet; If the speech rate value is greater than the upper limit of speech rate, the minimum value is selected from the forward error correction redundancy and the upper limit of redundancy, and the speech coding packet is forward error correction encoded according to the minimum value to obtain a redundant packet. 10. The method according to any one of claims 1 to 5, characterized in that, before sending the redundant packet and the voice coding packet to the receiving end, the method further comprises: The voice encoding packet and the redundant packet are encapsulated using a real-time transmission protocol to obtain an encapsulated voice data packet; Sending the redundant packet and the voice coding packet to the receiving end includes: Send a voice data packet, which encapsulates the voice encoding packet and the redundant packet, to the receiving end. 11. A voice processing device, characterized in that the device comprises: The detection module is used to detect the speech rate of the acquired speech and obtain the speech rate value; The acquisition module is used to obtain the forward error correction redundancy. An adjustment module is used to calculate adjustment parameters based on the speech rate value when the speech rate value is greater than the lower limit of speech rate and less than the upper limit of speech rate; and to adjust the forward error correction redundancy according to the adjustment parameters to obtain the target redundancy. The first encoding module is used to encode the speech to obtain a speech encoding packet; The second encoding module is used to perform forward error correction encoding on the speech encoding packet according to the target redundancy to obtain a redundant packet; The sending module is used to send the redundant packet and the voice encoding packet to the receiving end. 12. The apparatus according to claim 11, wherein the detection module is further configured to: Voice recording; Identify the phoneme sequence from the speech; The speech rate value is determined based on the frequency of phoneme transitions in the phoneme sequence. 13. The apparatus according to claim 12, wherein the detection module is further configured to: The speech is subjected to pulse code modulation to obtain speech coded data; Speech segments are identified from the speech coding data; Phoneme sequences are identified from the speech segments of the speech encoded data. 14. The apparatus according to claim 12, wherein the detection module is further configured to: Extract speech features from the speech; The speech features are decoded to obtain the decoded speech features; Phoneme sequences are identified from the decoded speech features. 15. The apparatus according to claim 12, wherein the detection module is further configured to: The number of fundamental frequency jumps of a phoneme per unit time is detected in the phoneme sequence. The speech rate value is determined based on the number of jumps per unit time. 16. The apparatus according to any one of claims 11 to 15, wherein the forward error correction redundancy is configured according to network quality; or, The forward error correction redundancy is determined based on the packet loss rate, which is obtained based on the network quality prediction. 17. The apparatus according to claim 11, wherein the apparatus further comprises: The comparison module is used to compare the target redundancy with the upper limit of redundancy and the lower limit of redundancy, respectively. The second encoding module is further configured to perform forward error correction encoding on the speech encoding packet according to the target redundancy when the target redundancy is less than the upper limit of redundancy and greater than the lower limit of redundancy, so as to obtain a redundant packet. 18. The apparatus according to claim 17, wherein the second encoding module is further configured to: when the target redundancy is less than the lower limit of redundancy, perform forward error correction encoding on the speech encoding packet according to the lower limit of redundancy to obtain a redundant packet; and when the target redundancy is greater than the upper limit of redundancy, perform forward error correction encoding on the speech encoding packet according to the upper limit of redundancy to obtain a redundant packet. 19. The apparatus according to any one of claims 11 to 15, wherein the second encoding module is further configured to: if the speech rate value is less than the upper limit of speech rate, select the maximum value from the forward error correction redundancy and the lower limit of redundancy, and perform forward error correction encoding on the speech encoding packet according to the maximum value to obtain a redundant packet; if the speech rate value is greater than the upper limit of speech rate, select the minimum value from the forward error correction redundancy and the upper limit of redundancy, and perform forward error correction encoding on the speech encoding packet according to the minimum value to obtain a redundant packet. 20. The apparatus according to any one of claims 11 to 15, characterized in that the apparatus further comprises: An encapsulation module is used to encapsulate the voice encoding packet and the redundant packet using a real-time transmission protocol to obtain an encapsulated voice data packet; The sending module is also used to send a voice data packet, which encapsulates the voice encoding packet and the redundant packet, to the receiving end. 21. A computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to perform the steps of the method as claimed in any one of claims 1 to 10. 22. A computer device comprising a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the steps of the method as claimed in any one of claims 1 to 10.