CN101268936A - Electrocardio-compression method and decrypting method of wireless cardiogram monitor - Google Patents

Abstract

The invention belongs to the biomedical engineering and computer field, in particular relates to a method that carries through embedded zero-tree coding to realize the high-efficient compression of ECG data on the basis of lifting the optimal space of wavelet packet. The method carries through the parity sequence decomposition to ECG signals obtained by pretreatment and obtains a first-order general picture and a first-order detail node by the lifting algorithm; the method analyzes whether the first-order father node needs to be decomposed into child nodes according to the cost function and decomposes the first-order father node that needs to be divided into second-order child nodes by the lifting algorithm, in the same way, till the lifting wavelet packet space structure of the lowest information entropy is obtained; finally, the embedded zero-tree coding algorithm is carried out so as to realize the high-efficient compression to the ECG data according to the low redundancy mapping relation on the basis of optimally lifting the wavelet packet. The method aims at the system characteristics of the long-distance ECG monitoring terminal and realizes the high-efficient compression algorithm of the ECG data with transferable embedded monitoring terminal, thereby solving the problems of the data storage and the real-time transmission in the long-distance ECG monitoring.

Description

The electrocardio compression method and the coding/decoding method of radio electrocardiographicmonitoring monitoring instrument

Technical field

The invention belongs to the biomedical engineering technology field, be specifically related to the electrocardio compression method and the decompression method of radio electrocardiographicmonitoring monitoring instrument.

Background technology

In recent years, along with the increase of people's life and work pressure and the aging development of society, the sickness rate of all kinds of heart diseases is ascendant trend year by year.Aspectant diagnostic mode can not satisfy people's great demand growing to health care between traditional patient and the doctor, and to being centroclinal with family, people also pay attention to prevention of disease to medical system more gradually.These change directly orders about cardiac monitoring and develops into remote electrocardiogram monitor based on mobile communications network from the non-online monitoring of non real-time, and Wearable radio electrocardiographicmonitoring monitoring instrument arises at the historic moment thus.

Wearable radio electrocardiographicmonitoring monitoring instrument is as the diagnosis monitoring terminal of tele-medicine and mobile household health care system, not only need to finish the record and the storage of a large amount of electrocardiogram (ECG) datas, and, in order to realize that the patient is guarded timely and effectively, or even 24 hours all-weather monitors, more need to possess the real-time Transmission ability of electrocardiogram (ECG) data.Therefore, in order to achieve the above object, must carry out the efficient compression of electrocardiogram (ECG) data at monitoring terminal: on the one hand by improving data compression ratio, strengthen the data storage capacities of electrocardiogram monitor, and reduce transmission needed data volume during equal information with this, promote the real-time Transmission ability of electrocardiogram (ECG) data; On the other hand,, shorten consuming time that the electrocardiogram (ECG) data compression needs greatly, further promote the transfer of data real-time of monitoring terminal by improving the execution efficient of compression algorithm.This shows, realize being applicable to the efficient data compression algorithm of cardiac monitoring terminal, is the key technology that realizes the remote real-time electrocardio monitoring.

At present the electrocardiogram (ECG) data compression algorithm mainly is divided into two big classes, and a class is the time domain compression algorithm, and wherein using maximum is segmented fitting.This algorithm is carried out the efficient height, and compression ratio is big, is fit to be transplanted to the monitoring terminal of system resource anxiety, but there is the problem that can't overcome in this method, and very big distorted signals is promptly arranged behind the decompress(ion).Another kind of is the transform domain compression algorithm, mainly comprises KL (Karhunen-Loeve) conversion, discrete cosine transform and traditional wavelet.Wherein KL conversion and discrete cosine transform have good Static Compression Performance, early be introduced in the electrocardio compression, but because these two kinds of algorithms can not partly be analysed and the layering processing in make-game, can't transmit data progressively to appear mode in one's mind in real time, influence data transmission efficiency greatly.And on the other hand, the complexity degree height of these two kinds of algorithms requires to take more system resources, therefore seldom is used for the electrocardiogram (ECG) data compression of monitoring terminal.Relative, traditional wavelet then because of its time, frequency domain has good localization property simultaneously, can adopt sub-band, level coding technology to realize that the progression transfer encoding to solve transmission real-time problem, has produced wavelet field EZW algorithm thus.This method has been successfully applied among the compression of images JPEG2000 by means of the time frequency analysis advantage of wavelet transformation, and aspect compress ecg data, the application of this algorithm has also obtained certain progress, but up to the present, this algorithm is mainly used in the abundant service end electrocardio storage backup system of system resource, if be applied to guard the electrocardio compression of end, this algorithm must solve the two large problems that self exists: at first, wavelet field EZW algorithm exists compression result that big redundant bottleneck problem is arranged: see that from information theory view data compression essence is to extract the order of signal, remove redundancy, that is to say the process that reduces comentropy.In theory of information, comentropy has reflected the unordered degree of signal, and comentropy is big more, and signal is unordered more.Because traditional wavelet field EZW adopts the fixed frequency band is olation to tend to cause the generation of high comentropy isolated zero in the encoding stream, this will inevitably make, and comentropy can't reduce fully, compression result still has big redundant.Secondly, wavelet field EZW algorithm still exists higher algorithm complexity and resources occupation rate: adopt floating number to calculate in the algorithmic procedure, not only take ample resources, and it is low to carry out efficient; Still there are a lot of redundant steps in the algorithmic procedure,, then two extract half redundant data of throwing away on each yardstick, wasted memory headroom in a large number, and increased operand such as taking to do earlier each change of scale in the algorithm.

Summary of the invention

Purport of the present invention provides a kind of cardiac monitoring terminal that is suitable for and carries out electrocardiogram (ECG) data compression and coding/decoding method, promotes the data storage and the real-time Transmission ability of remote electrocardiogram monitor terminal with this.

For this reason, the present invention adopts following technical scheme:

A kind of electrocardio compression method of radio electrocardiographicmonitoring monitoring instrument comprises the following steps:

(1) electrocardiosignal that collects is carried out pretreatment;

(2) electrocardiosignal that pretreatment is obtained is carried out the sequence of parity division, then obtains single order detail signal node and single order general picture signal node by boosting algorithm;

(3) judge according to cost function whether needs are decomposed into child node for single order details and single order general picture father node, if desired, then it is carried out the sequence of parity division, then obtain corresponding second order child node by boosting algorithm;

(4) judge whether have needs to continue to be decomposed into corresponding three rank child nodes in the second order child node according to cost function, adopt boosting algorithm that these second order child nodes are decomposed, by that analogy, up to the optimal wavelet bag space structure that obtains the minimum information entropy;

(5) traversal wavelet packet all coefficients are obtained the maximum of coefficient absolute value, keep the highest order of this numerical value binary digit, with all the other low level zero setting, then with the gained result as initial threshold; Set up master meter, auxilliary table memory space, deposit the electrocardiosignal WAVELET PACKET DECOMPOSITION result who obtains above in master meter, open sub-scanning storehouse and coding result memory space; Give threshold value (T) with the initial threshold assignment;

(6) master meter is scanned, each node on the table all is divided into positive significant coefficient, bears a heavy burden and want coefficient, zerotree root, isolated zero according to threshold value (T), scanning result is deposited in the master meter;

(7) master meter is translated into binary code stream and deposited in the coding result memory space, will be labeled as positive significant coefficient in the master meter and bear a heavy burden and want the node of coefficient to move to the sub-scanning storehouse;

(8) the sub-scanning storehouse is carried out sub-scanning, concern according to threshold value (T) and node size and carry out " 0 ", " 1 " coding, data flow is deposited in the coding result memory space simultaneously;

(9) upgrading threshold value (T) is 1/2 of current threshold value, returns step 6, and the main and auxiliary scanning process of repeating step (6) to (9) is 0 until threshold value (T);

(10) set up the wirelessly transmitting data bag, record optimal wavelet bag tree construction and data volume size and initial threshold information send packet by wireless network in labeling head.

The present invention provides a kind of coding/decoding method of above-mentioned electrocardio coding simultaneously, comprises getting off step:

(1) reception is through the packet of wireless network transmissions;

(2) read head labelling obtains optimal wavelet bag tree construction, obtains data volume size, initial threshold, gives threshold value (T) with the initial threshold assignment;

(3) read the master meter data flow, the reconstruct Data Position is carried out respective markers, reconstruct data, if be positive and negative significant coefficient, then the initial reconstitution value is 3/2 times of the plus or minus of threshold value (T), remaining position is zero;

(4) read auxilliary table data flow,, otherwise on the basis of original value, add 1/4 of upper threshold value (T), progressively accurate reconstruction value if for " 0 " then the reconstruction value absolute value deducts 1/4 of threshold value (T) on the basis of former coefficient absolute value;

(5) upgrading threshold value (T) is 1/2 of current threshold value, and the scan code that returns step (3) to (5) flows through journey, up to finishing decoding.

The inventor tries hard to find out a kind of compression algorithm of electrocardio efficiently from the angle of theory analysis and practice demonstration at the problem of data storage that exists in the remote electrocardiogram monitor terminal system and electrocardio real-time Transmission.At first pass through the characteristics of analysis conventional wavelet transformation, find out the problem that it exists in the data compression applications of cardiac monitoring terminal, then carry out efficient, reduce the resources occupation rate aspect and set about from the compression ratio bottleneck problem that solves traditional wavelet compression method and boosting algorithm respectively, proposed to carry out the efficient compression that fast lifting EZW algorithm is realized electrocardiogram (ECG) data, realized the telescopic progression transfer encoding of information source reconstruction quality simultaneously based on optimal wavelet bag space.Through the analysis of experimental data checking, provided by the invention based on fast lifting wavelet packet EZW method, not only made full use of the frequency domain correlation of electrocardiosignal, greatly reduce the comentropy of electrocardiosignal, and by introducing the advantage of operation of butterfly computation former address and integer arithmetic, effectively raise algorithm and carry out efficient, the electrocardiogram (ECG) data of having realized the cardiac monitoring terminal efficiently compresses, thereby the cardiac monitoring specialist system that can realize client cardiac monitoring terminal and server end is accurately mutual in real time, makes the cardiac obtain the effective monitoring and the record of round-the-clock electrocardiogram (ECG) data.

Description of drawings

Fig. 1: the wavelet decomposition sketch map, Fig. 1 (a) is the wavelet decomposition tree construction, Fig. 1 (b) coefficient of wavelet decomposition graph of a relation;

Fig. 2: WAVELET PACKET DECOMPOSITION structural representation;

Fig. 3: the cost curve synoptic diagram of father and son's node in the wavelet package transforms;

Fig. 4: Le_Gall 5/3 small echo promotes structure chart;

Fig. 5: the electrocardiogram (ECG) data compression result contrast of three kinds of algorithms.

The specific embodiment

The present invention is further described from several aspects below in conjunction with accompanying drawing, principle and embodiment.

1. the spatial electrocardiogram (ECG) data zerotree image of the breakthrough of compression ratio bottleneck problem-optimal wavelet bag

Under the theoretical frame of multiresolution, S.Mallat designs based on the wavelet decomposition of orthogonal filter group and restructing algorithm, the Mallat algorithm of promptly being transferred.Wavelet transformation to electrocardiosignal s (n) under the Mallat algorithm is

$s^{\left(j\right)}\left(n\right)=\underset{k}{\mathrm{\Σ}}h\left(k\right)\·s^{(j-1)}(n-2^{j-1}k)---\left(1\right)$

$d^{\left(j\right)}\left(n\right)=\underset{k}{\mathrm{\Σ}}g\left(k\right)\·s^{(j-1)}(n-2^{j-1}k)---\left(2\right)$

J is a wavelet transform dimension in the formula; H (k) is a low pass filter, and signal is level and smooth step by step by h (k), reflects the profile information of primary signal; And g (k) is a high pass filter, d ^(j)(n) be s ^(j-1)(n) and s ^(j)(n) difference between has reflected the detail section of signal.Fig. 1 (a) is the wavelet decomposition sketch map, and wherein node S represents primary signal; Node L _jBe illustrated in the profile information on the yardstick j, i.e. s ^(j)(n); Node H _jBe illustrated in the detailed information on the yardstick j, i.e. d ^(j)(n).

Shapiro finds, signal is being done in the process of wavelet transformation, strides between the wavelet coefficient of frequency band to have significant correlation properties.As shown in Fig. 1 (b), under the Mallat algorithm, the wavelet coefficient that obtains in the sampling process is inverted pyramid structure, and its sum is constant, and the dependency of these wavelet coefficients shows two aspects: 1. for each detailed information coefficient H _{J, k}(k coefficient on the j yardstick) must find two corresponding with it under adjacent small scale correlation coefficient H _J-1,2kAnd H _J-1,2k+1And profile information with have coefficient relation one to one with the detailed information of yardstick.2. generally speaking, if wavelet coefficient H _{J, k}Meaningless under given coding threshold value, its pairing correlation coefficient H then _J-1,2kAnd H _J-1,2k+1Generally also meaningless under this threshold value.

Shapiro utilizes the above-mentioned characteristic of wavelet transformation, realizes the reduction of comentropy, has proposed wavelet field EZW algorithm, the 2. pairing wavelet coefficient H of characteristics _{J, k}Be called zerotree root, H _{J, k}Constituted zero the tree with its all correlation coefficienies, and zero tree has very low comentropy, only need represent by considerably less bit stream, this is the primary condition of this method realization compression just.But sufficient entropy subtracts because general wavelet decomposition still is unrealized, and therefore can have redundancy, makes its wavelet coefficient some special circumstances may occur: as wavelet coefficient H _{J, k}Meaningless under given coding threshold value, and significant situation appears in its pairing correlation coefficient, this coefficient H _{J, k}Be called as isolated zero.The existence of isolated zero, making can't be according to H _{J, k}Directly judge the information of its correlation coefficient, H therefore can not only encode _{J, k}Also must examination H _{J, k}Pairing all correlation coefficienies, this will increase the coefficient of required coding, also will increase the computing workload simultaneously, and these all are unfavorable for the realization of efficient compression.

In order to overcome the bottleneck of wavelet field EZW compression algorithm, the present invention has at first found out and has verified the key factor that this algorithm can't breakthrough bottleneck by data test and theory analysis, promptly electrocardiogram (ECG) data is being reduced in the process of comentropy, wavelet field EZW compression algorithm is taked fixed band decomposition mode, do not take into full account the diversity between frequency band, thereby can't realize the further reduction of comentropy.Therefore, the present invention considers to adopt new method to replace the fixed frequency band is olation of traditional wavelet, subtracts with the entropy that compresses the back signal of guaranteeing to encode with the appearance that reduces the embedded encoded middle isolated zero of traditional wavelet field.And wavelet package transforms just in time can address that need, because it can obtain corresponding optimal wavelet bag band decomposition structure according to the frequency domain autocorrelation of electrocardiosignal, the optimum frequency-domain structure sign indicating number that structure is corresponding, reach the further reduction of comentropy, so just be expected to solve the bottleneck of Signal Compression.

In order to obtain this optimum frequency-domain structure sign indicating number, need the high frequency details in the traditional wavelet (Fig. 1 (a)) is also taken in, carry out suitable decomposition.When all nodes all need to decompose, can obtain complete wavelet bag decomposition texture as shown in Figure 2.

Whether in actual applications, the process of obtaining optimum frequency-domain structure sign indicating number is the root node that decomposes from frequency domain, be worth decomposing by the node of weighing on each layer yardstick, and the complete wavelet pack arrangement is successively pruned to obtain optimum structure.Need in the pruning process by the foundation of suitable cost function, promptly will fully coordinate distortion rate and compress recently to make up cost function from judging the criterion of compression performance quality as the optimum frequency-domain structure of search.

Had after the cost function, just can decompose cost, obtained optimum wavelet packet space structure by weighing.The following derivation that provides cost function.

Definition D is the data distortion rate, and R is the data volume after quantizing, and then compresses purpose and can be expressed as: satisfying (D≤D under the distortion requirement prerequisite _b, D _bBe maximum admissible distortion), reach maximum compression (minR).Take all factors into consideration distortion rate and data volume, can obtain corresponding Lagrange cost function

J＝D+λR (3)

In the formula: J is the compression cost, and λ is the Lagrangian factor (λ 〉=0), and expression is transformed into distortion rate with data bit-rate and explains spatial quality, and definition λ=Δ D/ Δ R can require to choose according to compression quality.Thus, above-mentioned compression purpose can be exchanged into: under the prerequisite that distortion meets the demands, find the pairing frequency domain decomposition texture of minimal compression cost J.As shown in Figure 3, if D _C1+ λ ₂R _C1)+(D _C2+ λ ₃R _C2)≤D _P+ λ ₁R _P, then father node is decomposed into two child nodes.

Obtain criterion according to the optimal spatial that provides above, electrocardiosignal is carried out the wavelet packet optimal spatial decompose, then, carry out EZW according to correlative relationship corresponding between each frequency band space, break through original compression ratio bottleneck with this, realize the efficient compression of electrocardiogram (ECG) data.

2. algorithm is carried out the breakthrough-optimal wavelet bag space boosting algorithm of efficient

By the EZW algorithm being incorporated into the flexible and changeable wavelet packet space of band decomposition structure, the present invention has broken through the compression ratio bottleneck of wavelet space zerotree image algorithm, required data volume when having significantly reduced the equal information transmission.But, relative wavelet field zerotree image algorithm, the core algorithm that the EZW that influence Wavelet Packet Domain is carried out efficient does not change, and encryption algorithm complexity partly increased, and this just makes the execution efficient of compression algorithm further reduce.Therefore, the present invention is after the compression ratio bottleneck of breaking through traditional method, in order to develop the efficient data compression method that is suitable for the cardiac monitoring terminal, implementation to compression algorithm is further analyzed, by continuous theory analysis and putting into practice in the process of argumentation, finally having found out influences the key factor that algorithm is carried out efficient: though in front in the step, by the EZW algorithm being incorporated into the flexible and changeable wavelet packet space of band decomposition structure, broken through the compression ratio bottleneck, but core algorithm still adopts the Mallet algorithm flow, promptly adopt a kind of Shuangzi band conversion scheme to calculate realization step by step, in each step, be signal decomposition high-frequency sub-band and low frequency sub-band all, then it is extracted sampling, low frequency sub-band is constantly carried out recursive operation, up to reaching required decomposed class.If carry out monitoring terminal electrocardiogram (ECG) data compressed encoding but use above-mentioned algorithm, will expose a lot of shortcoming and defect parts, mainly comprise following several aspect: all will have the data of half to be dropped after (1) each grade filtering, and that is to say that the multiplication calculating of half will be invalid; (2) because of the existence of floating-point coefficient, traditional small echo may be realized the wavelet transformation of integer to integer hardly; (3) when electrocardiogram (ECG) data is decompressed, because implementing, the inverse transformation of traditional small echo needs to obtain by reconstruction filter after two interpolation, use trouble, operand is big; (4) can produce boundary effect in the processing on border, different continuation modes have different results, can't realize the harmless reconstruct of signal fully; (5) data that have more one times of low frequency sub-band are used in every grade of conversion, and are interrelated before and after the filtering operation of data, need other memory element.

Thus,, need to introduce a kind of algorithm elevation scheme, promote the execution efficient of compressed encoding, i.e. the lifting thought of algorithm at the above-mentioned shortcoming that exists in the algorithm.At present, aspect the utilization that promotes thought, fast fourier transform is to the lifting of traditional Fourier transform, lifting wavelet transform all is being successful aspect the raising of algorithm execution efficient to the lifting of traditional wavelet, but, promoting thought does not appear in the newspapers so far in the spatial utilization of wavelet packet, and in order to keep having broken through among the present invention the advantage of wavelet field zerotree image compression ratio, must be incorporated into the wavelet packet space with promoting thought, propose based on promoting the spatial EZW algorithm of optimal wavelet bag.

When carrying out the electrocardiogram (ECG) data compression, in order to obtain the lossless coding of electrocardiosignal, the present invention has introduced the lossless coding wavelet filter that present JPEG2000 recommends, Le_Gall 5/3 small echo, wherein 5 represent low pass filter length, 3 represent high pass filter length, and concrete wavelet coefficient is as shown in table 1 below.

Table 1Le_all 5/3 wavelet filter coefficient

N	Decompose low pass filter	Rebuild low pass filter	N	Decompose low pass filter	N	Rebuild low pass filter
							0	3/4	1	-1	1	+1	3/4
+1，-1	1/4	1/2	0，-2	-1/2	+2，0	-1/4
							+2，-2	-1/8				+3，-1	-1/8

Above-mentioned coefficient in the table 1 can represent with matrix form, that is:

$P\left(z\right)=\left[\begin{array}{cc}-\frac{1}{8}{Z}^{-1}+\frac{3}{4}-\frac{1}{8}Z& \frac{1}{4}+\frac{1}{4}Z\\ -\frac{1}{2}{Z}^{-1}-\frac{1}{2}& 1\end{array}\right]---\left(4\right)$

Be applied to the wavelet packet space in order to promote thought, at first need and above-mentioned FIR bank of filters calculating to resolve into some lifting step, make computational process more succinct by decomposing these wave filter, replace with the new algorithm that extracts filtering more earlier with this step that extracts after with first filtering in the Mallet algorithm.Therefore, the present invention decomposes P (z) according to promoting thought (leading diagonal that obtains each factor is a unit 1), obtains the following expression formula of being made up of 3 part factors (from left to right the called after odd even is decomposed factor, general picture factor, details factor respectively)

$P\left(z\right)=\left[\begin{array}{cc}1& 0\\ 0& 1\end{array}\right]\left[\begin{array}{cc}1& \frac{1}{4}+\frac{1}{4}\mathrm{<figure-callout\; id="0"\; label="Z"\; filenames="A20081005233700142.png"\; state="\{\{state\}\}">Z</figure-callout>}\\ 0& 1\end{array}\right]\left[\begin{array}{cc}1& 0\\ -\frac{1}{2}{Z}^{-1}-\frac{1}{2}& 1\end{array}\right]---\left(5\right)$

Can find from first factor (odd even decomposition factor), as long as electrocardiosignal is resolved into odd number sequence and even number sequence, just can save the back extraction step, directly obtain high frequency details and low frequency general picture signal, promptly can adopt merging process to substitute the process of getting rid of redundant data after the division again, reduce the complexity of algorithm.Therefore, for electrocardiosignal X (z), by being decomposed into odd sequence X _e(z) and even sequence X _o(z), just can adopt method for improving to distribute realizes:

$P\left(z\right)X\left(z\right)=\left[\begin{array}{cc}1& \frac{1}{4}+\frac{1}{4}\mathrm{<figure-callout\; id="0"\; label="Z"\; filenames="A20081005233700142.png"\; state="\{\{state\}\}">Z</figure-callout>}\\ 0& 1\end{array}\right]\left[\begin{array}{cc}1& 0\\ -\frac{1}{2}{Z}^{-1}-\frac{1}{2}& 1\end{array}\right]\left[\begin{array}{c}{X}_e\left(z\right)\\ X_o\left(z\right)\end{array}\right]---\left(6\right)$

The first step becomes sequence of parity to signal decomposition: and d (0, r)=x _n(2r+1); S (0, r)=x _n(2r);

In second step, realize the details factor $\left[\begin{array}{cc}1& 0\\ -\frac{1}{2}{Z}^{-1}-\frac{1}{2}& 1\end{array}\right]$ Time domain result of calculation:

D (j, r)=d (j, r)-[s (j, r)+s (j, r+1)+1]/2, j=1, the i.e. detailed information of the first rank yardstick.

In the 3rd step, realize the general picture factor $\left[\begin{array}{cc}1& \frac{1}{4}+\frac{1}{4}\mathrm{<figure-callout\; id="0"\; label="Z"\; filenames="A20081005233700142.png"\; state="\{\{state\}\}">Z</figure-callout>}\\ 0& 1\end{array}\right]$ Time domain result of calculation:

S (j, r)=s (j, r)-[d (j, r)+d (j, r-1)]/4, j=1, the i.e. profile information of the first rank yardstick.

Above-mentioned steps can represent with butterfly mapped structure figure, as shown in Figure 4.

We are as can be seen from figure: initial data resolves into sequence of parity d _o ⁰, s _o ⁰After, be d through first lifting result calculated _o ¹, it covers and is stored in d _o ⁰The former address on, d _o ¹, s _o ¹(s _o ¹=s _o ⁰) the second lift result calculated is s _l ¹, it covers and is stored in s _o ⁰The former address on, both hocket, and do not conflict mutually, so-called former address operation that Here it is.This method need not newly be opened the space, and initial data is progressively replaced by wavelet coefficient.Above-mentioned algorithm is because the factor 1/2,1/4 occurred, and decimal appears in result of calculation probably, and this will cause truncation effect in data compression, introduces quantization error, makes compression process irreversible.In order to address this problem, on the basis of existing small echo, constructed the small echo that has " integer is to integer transform character ".Promptly the result of calculation behind each multiplication is rounded, thus, corresponding general picture signal and detail signal are converted to respectively:

General picture: s (j, r)=s (j, r)-integral{[d (j, r)+d (j, r-1)]/4}

Details: d (j, r)=d (j, r)-integral{[s (j, r)+s (j, r+1)+1]/2}

This number " integer is to integer transform " of realizing in lifting process is fully feasible: because in lifting process, always there is one-component to remain unchanged, therefore as long as the final result of this lifting process rounded can guarantee that each step result is an integer, and this process completely reversibility--the result that-one integer set obtains by integer lifting wavelet transform remains the integer set.During inverse transformation, only need progressively from the result, deduct integral{[d (j, r)+d (j, r-1)]/4} and integral{[s (j, r)+s (j, r+1)+1]/2} gets final product, whole process decimal can not occur.

When needs carry out the wavelet package transforms of next yardstick, the sequence of parity division is carried out in all filtering outputs (comprising general picture signal and detail signal) that only need to go up a yardstick once more, then, can obtain the wavelet package transforms result of required yardstick as the list entries among Fig. 4.

3. algorithm and flow process

Before providing algorithm flow of the present invention, at first four class wavelet packet coefficients in the electrocardiosignal zerotree image process are provided definition.According to the relation of wavelet packet coefficient and threshold value T, can be divided into following 4 class coefficients:

(1) POS, positive significant coefficient (greater than the positive coefficient of threshold value T);

(2) NEG bears a heavy burden and wants coefficient (absolute value is greater than the negative coefficient of threshold value T);

(3) ZTR, zerotree root (its offspring is the less important coefficient of less important coefficient);

(4) IZ, lonely zero (the less important coefficient that significant coefficient is arranged among its offspring).

The coefficient classification is undertaken by main scanning.To a coefficient, with it and threshold, classify according to top method, if scanning is not considered the comparison that has carried out the front and carried out separately each time, algorithm will be carried out a lot of repetitive operations.In order to reduce operand, all should carry out labelling to the result of each scanning: if significant coefficient is POS or NEG according to sign flag just, and absolute value is pressed into storehouse for sub-scanning usefulness, simultaneously also will be with corresponding coefficient zero setting; If lonely zero a labelling IZ; If zero tree then is labeled as ZTR, its descendants no longer is scanned.

The process of coding not only comprises the main scanning that coefficient is classified, and also will quantize one by one significant coefficient, carries out interval mark at each stacked significant coefficient, and quantized one by one process is called as sub-scanning.It is to use threshold value T (0) one by one that what is called quantizes one by one, T (1), T (2) ... T (n), and deciding significant coefficient is to belong to still upper half [3T/2,2T] of bottom half [T, 3T/2], if just write " 1 " toward auxilliary table at upper half, otherwise just writes " 0 ".Wherein, choosing of threshold series is to carry out according to following formula:

Initial threshold is T (0)=2B, B=integral (log2 (max|X|)) wherein, and X is an array of depositing wavelet packet coefficient, max|X| is a maximum of asking element absolute value in the array; Integral is for rounding operation.All the other are obtained by recurrence formula T (i+1)=T (i)/2.Along with the increase that quantizes times N one by one, the bit number of coding back output also will increase thereupon, and the reconstructed value of wavelet packet coefficient is also more near original value, and the quality of recovering electrocardiosignal is also just high more.When decoding, can block code stream as required at any time, recover electrocardiosignal with the least possible bit number.

Provide algorithm flow of the present invention below, specific as follows,

(1) based on Lifting Wavelet bag EZW compression algorithm flow process:

(1) electrocardiosignal pretreatment: filtering baseline drift and power frequency are disturbed;

(2) electrocardiosignal that pretreatment is obtained is carried out the sequence of parity division, then obtains single order detail signal node and single order general picture signal node by boosting algorithm;

(3) judge according to cost function whether needs are decomposed into child node for single order details and single order general picture father node, if desired, then it is carried out the sequence of parity division, then obtain corresponding second order child node by boosting algorithm.

(4) judge whether have needs to continue to be decomposed into corresponding three rank child nodes in the second order child node according to cost function, adopt boosting algorithm that these second order child nodes are decomposed.By that analogy, up to the optimal wavelet bag space structure that obtains the minimum information entropy.

(5) initialization threshold value T sets up master meter, auxilliary table memory space, deposits the electrocardiosignal WAVELET PACKET DECOMPOSITION result who obtains above in master meter, opens sub-scanning storehouse and coding result memory space.

(6) master meter is scanned, each node on the table all is divided into positive significant coefficient, bears a heavy burden and want coefficient, zerotree root, isolated zero according to threshold value T, scanning result is deposited in the master meter.

(7) master meter being translated into binary code stream deposits in the coding result storage file.To be labeled as positive significant coefficient in the master meter and bear a heavy burden and want the node of coefficient to move to the sub-scanning storehouse.

(8) the sub-scanning storehouse is carried out sub-scanning, concern according to threshold value and node size and carry out " 0 ", " 1 " coding, simultaneously data flow is deposited in the coding result storage file.

(9) change threshold value, repeat main and auxiliary scanning and (promptly return step 6), until T=0.

(10) set up the wirelessly transmitting data bag, record optimal wavelet bag tree construction and data volume size and initial threshold information send packet by wireless network in labeling head.

(2) based on Lifting Wavelet bag EZW decompression algorithm flow process:

(1) reception is through the packet of wireless network transmissions;

(2) read head labelling obtains optimal wavelet bag tree construction, obtains information such as data volume size, initial threshold.

(3) read the master meter data flow, the reconstruct Data Position is carried out respective markers, reconstruct data, if be positive and negative significant coefficient, then the initial reconstitution value is the 3T/2 of plus or minus, remaining position is zero.

(4) read auxilliary table data flow,, otherwise on the basis of original value, add T/4, progressively accurate reconstruction value if for " 0 " then the reconstruction value absolute value deducts T/4 on the basis of former coefficient absolute value.

(5) upgrade threshold value, multiple scanning code stream.Up to finishing decoding.

The remote electrocardiogram monitor terminal comprises DSP digital sampling and processing, main controller module and MC35 wireless sending module three big modules.Wherein the DSP module is in the core of algorithm design, and it obtains data from acquisition module on the one hand, provides compression data packet to host computer again on the other hand.The present invention carries out transplanting under the DSP hardware platform with algorithm, and the function that then utilizes CCS auxiliary development software to provide comes the time and the space expense of algorithm for estimating.

Experimental data is taken from MIT arrhythmia data base, is downloaded in the system by the JTAG mouth by PC, has ten groups of data, and each process segments of data is 12000 bytes.DSP is outer to be 20MHz frequently, enable inner PLL=5, so the instruction cycle is 10ns.The developing instrument CCS that utilizes TI to provide uses the monitoring module of BIOS internal system to come locator(-ter), obtains running time.Because the BIOS system adopts idling cycle that program is moved and monitors, therefore can not influence the normal operation of program.

In order to provide compression effectiveness of the present invention, we adopt three kinds of algorithms to calculate to each group data: traditional small echo EZW algorithm, tradition wavelet packet EZW algorithm and Lifting Wavelet bag EZW algorithm of the present invention, result of calculation as shown in Figure 5.

As can be seen from Figure 5, by the zerotree image algorithm being incorporated into traditional wavelet packet space, (the average compression ratio that traditional wavelet packet and zero tree-encoding obtains is 13.9 can to obtain better compression effectiveness, with respect to 9.8 of traditional ZT coding, be greatly improved), but adopt traditional wavelet packet and zero tree-encoding method, the desired compression time has but increased (average compression time is 8.7 seconds, and adopts Wavelet Zero-Tree Coding Algorithm only to need 6.1 seconds).

Provided by the invention the average compression ratio of electrocardiogram (ECG) data has reached 16.3 based on the spatial EZW algorithm of Lifting Wavelet bag by adopting, and corresponding needed average compression time decreased to 5.4 second.This shows that algorithm of the present invention can be realized the real-time high-efficiency compression of electrocardiogram (ECG) data.

Claims (2)

1. the electrocardio compression method of a radio electrocardiographicmonitoring monitoring instrument comprises the following steps:

(1) electrocardiosignal that collects is carried out pretreatment;

(2) electrocardiosignal that pretreatment is obtained is carried out the sequence of parity division, then obtains single order detail signal node and single order general picture signal node by boosting algorithm;

(3) judge according to cost function whether needs are decomposed into child node for single order details and single order general picture father node, if desired, then it is carried out the sequence of parity division, then obtain corresponding second order child node by boosting algorithm;

(4) judge whether have needs to continue to be decomposed into corresponding three rank child nodes in the second order child node according to cost function, adopt boosting algorithm that these second order child nodes are decomposed, by that analogy, up to the optimal wavelet bag space structure that obtains the minimum information entropy;

(5) traversal wavelet packet all coefficients are obtained the maximum of coefficient absolute value, keep the highest order of this numerical value binary digit, with all the other low level zero setting, then with the gained result as initial threshold; Set up master meter, auxilliary table memory space, deposit the electrocardiosignal WAVELET PACKET DECOMPOSITION result who obtains above in master meter, open sub-scanning storehouse and coding result memory space; Give threshold value (T) with the initial threshold assignment;

(6) master meter is scanned, each node on the table all is divided into positive significant coefficient, bears a heavy burden and want coefficient, zerotree root, isolated zero according to threshold value (T), scanning result is deposited in the master meter;

(7) master meter is translated into binary code stream and deposited in the coding result memory space, will be labeled as positive significant coefficient in the master meter and bear a heavy burden and want the node of coefficient to move to the sub-scanning storehouse;

(8) the sub-scanning storehouse is carried out sub-scanning, concern according to threshold value (T) and node size and carry out " 0 ", " 1 " coding, data flow is deposited in the coding result memory space simultaneously;

(9) upgrading threshold value (T) is 1/2 of current threshold value, returns step 6, and the main and auxiliary scanning process of repeating step (6) to (9) is 0 until threshold value (T);

(10) set up the wirelessly transmitting data bag, record optimal wavelet bag tree construction and data volume size and initial threshold information send packet by wireless network in labeling head.

2. the coding/decoding method of a radio electrocardiographicmonitoring monitoring instrument is used to the coding that decompresses and adopt the described compression method of claim 1 to obtain, it is characterized in that comprising the following steps:

(1) reception is through the packet of wireless network transmissions;

(2) read head labelling obtains optimal wavelet bag tree construction, obtains data volume size, initial threshold, gives threshold value (T) with the initial threshold assignment;

(3) read the master meter data flow, the reconstruct Data Position is carried out respective markers, reconstruct data, if be positive and negative significant coefficient, then the initial reconstitution value is 3/2 times of the plus or minus of threshold value (T), remaining position is zero;

(4) read auxilliary table data flow,, otherwise on the basis of original value, add 1/4 of upper threshold value (T), progressively accurate reconstruction value if for " 0 " then the reconstruction value absolute value deducts 1/4 of threshold value (T) on the basis of former coefficient absolute value;

(5) upgrading threshold value (T) is 1/2 of current threshold value, and the scan code that returns step (3) to (5) flows through journey, up to finishing decoding.

Images