CN113143284B - Electrocardiosignal compression method based on wavelet transformation and dual-mode prediction - Google Patents

Abstract

Translated fromChinese

一种基于小波变换和双模预测的心电信号压缩方法，所述方法包括以下步骤：第一步、信号分解：对心电信息的采样信号进行一级的提升小波变换，得到相同数量的高频系数和低频系数，去除高频系数保留低频系数；第二步、缩放平滑：对得到的小波系数低频部分做除法缩放和平滑操作；第三步、预测：对系数采用线性预测和模板预测相结合的方法进行预测并得到预测误差；第四步、编码：将预测误差应用二级golomb‑rice编码方式编码；第五步、封装：将编码值和预测需要的信息一起打包形成最后的压缩数据流。本发明在较低的失真率下实现了较高的压缩效果，适用于可穿戴心电监测设备。

An electrocardiographic signal compression method based on wavelet transform and bimodal prediction, the method includes the following steps: the first step, signal decomposition: performing a first-level lifting wavelet transform on the sampled signal of electrocardiographic information to obtain the same number of high Frequency coefficients and low-frequency coefficients, remove high-frequency coefficients and retain low-frequency coefficients; second step, scaling and smoothing: perform division, scaling and smoothing operations on the low-frequency part of the obtained wavelet coefficients; third step, prediction: use linear prediction and template prediction for coefficients. The combined method is used to predict and obtain the prediction error; the fourth step, encoding: the prediction error is encoded by the second-level golomb-rice encoding method; the fifth step, encapsulation: the encoded value and the information required for prediction are packaged together to form the final compressed data flow. The present invention achieves a higher compression effect at a lower distortion rate, and is suitable for wearable ECG monitoring equipment.

Description

Translated fromChinese

基于小波变换和双模预测的心电信号压缩方法ECG Signal Compression Method Based on Wavelet Transform and Bimodal Prediction

技术领域technical field

本发明涉及生物医学信息处理领域，具体涉及一种基于小波变换和双模预测的心电信号压缩方法The invention relates to the field of biomedical information processing, in particular to an electrocardiographic signal compression method based on wavelet transform and bimodal prediction

背景技术Background technique

随着科技的发展和人类生活质量的提高，人们对自身的健康状况越来越重视，全球医疗费用也逐年增长。同时，老龄化的加剧使得更多的人需要实时的照顾，对医疗资源的需求日益增加。这一系列的社会问题极大促进了远程可穿戴健康监测设备的快速发展。可穿戴远程医疗设备通过实时监测，使得病人一旦产生患病趋势，在病情发展初期便能采取相应措施加以应对，极大地降低了病情进一步恶化的概率，同时，病人也能足不出户就能完成各项医疗行为，不必再定期往返医院做医疗检查，节省了大量的时间和经济成本。近年来，伴随物联网和无线通信的迅速发展，具有低成本、高效率等优点的可穿戴远程医疗设备开始得到应用，极大地降低了人工的成本，对现代医疗的发展具有重要意义。With the development of science and technology and the improvement of human life quality, people pay more and more attention to their own health status, and the global medical expenses are also increasing year by year. At the same time, the aggravation of aging makes more people need real-time care, and the demand for medical resources is increasing. This series of social problems has greatly promoted the rapid development of remote wearable health monitoring devices. Through real-time monitoring, wearable telemedicine devices enable patients to take corresponding measures in the early stage of disease development once they develop a disease trend, which greatly reduces the probability of further deterioration of the disease. At the same time, patients can also stay home. After completing various medical behaviors, there is no need to go to the hospital for medical examinations regularly, which saves a lot of time and economic costs. In recent years, with the rapid development of the Internet of Things and wireless communication, wearable telemedicine devices with the advantages of low cost and high efficiency have begun to be applied, which greatly reduces labor costs and is of great significance to the development of modern medical care.

心电信号(ECG)是一项重要的人体生理信号，它以时间为单位记录人体心脏的各种生理活动，可以反映出心脏的节律以及其电传导的生理信息，可以较为客观地反映出心脏各个部位的生理情况,被广泛应用于心血管疾病的诊断。远程可穿戴心电设备可以长时间监测人体的心电信号，用于诊断和预防可能发生的心血管疾病，相对于临床的检查和诊断具有明显的优势。为了应对突发的心脏疾病，需要长时间不间断地采集心电信号，这对心电信号地传输和存储造成了较大地压力，在可穿戴心电监测设备中，无线传输所占的功耗超过了70％，于是对心电信号进行压缩就变得尤为重要。Electrocardiogram (ECG) is an important human physiological signal. It records various physiological activities of the human heart in units of time, which can reflect the rhythm of the heart and the physiological information of its electrical conduction, and can objectively reflect the heart. Physiological conditions of various parts are widely used in the diagnosis of cardiovascular diseases. Remote wearable ECG devices can monitor the human body's ECG signals for a long time to diagnose and prevent possible cardiovascular diseases, which has obvious advantages over clinical examination and diagnosis. In order to deal with sudden heart diseases, it is necessary to collect ECG signals for a long time without interruption, which causes great pressure on the transmission and storage of ECG signals. In wearable ECG monitoring equipment, the power consumption occupied by wireless transmission If it exceeds 70%, it is particularly important to compress the ECG signal.

发明内容SUMMARY OF THE INVENTION

为了克服现有技术的不足，本发明提供了一种基于小波变换和双模预测的心电信号压缩方法，所述方法基于小波变换和双模预测对心电信号进行压缩，实现了较高的压缩率，同时保留了绝大部分有效信息，使压缩信号重构后的心电信号失真度较小，能够完成各种心脏疾病的诊断。In order to overcome the deficiencies of the prior art, the present invention provides an ECG signal compression method based on wavelet transform and bimodal prediction. The method compresses the ECG signal based on wavelet transform and bimodal prediction, and achieves higher At the same time, most of the effective information is retained, so that the distortion of the ECG signal after the reconstruction of the compressed signal is small, and the diagnosis of various heart diseases can be completed.

本发明可通过如下的技术方案实现：The present invention can be realized through the following technical scheme:

一种基于小波变换和双模预测的心电信号压缩方法，所述方法包括以下步骤：An electrocardiographic signal compression method based on wavelet transform and bimodal prediction, the method comprises the following steps:

第一步、信号分解：对心电信息的采样信号进行一级的提升小波变换，得到相同数量的高频系数和低频系数，去除高频系数保留低频系数；The first step, signal decomposition: perform first-level lifting wavelet transform on the sampled signal of ECG information to obtain the same number of high-frequency coefficients and low-frequency coefficients, remove the high-frequency coefficients and retain the low-frequency coefficients;

第二步、缩放平滑：对得到的小波系数低频部分做除法缩放和平滑操作；The second step, scaling and smoothing: perform division, scaling and smoothing operations on the low-frequency part of the obtained wavelet coefficients;

第三步、预测：对系数采用线性预测和模板预测相结合的方法进行预测并得到预测误差；The third step, prediction: the coefficient is predicted by a combination of linear prediction and template prediction, and the prediction error is obtained;

第四步、编码：将预测误差应用二级golomb-rice编码方式编码；The fourth step, encoding: the prediction error is encoded by the secondary golomb-rice encoding method;

第五步、封装：将编码值和预测需要的信息一起打包形成最后的压缩数据流。The fifth step, encapsulation: encapsulate the encoded value and the information required for prediction together to form the final compressed data stream.

进一步，所述第一步中，所述小波变换采用了5/3提升小波变换，保留低频系数，其表达式为：Further, in the first step, the wavelet transform adopts 5/3 lifting wavelet transform, and the low-frequency coefficient is reserved, and its expression is:

式中X[2n+1]和X[2n]分别为输入信号X[n]分裂得到的奇偶两个序列，d[n]为提升小波变换得到的高频小波系数序列，s[n]为提升小波变换得到的低频尺度系数序列，在计算高频系数d[n]时，需要同时取得奇序列的值X[2n+1]和前后两个偶序列的值X[2n]和X[2n+2]进行计算，计算低频系数s[n]时，需要同时取得偶序列的值X[2n]和该偶序列值前后两个高频系数d[n]和d[n-1]的值。In the formula, X[2n+1] and X[2n] are the parity sequences obtained by splitting the input signal X[n], respectively, d[n] is the high-frequency wavelet coefficient sequence obtained by lifting wavelet transform, and s[n] is When calculating the high-frequency coefficient d[n] of the low-frequency scale coefficient sequence obtained by lifting the wavelet transform, it is necessary to obtain the value X[2n+1] of the odd sequence and the values X[2n] and X[2n of the two even sequences at the same time. +2] for calculation, when calculating the low-frequency coefficient s[n], it is necessary to obtain the value X[2n] of the even sequence and the values of the two high-frequency coefficients d[n] and d[n-1] before and after the even sequence value at the same time .

再进一步，所述第二步中，对得到的小波系数低频部分做除法缩放和平滑操作如下：对低频信号进行缩放，获得多种不同信号质量的心电信号，对小波系数做了平滑处理，对连续的三个小波系数x₁、x₂、x₃,若满足x₁＝x₃，|x₂-x₁|＝1则使x₂＝x₁，即将中间点视为毛刺，修改该值使其与前后值相等。Further, in the second step, the division, scaling and smoothing operations are performed on the low-frequency part of the obtained wavelet coefficients as follows: the low-frequency signal is scaled to obtain a variety of ECG signals with different signal qualities, and the wavelet coefficients are smoothed. For three consecutive wavelet coefficients x₁ , x₂ , x₃ , if x₁ =x₃ , |x₂ -x₁ |=1, then x₂ =x₁ , that is, the middle point is regarded as a burr, and the value to make it equal to the value before and after.

所述第三步中，双模预测采用线性预测与模板预测相结合的方式，对于心电信号非QRS区域采用0阶线性预测，对于QRS区域采用模板预测和2阶线性预测。In the third step, the bi-modal prediction adopts a combination of linear prediction and template prediction, and adopts 0-order linear prediction for the non-QRS region of the ECG signal, and adopts template prediction and 2-order linear prediction for the QRS region.

所述第四步中，二级golomb-rice编码方式编码，参数k的大小与编码的效率紧密相关，即k为

或

时编码的码长最小，m为待编码的值，在进行golomb-rice编码前先对参数k的值进行预测，从而获得比较高的编码效率，预测的公式表示为：

其中d为临时变量，初始值为64。In the fourth step, the second-level golomb-rice encoding method is encoded, and the size of the parameter k is closely related to the encoding efficiency, that is, k is

or

When the code length of the encoding is the smallest, m is the value to be encoded. Before the golomb-rice encoding, the value of the parameter k is predicted to obtain a relatively high encoding efficiency. The prediction formula is expressed as:

where d is a temporary variable with an initial value of 64.

所述第四步中，将预测误差应用二级golomb-rice编码方式编码，为对golomb-rice编码算法做出改进使压缩性能更好，通过增加一前缀码“1110”，对于商值大于等于3且小于8的编码，需在商值的一元编码前增加一个比特“1”，当出现连续的数据0后，则使用该前缀码后配合自定义的游程编码对该片段进行编码，当出现QRS区间或连续数据0时需要使用前缀码和再次查表。In the fourth step, the prediction error is encoded by applying the second-level golomb-rice encoding method to improve the compression performance of the golomb-rice encoding algorithm. By adding a prefix code "1110", for the quotient greater than or equal to For codes of 3 and less than 8, a bit "1" needs to be added before the unary encoding of the quotient value. When there iscontinuous data 0, the prefix code is used and the custom run-length encoding is used to encode the segment. When the QRS interval or continuous data is 0, you need to use the prefix code and look up the table again.

所述模板预测为：对于每个QRS区间的系数，都用已有的QRS模板中的值进行预测，在不同的模板中选取最为接近的一个进行预测，同时和应用2阶线性预测得到的值进行比较，选择误差最小的结果作为最终的预测结果。The template prediction is: for the coefficient of each QRS interval, use the value in the existing QRS template to predict, select the closest one in different templates to predict, and use the value obtained by applying the second-order linear prediction simultaneously. For comparison, the result with the smallest error is selected as the final prediction result.

进一步地，所述第五步中，将预测误差应用二级golomb-rice编码方式编码，和预测需要的信息一起打包形成最后的压缩数据流包括编码和数据封装，采用基于golomb-rice编码算法改进得到的二级golomb-rice编码对预测误差进行编码，将信号编码值和解码需要的信息一起封装。Further, in the described 5th step, the prediction error is encoded by applying the secondary golomb-rice encoding method, and the information required for the prediction is packaged together to form the final compressed data stream including encoding and data encapsulation, and the improvement based on the golomb-rice encoding algorithm is adopted. The resulting two-level golomb-rice encoding encodes the prediction error, encapsulating the encoded value of the signal with the information needed for decoding.

本发明与现有的技术相比，有以下优点：Compared with the prior art, the present invention has the following advantages:

(1)本发明提供了一种基于小波变换和双模预测的心电信号压缩方法，在较低失真度下实现了较高的心电信号压缩率，同时也支持多种不同的压缩等级以适应不同的应用场景。(1) The present invention provides an electrocardiographic signal compression method based on wavelet transform and bimodal prediction, which achieves a higher electrocardiographic signal compression rate under lower distortion, and also supports a variety of different compression levels to Adapt to different application scenarios.

(2)应用5/3提升小波变换处理信号，降低数据的相关性提升压缩率，同时5/3提升小波变换只需要简单的加法运算，相比需要大量乘法运算的传统小波，拥有更低的运算复杂度(2) Apply 5/3 lifting wavelet transform to process signals, reduce the correlation of data and improve the compression rate. At the same time, 5/3 lifting wavelet transform only needs simple addition operation. Compared with traditional wavelet which requires a lot of multiplication operations, it has lower operational complexity

(3)将小波系数分为QRS区和非QRS区，对于非QRS区，采用线性预测的方法进行预测，对于QRS区域应用线性预测和模板预测相结合的方法进行预测，从而进一步降低数据间的相关性以实现更高的压缩率。(3) The wavelet coefficients are divided into QRS area and non-QRS area. For the non-QRS area, the linear prediction method is used for prediction. For the QRS area, the combination of linear prediction and template prediction is used for prediction, so as to further reduce the difference between the data. correlation to achieve higher compression ratios.

(4)设计的一种新型的编码方式——二级golomb-rice编码，将不同特性区域分开编码，进而得到比用其他编码方式更好的压缩效果。(4) A new type of coding method is designed—two-level golomb-rice coding, which encodes different characteristic regions separately, and then obtains a better compression effect than other coding methods.

附图说明Description of drawings

图1为本发明的流程图。FIG. 1 is a flow chart of the present invention.

图2为基于MIT-BIH数据库的实验效果图，其中，(a)是原始心电采样信号，(b)是提升小波变换后得到的高频系数，(c)是提升小波变换后得到的低频系数，(d)是对低频系数做缩放处理后得到的值，(e)是对缩放后的系数进一步做平滑操作得到的值，(f)是对平滑后的系数进行双模预测，最终得到的预测误差值。Figure 2 shows the experimental results based on the MIT-BIH database, in which (a) is the original ECG sampled signal, (b) is the high-frequency coefficient obtained after boosting the wavelet transform, and (c) is the low-frequency coefficient obtained after boosting the wavelet transform coefficient, (d) is the value obtained after scaling the low-frequency coefficient, (e) is the value obtained by further smoothing the scaled coefficient, (f) is the bimodal prediction of the smoothed coefficient, and finally obtain prediction error value.

图3为数据封装形式的示意图。FIG. 3 is a schematic diagram of a data encapsulation form.

具体实施方式Detailed ways

为了本发明的技术方案更加清晰，下面结合附图和实施例对本发明进行更完整地表述。本发明的实施包括下面的实施例但不仅限于此。基于本发明中的实施例，本领域技术人员在没有做出创造性劳动前提所获得的所有其他实施例，都属于本发明保护的范围。In order to make the technical solutions of the present invention clearer, the present invention will be more completely described below with reference to the accompanying drawings and embodiments. The implementation of the present invention includes the following examples but is not limited thereto. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative efforts shall fall within the protection scope of the present invention.

参照图1～图3，一种基于小波变换和双模预测的心电信压缩方法，具体流程如图1所示，包括以下步骤：Referring to Fig. 1 to Fig. 3 , a method for compressing ECG based on wavelet transform and bimodal prediction, the specific process is shown in Fig. 1, including the following steps:

第一步，应用5/3提升小波对心电信号采样值进行分解，具体为将心电信号采样序列X[n]分为奇偶两个序列X[2n+1]和X[2n]，n为大于等于1的正整数，将这两个序列通过公式进行转换为低频系数和高频系数，转换公式表达为：The first step is to use 5/3 lifting wavelet to decompose the ECG signal sampling value, specifically, dividing the ECG signal sampling sequence X[n] into two odd and even sequences X[2n+1] and X[2n], n For a positive integer greater than or equal to 1, the two sequences are converted into low-frequency coefficients and high-frequency coefficients through the formula, and the conversion formula is expressed as:

式中X[2n+1]和X[2n]分别为输入信号X[n]分裂得到的奇偶两个序列，d[n]为提升小波变换得到的高频小波系数序列，s[n]为提升小波变换得到的低频尺度系数序列，在计算高频系数d[n]时，需要同时取得奇序列的值X[2n+1]和前后两个偶序列的值X[2n]和X[2n+2]进行计算，计算低频系数s[n]时，需要同时取得偶序列的值X[2n]和该偶序列值前后两个高频系数d[n]和d[n-1]的值。In the formula, X[2n+1] and X[2n] are the parity sequences obtained by splitting the input signal X[n], respectively, d[n] is the high-frequency wavelet coefficient sequence obtained by lifting wavelet transform, and s[n] is When calculating the high-frequency coefficient d[n] of the low-frequency scale coefficient sequence obtained by lifting the wavelet transform, it is necessary to obtain the value X[2n+1] of the odd sequence and the values X[2n] and X[2n of the two even sequences at the same time. +2] for calculation, when calculating the low-frequency coefficient s[n], it is necessary to obtain the value X[2n] of the even sequence and the values of the two high-frequency coefficients d[n] and d[n-1] before and after the even sequence value at the same time .

第二步，对分解得到的低频系数进行缩放，即对系数做除法，根据不同的应用场景选择不同的除数为2、4、8、16、32，以利于实际电路的实现，对缩放后的值做进一步的平滑操作增大最终的压缩率，对连续的三个小波系数x₁、x₂、x₃，若满足x₁＝x₃且|x₂-x₁|＝1，则使x₂＝x₁。即将中间点视为毛刺，修改该值使其与前后值相等，从而使得小波变换后的系数变化更加平滑，更有利于后续的预测。The second step is to scale the low-frequency coefficients obtained by decomposition, that is, to divide the coefficients, and select different divisors of 2, 4, 8, 16, and 32 according to different application scenarios to facilitate the realization of the actual circuit. The value is further smoothed to increase the final compression rate. For three consecutive wavelet coefficients x₁ , x₂ , x₃ , if x₁ =x₃ and |x₂ -x₁ |=1, then x₂ = x₁ . That is, the middle point is regarded as a burr, and the value is modified to be equal to the value before and after, so that the coefficient change after wavelet transformation is smoother, which is more conducive to subsequent prediction.

第三步，对平滑后的小波系数进行预测，过程如下：The third step is to predict the smoothed wavelet coefficients. The process is as follows:

对非QRS区间采用0阶线性预测的方法进行预测，即对每个小波系数的值，用前一个系数预测，保留预测误差值，也就是当前的系数减去前一个系数得到的差值。The non-QRS interval is predicted by the 0-order linear prediction method, that is, the value of each wavelet coefficient is predicted by the previous coefficient, and the prediction error value is retained, that is, the difference obtained by subtracting the previous coefficient from the current coefficient.

应用R峰检测算法检测出QRS区间的位置，当心电信号采样进入QRS区域时，开始用模板预测和2阶线性预测相结合的方法对低频系数进行预测，模板数量设为N_t，对于采样频率为f的心电信号，对应的QRS区间内低频小波系数的个数为W_qrs＝f/20，因此将模板长度设为W_qrs，也就是每个模板可以预测的系数个数，初始值都为0。The R peak detection algorithm is used to detect the position of the QRS interval. When the ECG signal is sampled into the QRS area, the combination of template prediction and second-order linear prediction is used to predict the low-frequency coefficients. The number of templates is set to N_t , for the sampling frequency For the ECG signal of f, the number of low-frequency wavelet coefficients in the corresponding QRS interval is W_qrs =f/20, so the template length is set to W_qrs , that is, the number of coefficients that can be predicted by each template, and the initial value is is 0.

具体的，预测首先需要在N_t个模板预测器和1个二阶线性预测器中选择一个让QRS区间所有系数预测误差绝对值的和最小的预测器，其中二阶线性预测的公式表达为：Specifically, the prediction first needs to select a predictor that minimizes the sum of the absolute values of the prediction errors of all coefficients in the QRS interval among the N_t template predictors and one second-order linear predictor, where the second-order linear prediction formula is expressed as:

其中x[n-1]、x[n-2]、x[n-3]分别是过去的三个系数值，求得QRS区间内所有W_qrs个系数的预测值，将所有预测的值和原始值做差得到W_qrs个预测误差值，将其求和得道预测误差总和。然后用所有N_t个模板预测器存储的值对当前QRS区间的值进行预测并分别求得N_t个预测误差总和，将这N_t个预测误差总和与二阶线性预测器的误差总和一共N_t+1个值相比较，选取预测误差总和最小的那个预测器，对当前QRS区间系数值进行预测。每轮QRS区间预测完成时，用当前QRS区间的所有系数值更新前面最早更新的一个模板。Where x[n-1], x[n-2], x[n-3] are the past three coefficient values, respectively, to obtain the predicted values of all W_qrs coefficients in the QRS interval, and sum all predicted values to The difference between the original values is obtained to obtain W_qrs prediction error values, which are summed to obtain the sum of the prediction errors. Then use the values stored in all N_t template predictors to predict the value of the current QRS interval and obtain the sum of N_t prediction errors respectively. The sum of these N_t prediction errors and the sum of the errors of the second-order linear predictor is a total of N_t + 1 values are compared, and the predictor with the smallest sum of prediction errors is selected to predict the coefficient value of the current QRS interval. When each round of QRS interval prediction is completed, the earliest updated template is updated with all the coefficient values of the current QRS interval.

第四步，对预测误差编码，首先先将整数预测误差映射为非负整数m，再对映射值进行二级golomb-rice编码，在进行golomb-rice编码前先对参数k的值进行预测，从而获得比较高的编码效率，预测的公式表示为

和

d的初始值设为64，编码方式如下：The fourth step, encoding the prediction error, firstly maps the integer prediction error to a non-negative integer m, and then performs two-level golomb-rice encoding on the mapped value, and predicts the value of the parameter k before performing the golomb-rice encoding. In order to obtain relatively high coding efficiency, the prediction formula is expressed as

and

The initial value of d is set to 64, and the encoding method is as follows:

对于非0值或连续为0数量小于8的数据，对每个数据我们的编码形式为常规golomb-rice编码，根据不同商值得范围在一元编码前加上相应个数的比特“1”。For data with a non-zero value or a number of consecutive zeros less than 8, our encoding form for each data is conventional golomb-rice encoding, and a corresponding number of bits "1" are added before unary encoding according to different quotient value ranges.

2.对于连续为“0”的数量大于等于8的片段，我们对该片段整体进行编码，编码形式为“1110”+该片段数据个数的golomb-rice编码值。2. For the segment with the number of consecutive "0" greater than or equal to 8, we encode the segment as a whole, and the encoding format is "1110" + the golomb-rice encoded value of the number of data in the segment.

其中golomb-rice编码值表示为{一元编码

0，二元编码(m mod 2^k)}。where the golomb-rice encoded value is expressed as {unary encoding

0, binary code (m mod 2^k )}.

第五步，对编码值进行封装，封装的数据包括所有解码需要的信息，具体形式如图3所示,其中t的值为log₂(N_t+1)。首先对前三个小波系数的二进制编码进行封装，解码时可根据前三个值进行后续的预测。然后按照非QRS区间误差编码、QRS指示码、QRS模板索引、QRS区间误差编码的顺序对每个心电周期进行封装，这样可以达到实时编解码。同时在每个非QRS区间和QRS区间存在多个由前缀码“1110”指示的连续0值片段的编码。The fifth step is to encapsulate the encoded value, and the encapsulated data includes all the information required for decoding. The specific form is shown in Figure 3, where the value of t is log₂ (N_t +1). Firstly, the binary codes of the first three wavelet coefficients are encapsulated, and the subsequent prediction can be performed according to the first three values during decoding. Then, encapsulate each ECG cycle in the order of non-QRS interval error coding, QRS indication code, QRS template index, and QRS interval error coding, so that real-time encoding and decoding can be achieved. At the same time, in each non-QRS interval and QRS interval, there are multiple codes of consecutive 0-valued segments indicated by the prefix code "1110".

以上所述，仅为本发明较佳的实施例，但本发明的保护范围并不仅仅局限于此，热火和采用本发明设计的原理，及在本发明基础上做出的非创造性的变化，都属于本发明专利的保护范围内。The above are only preferred embodiments of the present invention, but the protection scope of the present invention is not limited to this. All belong to the protection scope of the patent of the present invention.

Claims (7)

1. An electrocardiosignal compression method based on wavelet transformation and dual-mode prediction is characterized by comprising the following steps:

first step, signal decomposition: performing one-level lifting wavelet transformation on a sampling signal of the electrocardiogram information to obtain high-frequency coefficients and low-frequency coefficients with the same quantity, removing the high-frequency coefficients and reserving the low-frequency coefficients;

step two, scaling and smoothing: carrying out division scaling and smoothing operation on the low-frequency part of the obtained wavelet coefficient;

and step three, prediction: predicting the coefficient by adopting a method of combining linear prediction and template prediction to obtain a prediction error;

step four, encoding: coding the prediction error by applying a secondary golomb-rice coding mode; the process is as follows:

for the prediction error coding, firstly, the integer prediction error is mapped into a non-negative integer m, then secondary golomb-rice coding is carried out on the mapping value, the value of a parameter k is predicted before the golomb-rice coding is carried out, thereby obtaining higher coding efficiency, and the prediction formula is expressed as

And

the initial value of d is 64, and the encoding scheme is as follows:

for the data with the value of non-0 or the continuous 0 and the quantity less than 8, the coding form of each data is the conventional golomb-rice coding, and the range is obtained according to different quotient values and the corresponding number of bits 1 is added before the unitary coding;

for fragments with the number of continuous '0' being more than or equal to 8, the whole fragments are coded, and the coding form is '1110' + the golomb-rice coding value of the data number of the fragments;

wherein the golomb-rice code value is represented as

Fifthly, packaging: the encoded values are packed together with the information needed for prediction to form the final compressed data stream.

2. The wavelet transform and dual-mode prediction based electrocardiosignal compression method as claimed in claim 1, wherein in the first step, the wavelet transform adopts 5/3 lifting wavelet transform, and low-frequency coefficients are retained, and the expression is as follows:

wherein, X [2n +1] and X [2n ] are respectively two odd-even sequences obtained by splitting an input signal X [ n ], d [ n ] is a high-frequency wavelet coefficient sequence obtained by lifting wavelet transformation, s [ n ] is a low-frequency scale coefficient sequence obtained by lifting wavelet transformation, when calculating a high-frequency coefficient d [ n ], a value X [2n +1] of an odd sequence and values X [2n ] and X [2n +2] of two front and back even sequences need to be simultaneously obtained for calculation, and when calculating a low-frequency coefficient s [ n ], a value X [2n ] of an even sequence and values of two high-frequency coefficients d [ n ] and d [ n-1] before and after the even sequence value need to be simultaneously obtained.

3. The wavelet transform and bi-mode prediction based electrocardiographic signal compression method according to claim 1 or 2, wherein in the second step, the wavelet system obtained is subjected to wavelet transformThe division scaling and smoothing operations for the low frequency part are as follows: zooming the low-frequency signal to obtain electrocardiosignals with different signal qualities, smoothing wavelet coefficients, and continuously processing three wavelet coefficients x₁ 、x₂ 、x₃ If x is satisfied₁ ＝x₃ ，|x₂ -x₁ If 1 then x₂ ＝x₁ I.e. the middle point is considered as a spur, the value is modified to be equal to the previous and subsequent values.

4. The wavelet transform and bi-mode prediction based electrocardiosignal compression method as claimed in claim 1 or 2, wherein in the third step, the bi-mode prediction adopts a mode of combining linear prediction and template prediction, 0-order linear prediction is adopted for the non-QRS region of the electrocardiosignal, and template prediction and 2-order linear prediction are adopted for the QRS region.

5. The wavelet transform and dual-mode prediction based electrocardiograph signal compression method according to claim 1, wherein a prefix "1110" is added, and for codes with quotient greater than or equal to 3 and less than 8, a bit "1" is added before unary coding of quotient.

6. The wavelet transform and dual-mode prediction based electrocardiograph signal compression method according to claim 4, wherein the template prediction is: and for the coefficient of each QRS interval, predicting by using the value in the existing QRS template, selecting the closest one from different templates for prediction, simultaneously comparing the value with the value obtained by applying 2-order linear prediction, and selecting the result with the minimum error as the final prediction result.

7. The wavelet transform and dual-mode prediction-based electrocardiosignal compression method according to claim 4, wherein in the fifth step, the prediction error is encoded by applying a secondary golomb-rice coding method, and is packed with information required for prediction to form a final compressed data stream, and the final compressed data stream comprises encoding and data packaging, the prediction error is encoded by adopting a secondary golomb-rice coding obtained by improving a golomb-rice coding algorithm, and a signal encoding value and information required for decoding are packaged together.

Images