CN100505714C - Drop-frame processing device and method based on ADPCM - Google Patents
Drop-frame processing device and method based on ADPCM Download PDFInfo
- Publication number
- CN100505714C CN100505714C CNB2005100338748A CN200510033874A CN100505714C CN 100505714 C CN100505714 C CN 100505714C CN B2005100338748 A CNB2005100338748 A CN B2005100338748A CN 200510033874 A CN200510033874 A CN 200510033874A CN 100505714 C CN100505714 C CN 100505714C
- Authority
- CN
- China
- Prior art keywords
- frame losing
- frame
- module
- adaptive
- losing
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Landscapes
- Compression, Expansion, Code Conversion, And Decoders (AREA)
Abstract
This invention relates to a missing frame processing device and method basing on self-adapting difference impulse code modulation, when ADPCM decoder receives frame, decoding directly,then depositing the decoded data and outputting the data; when the frame is missing, complementing frame according to the deposited voice information, and using the missing frame complement value combining with the state memorizer value of ADPCM decoder to rectify the ADPCM decoder's state. This invention also detects the static voice of the missing frame complement value, which includes the following steps: first it bases on that whether missing frame complement valued produces inhibit windows with different sizes, and then it operates noise lowering procession of different information lengths for the ADPCM decoder's output according to the inhibit windows' size.
Description
Technical field
The present invention relates to the IP network technical field, relate in particular to the drop-frame processing device and the method for modulating (ADPCM) in a kind of IP network based on adaptive difference pulse code.
Background technology
IP (Internet Protocol) phone is a kind of digital telephone, it is a kind of communication service business of technological innovation, it voice, compressed encoding, packing grouping, distribute route, memory transactions, unpack exchange processing such as decompress(ion) and on IP network or the Internet, realize voice communication, it has promoted network resource usage, reduce the speech business cost, therefore obtained development rapidly in the world.Can be described as with fastest developing speed in the world today, popularize one of the fastest application service technology, also be one of focus of paying close attention to of computer network circle.
VoIP (being carried on the voice on the IP, i.e. IP phone) by voice signal is carried out encode digitalized, compression is processed into condensed frame, be converted to the IP packet then and on IP network, transmit, thereby reached purpose in the communication of the enterprising lang sound of IP network.Be illustrated in figure 1 as the transmission course schematic diagram of VoIP, the VoIP transmission course can be divided into the grouping of voice digitization, signal encoding, signal packing and transmit, unpacks and plurality of processes such as decompression process and digital speech be simulated.
The key technology of VoIP comprises signaling technology, coding techniques, Real-time Transmission, service quality (QoS) assurance technology and network transmission technology etc.Wherein coding techniques is divided into waveform coding, parameter coding and hybrid coding again, wherein wave coder as far as possible structure go out to comprise the analog waveform of background noise.Because wave coder acts on all input signals, therefore can produce high-quality sample value.
For waveform coding, G.711, ITU-T advises having stipulated that A leads and Mu leads PCM (pulse code modulation) segmentation logarithm quantization algorithm, can on the speed of 64kbit/s, obtain gratifying effect, be widely used in plurality of communication systems such as cable, microwave, satellite, optical cable.Owing to G.711 only utilized the one dimension statistical nature of voice signal, when code rate further reduces, speech quality will descend obviously, and for many application scenarios, the frequency band that the code rate of 64kbit/s takies is too wide, communication cost is too big, so people seek to obtain the code decode algorithm of high-quality speech always on low rate more.
ADPCM is a kind of encoding and decoding technique that utilizes adaptive technique and differential coding technology to combine, can make the pulse code data of the further boil down to 32kbit/s of pulse code (PCM) signal of 64kbit/s, make the transmission pulse needed bandwidth of encoding reduce half, improved the utilization ratio of channel, and the communication quality of pulse code modulating system (PCM System) is improved.The basic principle of ADPCM is to utilize the sample value of several sample values in past being predicted current input, and make prediction circuit have adaptive forecast function and the actual detected value compares, at any time the difference that records is quantized differential processing automatically, make it to remain with signal Synchronization and change.
CCITT proposed G.721 suggestion in 1984, and this is a kind of speech coding algorithm standard based on ADPCM (adaptive differential PCM) technology, and code rate is 32kbit/s, and each voice sample value is equivalent to encode with 4bit; After this G.723 replenishing 40kbit/s and two kinds of code rates of 24kbit/s in the suggestion, and G.726 replenishing the 16kbit/s code rate in the suggestion, thereby forming the ADPCM algorithm standard of the many speed of a cover in nineteen ninety.Wherein 24 and the channel of 16bits/s be mainly used in overload transferring voice in the digital circuit multiplication equipment (DCME); 40bits/s is mainly used to transmit data modulated signal in DCME, be used for transmitting the modulation signal greater than 4800bits/s especially.G.726 algorithm can be in 40kbit/s, 32kbit/s and three kinds of speed of 24kbit/s be dynamically adjusted, and reaching the purpose that increases capacity in given channel, thereby is used widely in applications such as voice storage and voice transfer.
But IP wraps in the network transport process, can't avoid wrapping in destroyed in the process by Network Transmission, bag because network congestion (formation of network node is full) is dropped, wraps the fault of network lose or only owing to arrive receiving terminal and can't be included in the playback voice too late and be dropped.Too much frame losing meeting causes that voice quality seriously descends, for code decode algorithm commonly used among the VoIP such as G.711, G.723.1, G.729, because the frame-losing hide module that itself has makes it still to keep higher voice quality when lower frame loss rate, and is comparatively commonly used in VoIP.And for the ADPCM codec, though as G.726 having good compression function and lower disposal ability, be particularly suitable for transmitting signal tone (DTMF, single frequency tone, modulation signal etc.), had both G.711 and both advantages G.729, but owing to itself there is not the frame-losing hide function, having under the situation of frame losing, voice quality descends fast, have a strong impact on it and use on the net, can only be limited on the purposes such as storing forwarding.
At the research of ADPCM frame losing or bit error performance, various thinking was in the industry cycle proposed, relatively be typically the noise reduction method.The applying date is on January 20th, 1999, the patent No. is US6578162, the United States Patent (USP) that name is called " Error recovery method and apparatus for ADPCMencoded speech " provides a kind of error recovery method and equipment of ADPCM encoded voice, the core technology of this patent be eliminate since cause during error code " clickly " sound, its basic ideas are according to being to decide the signal for output to carry out denoising Processing according to whether current signal wrong.
Be illustrated in figure 2 as the principle schematic of this patent, this patent mainly is to be applied in PSTN or wireless transmission aspect, when signal comes in to obtain actual signal by demodulation, the reformatting module is by after the correction process, signal is input to wrong frame detector and inspects current data and whether have error code or frame losing, suppressing the window generator decides according to the situation of error code and produces the inhibition window that a size is n (according to wrong frame condition decision), nonlinear processor comes signal is carried out non-linear de-noising according to the size that suppresses window, purpose is the noise that elimination error code or frame losing cause, the decay adjuster is decayed for the output of nonlinear processor, and whether the effect of transducer is to make mistakes according to current demand signal to decide current output.
The emphasis of this scheme cause after for frame losing " clickly " noise suppresses, and do not compensate for the frame of having lost, and also carries out recovering state for decoder, can't really improve voice quality.And this scheme only depends on PSTN or wireless transmission, can not satisfy the transmission on IP network.
Summary of the invention
Technical problem to be solved by this invention is: overcome prior art based on the ADPCM code decode algorithm on IP network during transferring voice, can be because frame losing causes the shortcoming that voice quality descends fast, a kind of drop-frame processing device and method of ADPCM encoding and decoding are provided, reduce the caused voice quality of frame losing and descend, improve the transmission quality of voice.
The present invention solves the problems of the technologies described above the technical scheme that is adopted to be:
This drop-frame processing device based on the adaptive difference pulse code modulation, comprise the adaptive differential constant-delay discriminator, on the adaptive differential constant-delay discriminator, increase by an input, on this input, be provided with frame losing compensating module and state correction module, the frame losing compensating module is connected with the state correction module, the state correction module is connected with the adaptive differential constant-delay discriminator is two-way, output at the adaptive differential constant-delay discriminator is set up buffer module, and buffer module also is connected with described frame losing compensating module; Described frame losing compensating module is used for when frame losing occurring, carries out the frame losing compensation according to the voice that kept in the described buffer module, and the frame losing offset is imported described state correction module; Described state correction module is used for the status register value of combining adaptive differential pulse decoder, and the state of adaptive differential constant-delay discriminator is proofreaied and correct.
Also can connect one on the described frame losing compensating module and suppress the window generator, suppressing the window generator is connected with a noise reduction module, noise reduction module is positioned at the output of adaptive differential constant-delay discriminator, link to each other with buffer module with described adaptive differential constant-delay discriminator, described inhibition window generator produces an inhibition window according to the frame losing offset of frame losing compensating module output, controls the length of described noise reduction module processing signals.Described frame losing compensating module can be two-way the connection with buffer module, and the frame losing offset of frame losing compensating module output also directly outputs to described buffer module, preserves output by buffer module.Described state correction module adopts the adaptive difference pulse code device.
Corresponding a kind of frame losing processing method based on the adaptive difference pulse code modulation when the adaptive differential constant-delay discriminator has received frame, is directly decoded, and decoded data is carried out exporting behind the buffer memory; When frame losing occurring, the voice according to institute's buffer memory carry out the frame losing compensation earlier, utilize the status register value of frame losing offset and combining adaptive differential pulse decoder then, and the adaptive differential constant-delay discriminator is carried out state correction.
When carrying out the frame losing compensation, can adopt pitch period copy method to carry out the frame losing compensation according to the voice of institute's buffer memory.When carrying out state correction, status register value with described adaptive differential constant-delay discriminator copies in the adaptive difference pulse code device earlier, the adaptive difference pulse code device is encoded to described frame losing offset then, value with the status register of adaptive difference pulse code device copies the adaptive differential constant-delay discriminator to again, thereby the state of adaptive differential constant-delay discriminator is proofreaied and correct.
Can also carry out silence detection to described frame losing offset, according to the frame losing offset is that the quiet right and wrong of going back are quiet, produce an inhibition window that varies in size, and the output of adaptive differential constant-delay discriminator is carried out the noise reduction process of unlike signal length according to the size that suppresses window.Described frame losing offset can directly carry out exporting behind the buffer memory.Whether described inhibition window generator can detect the frame losing offset by a silence detection module is quiet, or is judged by the energy of the frame data before the frame losing, when energy during less than quiet threshold value, is judged as mute signal, otherwise is voice signal.
Beneficial effect of the present invention is: the present invention overcome prior art based on the ADPCM code decode algorithm on IP network during transferring voice, can be because frame losing causes the shortcoming that voice quality descends fast, a kind of drop-frame processing device and method of ADPCM code decode algorithm are provided, defective at the ADPCM codec, adopt the frame losing predictive compensation, means such as decoder states correction and adaptive noise reduction processing have been accelerated the recovery of adpcm decoder state, the Frame of losing is hidden, eliminate produced " clickly " noise, thereby solved when the ADPCM code decode algorithm transmits on IP network because the problem that the voice quality that frame losing causes descends has fast improved subjective speech quality.
Description of drawings
Fig. 1 is the transmission course schematic diagram of VoIP;
Fig. 2 is the principle schematic of US6578162 patent;
Fig. 3 is an ADPCM frame losing handling principle schematic diagram of the present invention;
Fig. 4 is a code synchronism correcting process schematic diagram of the present invention;
Fig. 5 suppresses window generator process chart for the present invention;
Fig. 6 is a frequency domain noise reduction process schematic diagram of the present invention.
Embodiment
With embodiment the present invention is described in further detail with reference to the accompanying drawings below:
Because the ADPCM codec is to belong to waveform coding, in decoding end if frame loss condition, except when beyond the preceding LOF, the recovery of decoder states is more than slow many of PCM encoder, need under the normal condition could recover normal after the 80ms, thereby the voice quality that makes it to cause when frame losing occurring descends fast, and bring very big " clickly " noise, this drawbacks limit the application of ADPCM codec in IP network.Under present network condition, frame losing is inevitable.PSTN or wireless frame losing have very big difference with IP network, the former wrong frame is often because the transmission line quality, and perhaps error code influence, the data sample random distribution of makeing mistakes appear in demodulation; And IP network is owing to shake, time delay, factor affecting such as congested, each minimumly loses a packet, and the data sample of losing is concentrated.Because the problem that the voice quality that frame losing causes descends fast will compensate for the Frame of losing at least, and prior art is just judged data and is transmitted and make mistakes and adopt denoising Processing when solving the ADPCM code decode algorithm and transmitting on IP network.
The present invention proposes a kind of drop-frame processing device and method of ADPCM code decode algorithm, defective at the ADPCM codec, adopt means such as frame losing predictive compensation, decoder states correction and adaptive noise reduction processing to accelerate the recovery of decoder states, the Frame of losing is hidden, eliminate produced " clickly " noise, thereby the problem that the voice quality that solution ADPCM code decode algorithm causes owing to frame losing when transmitting on IP network descends fast improves subjective speech quality.
Be illustrated in figure 3 as ADPCM frame losing handling principle schematic diagram of the present invention, the present invention is based on the drop-frame processing device of adaptive difference pulse code modulation, comprise adpcm decoder, on adpcm decoder, increase by an input, be provided with frame losing compensating module and state correction module on this input, the state correction module adopts an adpcm encoder.The frame losing compensating module is connected with the state correction module, and the state correction module is connected with adpcm decoder is two-way, sets up buffer module at the output of adpcm decoder, and buffer module also is connected with the frame losing compensating module.
Also be connected with one on the frame losing compensating module and suppress the window generator, suppressing the window generator is connected with a noise reduction module, noise reduction module is positioned at the output of adpcm decoder, suppress the frame losing offset generation inhibition window of window generator, control the length of noise reduction module processing signals according to the output of frame losing compensating module.The frame losing compensating module is two-way the connection with buffer module, and the frame losing offset of frame losing compensating module output also directly outputs to buffer module, preserves output by buffer module.
When adpcm decoder has received frame, directly decode, and decoded data is carried out exporting behind the buffer memory; When frame losing occurring, earlier the voice according to institute's buffer memory carry out frame losing compensation (can adopt pitch period copy method), utilize the frame losing offset then and in conjunction with the status register value of adpcm decoder, adpcm decoder are carried out state correction.By suppressing the window generator described frame losing offset is detected, whether according to the frame losing offset is that different sizes of quiet generation suppress window, it is little to suppress window when quiet, it is big to suppress window when not being quiet, and the output of adpcm decoder is carried out the noise reduction process of unlike signal length according to the size that suppresses window.
When on network, transmitting, whether exist frame loss condition to judge that by other functional module (whether good frame is received by receiver judges when IP wraps, when certain IP wraps in when not arriving buffering area in the stipulated time, receiver is judged as packet loss, thus the notice adpcm decoder).The detection module that in drop-frame processing device of the present invention, does not have frame loss condition, when having received frame, walk normal ADPCM decoding process, promptly handle by adpcm decoder, carry out preserving output after the noise reduction process, and the degree of noise reduction process and size are 0 by the size decision that suppresses window when suppressing the window initialization, at this moment suppress window and are actually and do not work.When frame losing occurring, carry out pitch analysis for the sampling point of preserving earlier, obtain the pitch period of latest data, thereby obtain the output of frame losing compensation, for the data of having lost, the data of the output place of lost of frame losing compensation, simultaneously, the state correction module utilizes the output of frame losing compensating module as input, carries out decoder states and proofreaies and correct, and makes it to approach the state after the frame losing.At this moment suppress window and just start, the size that suppresses window is by the serious situation of frame losing and whether the current speech data are quiet or voice decide.The effect of buffer module is to save the data in the buffering area, and as the input of frame losing compensation, time-delay is carried out exporting after the smoothing processing from buffer module at last.
Respectively frame losing compensating module, state correction module, inhibition window generator, noise reduction module are specifically described below:
1, frame losing compensating module
The operation principle of frame losing compensation is on the net behind the packet loss, and adpcm decoder compensates for the data of losing, and reaching good auditory effect, great majority are predicted the data of place of lost for the feature of utilizing the data before the packet loss.The method that the frame losing compensation realizes has a lot, comprising elimination method, quiet method of substitution, cladding method for making, template matching method, pitch period copy method etc.According to the raising of computation complexity, time delay and tonequality, pitch period copy method is the reasonable a kind of method of resultant effect, and pitch period copy method mainly is the pitch period clone method of recommending with reference to ITU_T.
The purpose of frame losing compensating module is the data of the current frame losing of compensation, and as the input of state correction module.Present embodiment frame losing compensating module adopts pitch period copy method to carry out the frame losing compensation, utilizes the sampling point that keeps in the buffer module to carry out the pitch period search, carries out the data copy according to its fundamental tone then and smoothly obtains output valve.
The frame lost in the pitch period copy method repeats the data of a pitch period of previous frame signal, and the seamlessly transitting between primary signal and the composite signal of having adopted " phase matched " technique guarantee simultaneously reduced the voice distortion that causes because of simple repeating data.Because voice signal is standard time series stably, especially voiced sound signal, has certain quasi periodicity, it is rational therefore adopting the approximate frame losing speech data of the preceding speech data of frame losing.In the ideal case, synthetic speech with lose voice and have identical time domain and frequency domain characteristic, can guarantee that synthetic speech is a nature.
Be provided with a fundamental tone buffering area in the frame losing compensating module, when carrying out frame-losing hide, up-to-date speech data in the above-mentioned buffer module is saved in the fundamental tone buffering area, is used to estimate the pitch period of current speech signal, pitch period is calculated by top described method.Nearest a pitch period and its 1/4 pitch period data are before taken out from the fundamental tone buffering area, be used for compensation the frame losing data.Wherein preceding 1/4 pitch period data be used for frame losing before voice signal carry out overlap-add (Overlap_Add), to guarantee seamlessly transitting between primary signal and compensating signal.Offset data simply repeats the pitch period data of taking out, up to whole covered an information frame till, if the next frame data are not lost, then continuation 1/4 pitch period data are carried out overlap-add with the data that accurately receive again, and purpose is to seamlessly transit in order to guarantee equally.Can adopt triangular window or Hanning window (how can adopt triangular window during realization) during overlap-add because it is more simpler than Hanning window.If next frame is still lost, then data of extracting a pitch period are used for compensation more, can extract 1 pitch period at most.Information frame is lost many more, and it is just big more that synthetic speech and actual speech differ.Therefore, except first frame data, the data of drop-out frame will have certain decay when compensation continuously, and the signal of compensation carries out linear attenuation with the speed of every frame 20%, if there are continuous 6 frame informations to lose, then the 6th vertical shading signal is 0 like this.
2, state correction module
According to the characteristics of ADPCM codec, the present invention utilizes encoder to carry out synchronous correction for decoder, is input as the output of frame losing compensation.For the decoder of ADPCM, wherein include encoder, during operate as normal, encoder is identical with the status register of decoder, and calculating process also is the same.Status register mainly refers to the parameter of each submodule in the coding and decoding device.Because frame losing, decoder is not decoded in 80 (frame data) data, perhaps decodes for the data of mistake, thereby causes the parameter of each submodule deviation to occur.
After the frame losing, except the data of being lost, the state of each register of decoder is bigger with the perfect condition difference, and this brings very big influence for the normal decoder of ADPCM after frame losing, during practical application, for quiet frame, after the frame losing, the recovery that decoder can be very fast is normal, for speech frame, often occur after the frame losing in the 80ms, the state value of adpcm decoder can't be restrained, and becomes the bottleneck of ADPCM frame-losing hide algorithm maximum.By actual emulation,, in most of the cases will accelerate the recovering state of decoder if find at this time to utilize encoder to correct synchronously. and the input of encoder should be the output valve of frame losing compensation.
State correction module of the present invention mainly is to have comprised an adpcm encoder.When carrying out state correction, status register value with adpcm decoder copies in the adpcm encoder earlier, adpcm encoder is encoded to the frame losing offset then, value with the status register of adpcm encoder copies adpcm decoder to again, thereby the state of adpcm decoder is proofreaied and correct.After increasing the encoder synchronous correction, the state of decoder can be restrained fast.Be illustrated in figure 4 as code synchronism correcting process schematic diagram, step is:
A, decoder copy the value of status register to encoder.
B, encoder are encoded the state value of correcting encoder device to the output of frame losing compensation.
The value of C, coder state register copies decoder to.
3, suppress the window generator
Statistical results show, in two personal comminication's processes, the time that everyone on average has 60% is to be in silent status, the discontinuity of the voice that this just often says.According to the actual emulation analysis, when quiet frame frame losing, the decoding dateout that is influenced seldom just can restrain normally in 20-30ms, and when speech frame, the Frame that is had influence on is a lot, often could restrain behind 80ms normally.So, need to judge whether current data are speech frame, whether are quiet or efficient voice according to current demand signal, suppress the window generator and produce an inhibition window not of uniform size, the inhibition window of efficient voice than quiet greatly.The effect that suppresses window is the length of noise reduction module processing signals after the decision frame losing, when frame losing compensation output be the efficient voice signal time, the length of noise reduction process signal is 80ms, is output as when quiet when frame losing compensates, the length of noise reduction process signal is 30ms.
Be illustrated in figure 5 as and suppress window generator process chart, suppress the output that is input as the frame losing compensation of window generator.Utilize the silence detection module, as the VAD detection module among the appendix II G.711, whether the data that detect current input are quiet frame.It realizes that substantially principle is to extract the speech parameter of present frame (full range energy, low frequency energy, zero-crossing rate, frequency spectrum distortion), compares with mean parameter then, result is relatively formed a plurality of boundary conditions judge.If be quiet frame, suppressing window output size is 30ms; If be speech frame, suppressing window output size is 80ms.Consider the actual treatment ability, the VAD detection module can be simplified to be judged the energy of the frame data before the frame losing, when energy during less than quiet threshold value, is judged as mute signal, otherwise is voice signal.
4, noise reduction module
The recovering state time that state correction and frame losing compensation has just been accelerated decoder to a certain extent, but do not solve fully frame losing caused " clickly ".So will carry out noise reduction process for output, noise reduction process can adopt frequency domain noise reduction method or nonlinear noise reduction processing etc., the size of process data block is by suppressing the window decision.As shown in Figure 6, as when adopting frequency domain noise reduction method, can adopt the FFT on 128 rank (is the fast discrete Fourier conversion, signal is turned to the calculating of frequency domain from time domain) and IFFT (be reverse fast discrete Fourier conversion, signal is turned to the calculating of time domain from frequency domain), once import 80 data, add 48 0 values, after signal is transformed into frequency domain through FFT, carry out the noise amplitude limit, the part signal that spectrum energy is not concentrated is decayed, then by behind the IFFT, noise section obtains decay, thereby reduces noise.
The invention provides a kind of drop-frame processing device and method of ADPCM code decode algorithm, defective at the ADPCM codec, adopt means such as frame losing predictive compensation, decoder states correction and adaptive noise reduction processing to accelerate the recovery of adpcm decoder state, the Frame of losing is hidden, eliminate produced " clickly " noise, thereby solved when the ADPCM code decode algorithm transmits on IP network because the problem that the voice quality that frame losing causes descends has fast improved subjective speech quality.
Those skilled in the art do not break away from essence of the present invention and spirit, can there be the various deformation scheme to realize the present invention, the above only is the preferable feasible embodiment of the present invention, be not so limit to interest field of the present invention, the equivalent structure that all utilizations specification of the present invention and accompanying drawing content are done changes, and all is contained within the interest field of the present invention.
Claims (10)
1, a kind of drop-frame processing device based on the adaptive difference pulse code modulation, it is characterized in that: comprise the adaptive differential constant-delay discriminator, on the adaptive differential constant-delay discriminator, increase by an input, on this input, be provided with frame losing compensating module and state correction module, the frame losing compensating module is connected with the state correction module, the state correction module is connected with the adaptive differential constant-delay discriminator is two-way, output at the adaptive differential constant-delay discriminator is set up buffer module, and buffer module also is connected with described frame losing compensating module;
Described frame losing compensating module is used for when frame losing occurring, carries out the frame losing compensation according to the voice that kept in the described buffer module, and the frame losing offset is imported described state correction module;
Described state correction module is used for the status register value of combining adaptive differential pulse decoder, and the state of adaptive differential constant-delay discriminator is proofreaied and correct.
2, the drop-frame processing device based on the adaptive difference pulse code modulation according to claim 1, it is characterized in that: also be connected with one on the described frame losing compensating module and suppress the window generator, suppressing the window generator is connected with a noise reduction module, noise reduction module is positioned at the output of adaptive differential constant-delay discriminator, link to each other with buffer module with described adaptive differential constant-delay discriminator, described inhibition window generator produces an inhibition window according to the frame losing offset of frame losing compensating module output, controls the length of described noise reduction module processing signals.
3, the drop-frame processing device based on the adaptive difference pulse code modulation according to claim 1 and 2, it is characterized in that: described frame losing compensating module is two-way the connection with buffer module, the frame losing offset of frame losing compensating module output also directly outputs to described buffer module, preserves output by buffer module.
4, the drop-frame processing device based on the adaptive difference pulse code modulation according to claim 3, it is characterized in that: described state correction module adopts the adaptive difference pulse code device.
5, a kind of frame losing processing method based on the adaptive difference pulse code modulation is characterized in that, may further comprise the steps:
When the adaptive differential constant-delay discriminator has received frame, directly decode, and decoded data is carried out exporting behind the buffer memory;
When frame losing occurring, the voice according to institute's buffer memory carry out the frame losing compensation earlier, utilize the status register value of frame losing offset and combining adaptive differential pulse decoder then, and the adaptive differential constant-delay discriminator is carried out state correction.
6, the frame losing processing method based on the adaptive difference pulse code modulation according to claim 5 is characterized in that: when carrying out the frame losing compensation, according to the voice of institute's buffer memory, adopt pitch period copy method to carry out the frame losing compensation.
7, the frame losing processing method based on the adaptive difference pulse code modulation according to claim 5, it is characterized in that: when carrying out state correction, status register value with described adaptive differential constant-delay discriminator copies in the adaptive difference pulse code device earlier, the adaptive difference pulse code device is encoded to described frame losing offset then, value with the status register of adaptive difference pulse code device copies the adaptive differential constant-delay discriminator to again, thereby the state of adaptive differential constant-delay discriminator is proofreaied and correct.
8, according to claim 5,6 or 7 described frame losing processing methods based on the adaptive difference pulse code modulation, it is characterized in that: also described frame losing offset is carried out silence detection, according to the frame losing offset is that the quiet right and wrong of going back are quiet, produce an inhibition window that varies in size, and the output of adaptive differential constant-delay discriminator is carried out the noise reduction process of unlike signal length according to the size that suppresses window.
9, the frame losing processing method based on the adaptive difference pulse code modulation according to claim 8, it is characterized in that: described frame losing offset directly carries out exporting behind the buffer memory.
10, the frame losing processing method based on the adaptive difference pulse code modulation according to claim 8, it is characterized in that: whether described inhibition window generator detects the frame losing offset by a silence detection module is quiet, or judge by the energy of the frame data before the frame losing, when energy during less than quiet threshold value, be judged as mute signal, otherwise be voice signal.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2005100338748A CN100505714C (en) | 2005-03-25 | 2005-03-25 | Drop-frame processing device and method based on ADPCM |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2005100338748A CN100505714C (en) | 2005-03-25 | 2005-03-25 | Drop-frame processing device and method based on ADPCM |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN1838651A CN1838651A (en) | 2006-09-27 |
| CN100505714C true CN100505714C (en) | 2009-06-24 |
Family
ID=37015906
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNB2005100338748A Active CN100505714C (en) | 2005-03-25 | 2005-03-25 | Drop-frame processing device and method based on ADPCM |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN100505714C (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101800608B (en) * | 2009-02-11 | 2014-03-05 | 晨星软件研发(深圳)有限公司 | Adaptive differential pulse-code modulation-demodulation system and method |
| EP3023983B1 (en) * | 2014-11-21 | 2017-10-18 | AKG Acoustics GmbH | Method of packet loss concealment in ADPCM codec and ADPCM decoder with PLC circuit |
| CN107919996A (en) * | 2016-10-10 | 2018-04-17 | 大唐移动通信设备有限公司 | A kind of data pack transmission method and equipment |
| CN111081264B (en) * | 2019-12-06 | 2022-03-29 | 北京明略软件系统有限公司 | Voice signal processing method, device, equipment and storage medium |
| CN111256589B (en) * | 2020-02-14 | 2020-12-01 | 长飞光纤光缆股份有限公司 | Image method size measurement compensation method and device based on frame loss detection |
-
2005
- 2005-03-25 CN CNB2005100338748A patent/CN100505714C/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN1838651A (en) | 2006-09-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US7286562B1 (en) | System and method for dynamically changing error algorithm redundancy levels | |
| US8428959B2 (en) | Audio packet loss concealment by transform interpolation | |
| US6990195B1 (en) | Voice and data exchange over a packet based network with resource management | |
| US7627467B2 (en) | Packet loss concealment for overlapped transform codecs | |
| US20070206645A1 (en) | Method of dynamically adapting the size of a jitter buffer | |
| CN106664161A (en) | System and method of redundancy based packet transmission error recovery | |
| US8898060B2 (en) | Source code adaption based on communication link quality and source coding delay | |
| JPS60116000A (en) | Voice encoding system | |
| US20170187635A1 (en) | System and method of jitter buffer management | |
| CN100505714C (en) | Drop-frame processing device and method based on ADPCM | |
| CN108696491B (en) | Audio data sending processing method and device and audio data receiving processing method and device | |
| US8787490B2 (en) | Transmitting data in a communication system | |
| US7450593B2 (en) | Clock difference compensation for a network | |
| WO2007091927A1 (en) | Variable frame offset coding | |
| WO2012072278A1 (en) | Source signal adaptive frame aggregation | |
| US20210125622A1 (en) | Digital Voice Packet Loss Concealment Using Deep Learning | |
| US7079498B2 (en) | Method, apparatus, and system for reducing memory requirements for echo cancellers | |
| CN101783142A (en) | Transcoding method, device and communication equipment | |
| WO2023071822A1 (en) | Packet loss protection method and apparatus, electronic device, and storage medium | |
| CN1338834A (en) | Low-speed speech encode method based on network protocol | |
| US20010012993A1 (en) | Coding method facilitating the reproduction as sound of digitized speech signals transmitted to a user terminal during a telephone call set up by transmitting packets, and equipment implementing the method | |
| US20040047369A1 (en) | Method and apparatus for using and combining sub-frame processing and adaptive jitter-buffers for improved voice quality in voice -over-packet networks | |
| US7929520B2 (en) | Method, system and apparatus for providing signal based packet loss concealment for memoryless codecs | |
| Skoglund et al. | Voice over IP: Speech transmission over packet networks | |
| US20040138878A1 (en) | Method for estimating a codec parameter |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |