CN101820542A - Video transmission method for effectively controlling unequal power consumption - Google Patents

Abstract

The invention discloses a video transmission method for effectively controlling unequal power consumption. The method comprises the following steps of: (1), video coding: coding a coding end information source through a video coder, and acquiring different priority code streams by starting data qualification of robust; (2), Turbo coding: performing joint source/channel coding by adopting a Turbo code, and coding the different priority code streams by using the Turbo code with the same code rate; (3), modulating: after the code streams are modulated with binary phase-shift keying (BPSK), sending the code streams by using a wireless channel; (4), demodulating: demodulating the code streams which are received by a decoding end from the channel and interfered by the noise by using the BPSK, and performing iterative Turbo code decoding on the demodulated code streams; (5), iterative Turbo code decoding: performing unequal decoding power allocation on the different priority code streams by using an optimized decoding power consumption method; and (6): video decoding: sending the code streams which are decoded by using the channel to a H.264 decoder to performing source decoding. By using the method, not only the balance of the video quality and the decoding power consumption can be obtained, but also the better quality of the video transmission can be obtained under the condition of finite power consumption.

Description

A kind of unequal power consumption is the video transmission method of control effectively

Technical field

The present invention relates to a kind of multimedia video communication technical field, particularly relate to the effectively video transmission method of control of a kind of unequal power consumption.

Background technology

Along with video is more and more universal in the application of portable business field, also just more and more higher to the requirement of low-power consumption Video transmission system in the portable use.So effectively power consumption control becomes the key technology of third generation Multimedia Mobile Communication Systems.The design power consumption is the communication system of control effectively, at first will understand the multimedia communications system power consumption and how to consume.In general, the power consumption of Video transmission system comprises calculating power consumption, transmission power consumption and plays power consumption etc.Wherein very most power consumption is to calculate power consumption and transmission power consumption.Calculate power consumption and be used for operation systems soft ware, Code And Decode, for example the calculating power consumption of video coding end comprises estimation, motion compensation, dct transform and quantification etc.And the transmission power consumption is used for transmitting and receiving video signals, considers mainly how coding side carries out video encoding design and just can make video be transferred to the base station with more effective power consumption, and influence in the video transmission process power Elements Of Expense (as apart from etc.).From the angle of hardware, the VLSI system design of low-power consumption and the circuit optimization of low-power consumption RF will significantly reduce the consumption of power.

Make a general survey of present domestic and international research, the power consumption control of wireless video is based on the power managed of optimization method, need know the accurate model of system, by target function systematic function is described, systematic function is reached under the optimal target, finish calculating the calculating of power consumption and transmission power consumption, the power consumption number of calculating is loaded in real time give the user then.Make in this way the precise control of power consumption and clear and definite meaning (optimization objective function) is arranged, but amount of calculation is very big, is not suitable for dynamic environment.When environmental parameter changed, the optimal solution of originally trying to achieve was just no longer suitable, need rebulid system model and suboptimization again, is reaching the minimum power of consumption under the best QoS prerequisite of video, and finishing in real time of these tasks is to be difficult to realize.

Existing optimised power consumption transmission method, the simplest power consumption control are exactly the complexity that reduces encoding and decoding at the information source end, study its fast algorithm from each module (as motion estimation/compensation, infra-frame prediction and model selection etc.) of encoding and decoding and reduce the calculating power consumption.In the whole video transmission system, existing optimised power consumption transmission plan is often concentrated the processing power consumption and transmission power consumption of considering information source, channel, is obtaining to make whole transmission system power consumption minimum under the acceptable video quality prerequisite.But, do not consider power consumption and the optimization level of information source data and the characteristics of channel code of decoding end in the existing optimised power consumption transmission plan, be not suitable for the application of the universality of power consumption effective video transmission, particularly require the less mobile multi-media service of power consumption (as video mobile phones and handheld device).

Summary of the invention

The objective of the invention is to overcome the deficiencies in the prior art, the effectively video transmission method of control of a kind of unequal power consumption is provided.This method can promptly, obtain better video transmission quality reaching the minimum power of consumption under the best QoS prerequisite of video under the power consumption condition of limited.

For achieving the above object, the technical solution adopted in the present invention is:

Above-mentioned a kind of unequal power consumption is the video transmission method of control effectively, and its step comprises as follows:

(1) video coding: the coding side information source is encoded through video encoder, and the data qualification of enabling robust obtains the code stream of different priorities;

(2) Turbo coding: adopt Turbo code to carry out joint source-channel coding, with the Turbo code coding of different priorities code stream with same code rate;

(3) modulation: code stream sends through wireless channel after the strong control of two-phase phase shift (BPSK) modulation;

(4) demodulation: decoding end receives the code stream that the is subjected to noise jamming BPSK demodulation of sending from above-mentioned channel;

(5) Turbo code iterative decoding: each code stream that inspection is accessed carries out N _iInferior Turbo code iterative decoding carries out unequal decoding power consumption with the decoding power consumption of optimizing to the different priorities code stream and distributes, and is specially:

The coding/decoding method of optimizing, in the acceptable distortion range of video, R _TTotal bit number that expression can get, D _TBe expressed as tolerable distortion.Set up the resource allocation computing formula of power consumption minimum, its formula is:

$\min P\left(R\right)=\min (\underset{i=0}{\overset{n-1}{\mathrm{\Σ}}}{r}_s{R}_s^i+\underset{i=0}{\overset{n-1}{\mathrm{\Σ}}}(N_i\mathrm{\γ}+\mathrm{\λ})(R_s^i+R_c^i))---\left(1\right)$

Subject?to

End-to-end distortion limited condition is: D (R)≤D _T(2)

The limited bandwidth condition is: R≤R _T(3)

The priority confined condition is: N _i〉=N _I+m(m=0,1 ... n-i) (4)

The concrete steps of obtaining the resource allocation algorithm of power consumption minimum at above-mentioned confined condition are:

(501) Bit Allocation in Discrete of channel self-adapting;

(502) unequal decoding power consumption is distributed.

(6) video decode: carry out source coding in the decoder delivering to H.264 through the code stream of channel-decoding.

The beneficial effect that the video transmission method that a kind of unequal power consumption of the present invention is effectively controlled has compared with prior art is: this method adopts priority coding/decoding method-UDPA algorithm not only to obtain the balance of video quality and decoding power consumption, and distribute (Equal Decoding Power Allocation with traditional equalization decoding power consumption, EDPA) algorithm is compared, when obtaining identical video quality, reduce power consumption, promptly under the power consumption condition of limited, obtained better video transmission quality.This method is in order to design effective UDPA algorithm, and the video flowing of different priority has been distributed identical channel redundancy degree, and this just determines code check of channel easily, is convenient to the distribution of decoding end power consumption.Then, under the prerequisite that satisfies the video quality requirement, adjust the decoding iterations of Turbo code according to the variation of channel and realize minimum decoding power consumption.

Description of drawings

Fig. 1 is the effectively FB(flow block) of the video transmission method of control of a kind of unequal power consumption of the present invention;

Fig. 2 is the ratio schematic diagram that the decoding power consumption of Turbo code accounts for total processing power consumption under the decoding iterations of different Turbo codes;

Fig. 3 is the schematic diagram of the decoding power consumption of EDPA algorithm under different source bits rates and Turbo code iterations;

Fig. 4 is the schematic diagram of the video reconstruction quality of EDPA algorithm under different source bits rates and Turbo code iterations;

Fig. 5 is the decode time contrast that Fig. 5 provides UDPA algorithm and EDPA algorithm, and wherein (a) is when the signal to noise ratio of channel is 0.4db, the schematic diagram of the decoding power consumption of UDPA algorithm and EDPA algorithm contrast; (b) be when the signal to noise ratio of channel is 0.6d b, the schematic diagram of the decoding power consumption of UDPA algorithm and EDPA algorithm contrast; (c) be when the signal to noise ratio of channel is 0.8db, the schematic diagram of the decoding power consumption of UDPA algorithm and EDPA algorithm contrast;

Fig. 6 is the decoding sequence PSNR effect schematic diagram relatively under the different schemes;

Fig. 7 is the visual effect contrast of the 95th frame reconstruction figure of foreman sequence, and wherein (a) is at E _b/ N ₀The former figure that does not have error code for the 95th frame of 0.4dbForeman sequence; (b) be at E _b/ N ₀For the 95th frame UDPA algorithm of 0.4db Foreman sequence is rebuild design sketch; (c) be at E _b/ N ₀For the 95th frame EDPA algorithm of 0.4db Foreman sequence is rebuild design sketch.

Embodiment

Below in conjunction with the accompanying drawing example embodiments of the invention are described in further detail.

Provide the effectively video transmission method principle explanation of control of a kind of unequal power consumption of the present invention below:

In video transmission, total distortion comprises source encoding distortion and Channel Transmission distortion end to end, and defining end to end, the total distortion function is:

$D\left(R\right)=\underset{i=0}{\overset{n-1}{\mathrm{\Σ}}}(D_s\left(R_s^i\right)+D_c(R_c^i,R_s^i,P_e^i))---\left(5\right)$

Wherein, n represents the data qualification number of video flowing, R _s ⁱThe bit rate of i data classification of expression information source, R _c ⁱBe to be added in i the channel redundancy bit rate in the data classification, P _e ⁱBe that i data classification is through the channel bit error ratio behind the channel-decoding, D _c(R _c ⁱ, R _s ⁱ, P _e ⁱ) be the chnnel coding distortion function of i data classification, D _s(R _s ⁱ) be the source encoding distortion function of i data classification.

Bit rate R total in the Video transmission system is:

$R=\underset{i=0}{\overset{n-1}{\mathrm{\Σ}}}(R_s^i+R_c^i)---\left(6\right)$

Use Turbo code as FEC, the error correcting capability of Turbo code can be weighed with BER, and BER is low more, and the error correcting capability of Turbo code is just good more; BER is high more, and the error correcting capability of Turbo code is just poor more.In the whole video transmission system, the BER that the Turbo code decoding produces directly has influence on the quality of video transmission, and the BER of Turbo code is closely related with the iterations of Turbo code decoding.In order to oversimplify, consider that the BER of Turbo code can reduce along with the increase of Turbo code iterations, therefore the BER function P of Turbo code under the identical situation of other parameter (as the SNR of coding multinomial, block length, weaving length and channel) _eBe expressed as the function of the iterations of Turbo code approx by the method for curve fit, promptly

$P_e=\frac{\mathrm{\α}}{e^{\mathrm{\βN}}}(N\≤N_{\max })---\left(7\right)$

Wherein, N is the iterations of Turbo code, and it is the value (P when arriving certain value because of N of certain limit _eChanging can be very not big, does not then satisfy following formula).The value of α and β depends on other parameters such as the code check, SNR of channel, uses the nonlinear parameter method of estimation, just can access the value of parameter alpha and β.

Suppose that the bit error code mistake is independently, then the packet loss of k bag is:

${\mathrm{\ρ}}^k=1-{(1-P_e)}^{B_k}---\left(8\right)$

B _kFor the size of every bag, represent with bit.P _eIt is bit error rate.Can get formula (7) substitution formula (8)

${\mathrm{\ρ}}^k=1-{(1-\frac{\mathrm{\α}}{e^{\mathrm{\βN}}})}^{B_k}---\left(9\right)$

By formula (9) as can be known, packet loss ρ ^kRelevant with the size of the source encoding parameter of the iterations N of Turbo code, corresponding bag, packing and channel parameter (as the code check of channel, signal to noise ratio snr etc.).

The present invention represents the quality of video with the expected distortion of receiving terminal, and packet loss is not only depended in the distortion of primitive frame and received frame, and depends on the coding parameter (as quantizing) of information source.For packet loss, as top formula (9), the iterations that can adjust Turbo code is controlled the height of packet loss.

Suppose that transmitting terminal only knows the packet loss of receiving terminal, allow E[D ^k] k expected distortion that wraps of expression receiving terminal, suppose that the packet loss of k bag is ρ ^k, then the expected distortion function of k bag is:

$E\left[D^k\right]=(1-{\mathrm{\ρ}}^k)E\left(D_R^k\right)+{\mathrm{\ρ}}^{k}E\left(D_L^k\right)---\left(10\right)$

Wherein, E (D _R ^k) and E (D _L ^k) represent that respectively k is wrapped the expected distortion function that normal solution receives and loses, ρ ^kIt is the packet loss of k bag.If bag is correctly received expected distortion function E (D _R ^k) value depend on source encoding parameter (as quantizing distortion), fix as the coding parameter of information source, then E (D _R ^k) value also fix; If bag has been lost expected distortion function E (D _L ^k) value depend on the code error shielding strategy of decoding, the simplest method for shielding error code is exactly directly to duplicate a decode frame data to carry out code error shielding.

By formula (5), (6) and (9), then (10) formula can be written as:

$E\left(D\left(R\right)\right)=\underset{i=0}{\overset{n-1}{\mathrm{\Σ}}}\underset{k=1}{\overset{K_i}{\mathrm{\Σ}}}E\left(D(R_s^{\mathrm{ik}},R_c^{\mathrm{ik}},N_{\mathrm{ik}})\right)$ (11)

$=\underset{i=0}{\overset{n-1}{\mathrm{\Σ}}}\underset{k=1}{\overset{K_i}{\mathrm{\Σ}}}((1-{\mathrm{\ρ}}^{\mathrm{ik}})E\left(D_R^{\mathrm{ik}}\right)+{\mathrm{\ρ}}^{\mathrm{ik}}E\left(D_L^{\mathrm{ik}}\right))$

Wherein, K _iThe number of representing bag in i the data classification, N _IkThe iterations of representing the Turbo code of k bag in i the data classification, E (D _R ^Ik) and E (D _L ^Ik) represent that respectively k is wrapped the expected distortion function that normal solution receives and loses, ρ in i the data classification ^IkThe packet loss of then representing k bag in i the data classification.

In wireless video transmission system, total power consumption comprises handles power consumption and transmission power consumption.The present invention only considers the processing power consumption of receiver, and total processing power consumption of receiver then comprises source coding power consumption and channel-decoding power consumption, can be expressed as:

P(R)＝P _ds(R _s)+P _dc(R _s，R _c，N)????????????????????????????(12)

Wherein, P _Ds(R _s) be the source coding power consumption, it is information source code check R _sFunction; P _Dc(R _s, R _c, N) being channel-decoding power consumption (the decoding power consumption of Turbo code), it is source encoding code check R _s, Turbo code encoder bit rate R _cAnd the function of Turbo code decoding number of iterations N.

The decoding power consumption table of information source is shown:

$P_{\mathrm{ds}}\left(R_s\right)=\underset{i=0}{\overset{n-1}{\mathrm{\Σ}}}{r}_s{R}_s^i---\left(13\right)$

Wherein, r _sThe power consumption of the every bit consumption of expression source coding, it depends on the selection of each decoder module algorithm of decoding end.

The decoding power consumption of Turbo code can be expressed as:

$P_{\mathrm{dc}}(R_s,R_c,N)=\underset{i=0}{\overset{n-1}{\mathrm{\Σ}}}{r}_c(R_s^i+R_c^i)$ (14)

$=\underset{i=0}{\overset{n-1}{\mathrm{\Σ}}}(N_i\mathrm{\γ}+\mathrm{\λ})(R_s^i+R_c^i)$

Wherein, r _cThe decode power consumption of every bit consumption of expression Turbo code, N _iBe the iterations of the Turbo code decoding of i data classification, parameter γ and λ depend on the algorithm of Turbo code decoding, the signal to noise ratio (snr) and the channel model of channel.

Therefore, formula (12) can be written as

$P\left(R\right)=\underset{i=0}{\overset{n-1}{\mathrm{\Σ}}}{r}_s{R}_s^i+\underset{i=0}{\overset{n-1}{\mathrm{\Σ}}}(N_i\mathrm{\γ}+\mathrm{\λ})(R_s^i+R_c^i)---\left(15\right)$

As shown in Figure 1, a kind of unequal power consumption of the present invention is the video transmission method of control effectively, and its step comprises as follows:

(1) video coding: the coding side information source is encoded through video encoder, and the data qualification of enabling robust obtains the code stream of different priorities;

(2) Turbo coding: adopt Turbo code to carry out joint source-channel coding, with the Turbo code coding of different priorities code stream with same code rate;

(3) modulation: code stream sends through wireless channel after the strong control of two-phase phase shift (BPSK) modulation;

(4) demodulation: decoding end receives the code stream that the is subjected to noise jamming BPSK demodulation of sending from above-mentioned channel;

(5) Turbo code iterative decoding: each code stream that inspection is accessed carries out N _iInferior Turbo code iterative decoding carries out unequal decoding power consumption with the decoding power consumption of optimizing to the different priorities code stream and distributes, and is specially:

The coding/decoding method of optimizing, in the acceptable distortion range of video, R _TTotal bit number that expression can get, D _TBe expressed as tolerable distortion.Set up the resource allocation computing formula of power consumption minimum, its formula is:

$\min P\left(R\right)=\min (\underset{i=0}{\overset{n-1}{\mathrm{\Σ}}}{r}_s{R}_s^i+\underset{i=0}{\overset{n-1}{\mathrm{\Σ}}}(N_i\mathrm{\γ}+\mathrm{\λ})(R_s^i+R_c^i))---\left(1\right)$

Subject?to

End-to-end distortion limited condition is: D (R)≤D _T(2)

The limited bandwidth condition is: R≤R _T(3)

The priority confined condition is: N _i〉=N _I+m(m=0,1 ... n-i) (4)

The concrete steps of obtaining the resource allocation algorithm of power consumption minimum at above-mentioned confined condition are:

(501) Bit Allocation in Discrete of channel self-adapting:

If R _sThe code check of expression current information source, R _TRepresent the current transmission bandwidth that gets;

Distributed identical channel redundancy degree at the data qualification of different priorities, the code check of channel is R _c(R _c=R _T-R _s);

The channel redundancy degree of each data qualification is R _c/ n, wherein, n is the number of data qualification;

(502) unequal decoding power consumption is distributed:

The A initialization.

Priority classification is carried out in classification to data, is respectively 1,2 ... i ... n, the more little priority of i is high more;

The Turbo code iterations that each data qualification is selected is N ₁, N ₂... N _iN _n, i is more little, N _iBig more;

Setting maximum Turbo code iterations is: N _Max=max (N _i);

If minimum power consumption initial value is: P _Min=P (R, max (N _i));

If the iterations initial value is: N ₀=N _Max

The tolerable maximum distortion of setting video is D _T

The signal to noise ratio (snr) of Initial Channel Assignment;

The iterative algorithm that B optimizes

When each data qualification is selected different iterationses:

B1 calculates the packet loss of each data qualification:

${\mathrm{\ρ}}^k=1-{(1-\frac{\mathrm{\α}}{e^{\mathrm{\βN}}})}^{B_k}---\left(16\right)$

B2 calculates the message source and channel total distortion:

$E\left(D\left(R\right)\right)=\underset{i=0}{\overset{n-1}{\mathrm{\Σ}}}\underset{k=1}{\overset{K_i}{\mathrm{\Σ}}}E\left(D(R_s^{\mathrm{ik}},R_c^{\mathrm{ik}},N_{\mathrm{ik}})\right)$ (17)

$=\underset{i=0}{\overset{n-1}{\mathrm{\Σ}}}\underset{k=1}{\overset{K_i}{\mathrm{\Σ}}}((1-{\mathrm{\ρ}}^{\mathrm{ik}})E\left(D_R^{\mathrm{ik}}\right)+{\mathrm{\ρ}}^{\mathrm{ik}}E\left(D_L^{\mathrm{ik}}\right))$

B3 calculates power consumption: work as D _{1 ..i..n}≤ D _TThe time, calculate decoding power consumption P _1...i...n

$P\left(R\right)=\underset{i=0}{\overset{n-1}{\mathrm{\Σ}}}{r}_s{R}_s^i+\underset{i=0}{\overset{n-1}{\mathrm{\Σ}}}(N_i\mathrm{\γ}+\mathrm{\λ})(R_s^i+R_c^i)---\left(18\right)$

The selection of the optimum power consumption of B4: if P _1...i...n＜P _MinThen selecting current power consumption is minimum power consumption, otherwise minimum power consumption is an initial value;

And then, repeat the B1-B4 step, up to finding optimum P for the different number of iterations of each data qualification selection _Min

Table 1 be when the Turbo code maximum iteration time be 16, data qualification is 3 o'clock, example of the false code of unequal decoding power consumption allocation algorithm.

The false code special case of the unequal decoding power consumption of table 1 allocation algorithm

(6) carry out source coding delivering to H.264 in the decoder through the code stream of channel-decoding.

The UDPA algorithm of optimizing above is because the iteration of Turbo decoding may be brought very big delay, and in actual applications, information source and Turbo code encoding and decoding adopt hardware designs to finish real-time realization.

Experiment simulation environment of the present invention: encoding and decoding are adopted is reference software JM8.6 H.264/AVC, cycle tests adopts the foreman of Qcif form, sequence length is 100 frames, frame per second is made as 30fps, coding bit rate is 200kbps, the Turbo code of 1/3 code check, Turbo code coding/decoding method are log-MAP.

The processing power consumption of receiver comprises source coding power consumption and channel-decoding power consumption, and as shown in Figure 2, Turbo code decoding power consumption is occupied the overwhelming majority of total decoding power consumption.Turbo code that the ordinate of Fig. 2 is represented decoding power consumption (decode time of Turbo code) shared percentage in total decoding power consumption (decode time of Turbo code and source coding time sum), i.e. Turbo code decoding power consumption and and the ratio of total decoding power consumption.As can be seen from Figure 2, (Number ofIteration NOI) is at 7 o'clock, and the decoding power consumption of Turbo code accounts for 85% of total decoding power consumption at the decoding iterations of Turbo code.Can show that thus the decoding power consumption of Turbo code is to influence the key factor that receiver is always handled power consumption.And the decoding NOI of Turbo code is the principal element that influences Turbo code decoding power consumption, therefore, is supposing under the constant situation of source coding power consumption that the power consumption of the receiver that the present invention can obtain to optimize by the decoding NOI that adjusts Turbo code is distributed.

Carry out a large amount of emulation experiments below and assess the performance of UDPA algorithm proposed by the invention.Experiment porch is that H.264 video flowing transmits on awgn channel.

Encoding and decoding are adopted in the emulation experiment is reference software JM8.6 H.264/AVC, select QCIF (176 * 144) cycle tests " Foreman " for use, sequence length is 100 frames, frame per second is made as 30fps, maximum reference frame number is 5, maximum motion search range is made as ± 16 pixels, the structure of coding is " IPPPP ... "Emulation experiment adopts data qualification H.264, and video flowing is divided into A, B, three streams of C, and each Slice selects 200bytes for use.The encoder bit rate that test is adopted has 50kbps, 100kbps, 150kbps and 200kbps.

At the channel end, what select for use is two dimension [15,13] the Turbo code encoder with two RSC, and frame length is 4550 bits, and coding/decoding method is log-MAP.In order to reduce the average iterations of Turbo code, use the CRC coding method in the emulation experiment of the present invention and determine whether stopping iteration (the hard decision result to decoder output after each iteration carries out CRC check.If CRC check is all correct, then stops iteration, otherwise continue the next round iteration).In test, the maximum iteration time of Turbo code elects 16 as, even because higher iterations, the change of Turbo code performance neither be very big.Table 2 has been listed formula (7) and has been selected about the parameter (α, β) of Turbo code bit error rate function under different Eb/N0, obtains by curve-fitting method.In emulation of the present invention, suppose that bandwidth is known, the code check of Turbo code is 1/3, and in actual applications, used Turbo code code check is that the bandwidth by channel decides.

In emulation experiment, adopt the white noise random error mode of Gaussian Profile, error rate scope is from 1e ^-6To 1e ^-2, weigh the power consumption of decoding with decode time, emulation is to move on the PC of P4/2.4G/512, platform is Windows.

The parameter of table 2Turbo sign indicating number bit error rate is selected

??E B/N0	??α	??β
			??0.1	??0.1309	??0.1024
??0.2	??0.1463	??0.213
			??0.3	??0.1791	??0.4011
??0.4	??0.2034	??0.5834
			??0.5	??0.2183	??0.7367

??E B/N0	??α	??β
			??0.6	??0.2373	??0.9041
??0.7	??0.2558	??1.068
			??0.8	??0.2747	??1.241
??0.9	??0.2908	??1.407
			??1.0	??0.3076	??1.576

Be the validity verify that unequal decoding power consumption distributes (UDPA) algorithm and impartial decoding power consumption to distribute (EDPA) algorithm, Design of Simulation under the identical information source distortion factor and channel signal to noise ratio condition two kinds of algorithms:

1) UDPA algorithm: unite the unequal decoding power consumption allocation algorithm of considering message source and channel

2) EDPA algorithm: unite the equalization decoding power consumption allocation algorithm of considering message source and channel.

At first equalization decoding power consumption is distributed in the experiment and carried out emulation, channel E _b/ N ₀Be 0.4db.Under different source encoding code checks, what Fig. 3 represented is: under different bit rates, and the decoding power consumption of Turbo code under different iterationses, wherein ordinate is represented the power consumption of decoding, abscissa is represented the Turbo code iterations.What Fig. 4 represented is: corresponding video reconstruction quality, and its ordinate is represented the PSNR that video is rebuild, abscissa is represented the Turbo code iterations.From Fig. 3 and Fig. 4 as can be seen, under identical source bits rate, the Turbo code iterations is high more, and the decoding power consumption that is consumed is just big more, and in certain iteration scope, the decoding power consumption reaches capacity; Equally, the Turbo code iterations is high more, the PSNR value that video is rebuild is just high more, be that quality is just good more, under certain degree, video quality also reaches capacity, and is because when the Turbo code iterations reaches certain value, the bit error rate of channel is zero, only is that quantizing distortion exists for video.In addition, it can also be seen that from figure that the encoder bit rate of information source is high more, it is just high more to finish the required decoding power consumption of iteration, and the reconstruction quality of video is just high more, and promptly the source encoding code check is big more, and the decoding power consumed is just many more.

Then, emulation experiment has compared the performance of UDPA algorithm and EDPA algorithm, at different channel E _b/ N ₀(0.4,0.6,0.8) is tested.Fig. 5 provides the decode time contrast of two kinds of algorithms, as seen from the figure, under identical video quality (identical PSNR), the UDPA algorithm has been saved many power consumptions than EDPA algorithm, be because the UDPA algorithm that proposes has very big flexibility, the video flowing of different priorities carried out unequal decoding power consumption distributed.Simultaneously, comparison diagram 5 (a), Fig. 5 (b) and Fig. 5 (c), as can be seen from the figure: along with E _b/ N ₀Increase, channel bit error rate descends, under identical video quality (identical PSNR), the power consumption of decoding reduces.When low signal-to-noise ratio, the advantage that reduces power consumption in the UDPA algorithm of the present invention can be more obvious.For example, when Y-PSNR was approximately 26dB, Fig. 5 (a) UDPA algorithm had been saved about 11% power consumption than EDPA algorithm; Fig. 5 (b) UDPA algorithm has been saved about 9% power consumption than EDPA algorithm; Fig. 5 (c) UDPA algorithm has been saved about 8% power consumption than EDPA algorithm.

Then, Fig. 6 has provided above-mentioned two kinds of algorithms at E _b/ N ₀=0.4 and the source encoding code check be under the situation of 100kbps, and under the situation of decoding power consumption close (decode time is approximately 80s), the comparing result of the decoded picture PSNR of the every frame of Forman sequence, wherein the mean P SNR value of UDPA algorithm is 29.1db, and the mean P SNR value of EDPA algorithm is 26.7db.Obviously, under close decoding power consumption, adopt the video reconstruction quality of the UDPA algorithm of the present invention's proposition to be greatly improved, can reach 2-6dB than average signal-to-noise ratio gain based on the EDPA algorithm.The picture quality of UDPA algorithm changes milder, and EDPA algorithm picture quality fluctuation ratio is bigger.This is because the UDPA algorithm is very effective to the opposing error code to the protection of the more power consumptions of significant data, has improved the reconstruction quality of video.In addition the coding that adopts of the present invention structure is " IPPPP ... ", the quality of every two field picture not only is subjected to the influence of channel error code, and extended influence by the error code of previous frame.Particularly the P frame is vulnerable to the influence from front reference frame error propagation, so have lower PSNR value.

At last, Fig. 7 (a), Fig. 7 (b), Fig. 7 (c) have provided the visual effect contrast of the 95th frame reconstruction figure of foreman sequence.Wherein, Fig. 7 (a) is the former figure that does not have error code; Fig. 7 (b) is the reconstruction design sketch of UDPA algorithm; Fig. 7 (c) is the reconstruction design sketch of EDPA algorithm.Fig. 7 (b), Fig. 7 (c) compare with Fig. 7 (a), distortion has taken place in the many places among the figure, but the whole visual effect of Fig. 7 (b) is significantly better than Fig. 7 (c), this has further illustrated that from subjective angle the UDPA algorithm that the present invention proposes is better than the EDPA algorithm, can obtain better video reconstruction quality under identical power consumption.

Claims (3)

1. the effective video transmission method of control of a unequal power consumption is characterized in that the step of this method is as follows:

(1) video coding: the coding side information source is encoded through video encoder, and the data qualification of enabling robust obtains the code stream of different priorities;

(2) Turbo coding: adopt Turbo code to carry out joint source-channel coding, with the Turbo code coding of different priorities code stream with same code rate;

(3) modulation: code stream sends through wireless channel after the strong control modulation of two-phase phase shift;

(4) demodulation: decoding end receives the code stream that the is subjected to noise jamming BPSK demodulation of sending from above-mentioned channel;

(5) Turbo code iterative decoding: each code stream that inspection is accessed carries out N _iInferior Turbo code iterative decoding carries out unequal decoding power consumption with the decoding power consumption of optimizing to the different priorities code stream and distributes, and is specially:

The coding/decoding method of optimizing, in the acceptable distortion range of video, R _TTotal bit number that expression can get, D _TBe expressed as tolerable distortion.Set up the resource allocation computing formula of power consumption minimum, its formula is:

$\min P\left(R\right)=\min (\underset{i=0}{\overset{n-1}{\mathrm{\Σ}}}{r}_s{R}_s^i+\underset{i=0}{\overset{n-1}{\mathrm{\Σ}}}(N_i\mathrm{\γ}+\mathrm{\λ})(R_s^i+R_c^i))---\left(1\right)$

Subject?to

End-to-end distortion limited condition is: D (R)≤D _T(2)

The limited bandwidth condition is: R≤R _T(3)

The priority confined condition is: N _i〉=N _I+m(m=0,1 ... n-i) (4)

The concrete steps of obtaining the resource allocation algorithm of power consumption minimum at above-mentioned confined condition are:

(501) Bit Allocation in Discrete of channel self-adapting;

(502) unequal decoding power consumption is distributed.

(6) video decode: carry out source coding in the decoder delivering to H.264 through the code stream of channel-decoding.

2. a kind of unequal power consumption according to claim 1 is the video transmission method of control effectively, it is characterized in that, and the Bit Allocation in Discrete of the described channel self-adapting of step (501) in the above-mentioned steps (5), it is specific as follows:

If R _sThe code check of expression current information source, R _TRepresent the current transmission bandwidth that gets;

Distributed identical channel redundancy degree at the data qualification of different priorities, the code check of channel is: R _c(R _c=R _T-R _s);

The channel redundancy degree of each data qualification is: R _c/ n, wherein, n is the number of data qualification.

3. a kind of unequal power consumption according to claim 1 is the video transmission method of control effectively, it is characterized in that, the described unequal decoding power consumption of step (502) is distributed in the above-mentioned steps (5), and concrete steps are as follows:

The A initialization: priority classification is carried out in classification to data, is respectively 1,2 ... i ... n, the more little priority of i is high more; The Turbo code iterations that each data qualification is selected is N ₁, N ₂... N _iN _n, i is more little, N _iBig more; Setting maximum Turbo code iterations is: N _Max=max (N _i); If minimum power consumption initial value is: P _Min=P (R, max (N _i)); If the iterations initial value is: N ₀=N _MaxThe tolerable maximum distortion of setting video is D _TThe signal to noise ratio (snr) of Initial Channel Assignment;

The iterative algorithm that B optimizes: when each data qualification is selected different iterationses:

B1, calculate the packet loss of each data qualification:

${\mathrm{\ρ}}^k=1-{(1-\frac{\mathrm{\α}}{e^{\mathrm{\βN}}})}^{B_k}---\left(16\right)$

B2, calculating message source and channel total distortion:

$E\left(D\left(R\right)\right)=\underset{i=0}{\overset{n-1}{\mathrm{\Σ}}}\underset{k=1}{\overset{K_i}{\mathrm{\Σ}}}E\left(D(R_s^{\mathrm{ik}},R_c^{\mathrm{ik}},N_{\mathrm{ik}})\right)$

(17)

$=\underset{i=0}{\overset{n-1}{\mathrm{\Σ}}}\underset{k=1}{\overset{K_i}{\mathrm{\Σ}}}((1-{\mathrm{\ρ}}^{\mathrm{ik}})E\left(D_R^{\mathrm{ik}}\right)+{\mathrm{\ρ}}^{\mathrm{ik}}E\left(D_L^{\mathrm{ik}}\right))$

B3, calculating power consumption: work as D _{1 ..i..n}≤ D _TThe time, calculate decoding power consumption P _1...i...n

$P\left(R\right)=\underset{i=0}{\overset{n-1}{\mathrm{\Σ}}}{r}_s{R}_s^i+\underset{i=0}{\overset{n-1}{\mathrm{\Σ}}}(N_i\mathrm{\γ}+\mathrm{\λ})(R_s^i+R_c^i)---\left(18\right)$

The selection of B4, optimum power consumption: if P _1...i...n＜P _MinThen selecting current power consumption is minimum power consumption, otherwise minimum power consumption is an initial value;

And then, repeat the B1-B4 step, up to finding optimum P for the different number of iterations of each data qualification selection _Min

Images