JP6344679B2 - Biological information compression method and data compression apparatus - Google Patents

Description

  The present invention relates to a method for compressing biometric information smaller and a data compression apparatus that achieves low power consumption using the method.

  In recent years, in the medical field, not only biological information obtained from short-term examinations performed in hospitals but also biological information that is acquired over a long period of time while living daily life has been regarded as important for patients with chronic diseases such as hypertension. . This is because there may be cases where abnormalities are found in the biological information that has been acquired over a long period of time even if no abnormalities appear in the examination at the hospital. In order to reduce physical and mental burdens on patients due to long-term examinations in daily life, the biological information sensing device used for examinations must be minimally invasive, unconstrained, and unconscious, And it is necessary to be able to do in real time (for example, refer nonpatent literature 1).

  For this reason, the inspection equipment must be small, light and have low power consumption. Examples of the development of biological information sensing devices include capsule-type medical devices that measure biological information such as intravesical pressure data and rectal pressure data and transmit the information to external devices by wireless communication (for example, non-patented) Reference 2).

  Since the inspection device continues to measure biological information continuously for a long time, the amount of biological information is large, and the amount of power consumed for transmitting information to other devices or storing it in the sensing device increases. For this reason, the size of the battery of the device must be increased, and in that case, the device cannot be reduced in size.

  Consider reducing the power consumption during communication to reduce the size of the equipment. Therefore, it is necessary to reduce the amount of information by compressing the biological information to be transmitted. Furthermore, in order to perform a precise inspection, the compression needs to be a lossless compression that can be restored without loss of one bit.

  Various methods have been proposed for compressing biological information, but there are few methods for reversibly compressing biological information. For example, as a reversible compression method, there are miniLZO, Huffman, adaptive Huffman, arithmetic compression, and lexicographic compression (see, for example, Non-Patent Document 3). In addition, those that can be compressed smaller require a large amount of calculation and take a long time to execute, or require a lot of memory, and thus require many restrictions when used as an actual application. For example, most of what is considered as compression of electrocardiographic data is lossy compression and has a large amount of calculation (see Non-Patent Document 4, for example).

Next, the conventional method to implement as a device for performing these compressions is:
(A) Software implementation on a general-purpose processor (hereinafter referred to as conventional method (A))
(B) Implementation by dedicated hardware (ASIC: Application Specific Integrated Circuit) (hereinafter referred to as conventional method (B))
It can be classified into two types.

  FIG. 14 is a diagram illustrating a conventional compression implementation by software on a general-purpose processor. In FIG. 14, there are an instruction memory 101 and a data memory 102 outside the general-purpose processor 100. The instruction memory 101 holds instructions for compression processing. The general-purpose processor 100 performs a compression operation of compression processing according to the instruction read from the instruction memory 101. The general-purpose processor 100 holds data in the data memory 102 or reads the held data through the data bus 103 in accordance with the instruction read from the instruction memory 101. Further, the general-purpose processor 100 stores the operation result performed by the general-purpose processor 100 in the data memory 102 through the data bus 103.

  FIG. 15 is a diagram showing a conventional compression implementation using a general-purpose processor and a dedicated peripheral circuit (ASIC).

  In FIG. 15, there are an instruction memory 111, a data memory 112, and a compression processing circuit 113 outside the general-purpose processor 110. The instruction memory 111 holds an instruction for a compression operation of the compression process. The general-purpose processor 110 reads commands and data held in the data memory 112 through the data bus 114 in accordance with the instructions read out from the instruction memory 111, and performs processing to be executed on the compression processing circuit 113 through the data bus 114. Pass command and data to indicate

  In addition, the general-purpose processor 110 reads the status information and the operation result after the compression processing from the compression processing circuit 113, and stores the status information and the operation result in the data memory 112 through the data bus 114. The compression processing circuit 113 performs a compression operation of compression processing according to the instruction read from the instruction memory 111. Here, the compression processing circuit 113 uses an ASIC.

  The following trade-off exists between the conventional method (A) and the conventional method (B).

  The conventional method (A) has the following advantages (a) and (b) compared to the conventional method (B).

(A) The amount of hardware required for implementation is small (a memory for storing a program is sufficient).
(B) It is possible to flexibly cope with various biological information and specification changes (easily adaptable to changes in the type of biological information, sampling interval, and accuracy).

  However, the conventional method (A) has the following disadvantages (c) and (d) as compared with the conventional method (B).

(C) Long execution time (the number of execution cycles is large).
(D) Power consumption is large.

Masaharu Imai, Yoshinori Takeuchi, "High reliability and low power consumption of wireless communication in medical and healthcare biometric information sensing system", 2012 IEICE General Conference, pp. SS-43-46, March 2012 Hirofumi Iwato, Keishi Sakanushi, Yoshinori Takeuchi, and Masaharu Imai, "A Small-area and Low-power SoC for Less-invasive Pressure Sensing Capsules in Ambulatory Urodynamic Monitoring," IEICE TRANSACTIONS on Electronics, Vol. E95-C, No. 4 , pp. 487-494, 2012 Antti Koshi, "Lossless ECG encoding," Computer Methods and Programs in Biomedicine, ELSEVIER, Vol. 52, pp. 23-33, 997 Ziya Arnavut, "ECG Signal Compression Based on Burrows-Wheeler Transformation and Inversion Ranks of Linear Prediction," IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 54, NO. 3, MARCH 2007

  In the biological information sensing system, low power consumption in various processes is required for long-time use. The portion of the system that consumes the largest amount of power is the portion that transmits and receives biological information with other devices. In order to reduce the power consumption of that portion, it is conceivable to reduce the amount of information by compressing information. In that case, in order to handle biometric information, the compression must be reversible assuming a more precise inspection.

  Since compression also consumes power, it is necessary to perform compression with a low power consumption. When compression is implemented on a general-purpose processor, arithmetic instructions for efficiently calculating compression are not included. Therefore, in order to execute compression on software on a general-purpose processor, a long number of execution cycles is required and power consumption is also low. Become more. However, mounting with dedicated hardware (ASIC) has high performance and low power consumption, but it is difficult to flexibly cope with changes in the type of biological information, sampling interval, and accuracy.

  In other words, we develop a biometric information sensing system that uses high-performance and low-power-consumption packaging methods that can reversibly compress various biometric data with high compression ratio and low computational complexity, flexibly respond to specification changes This is an issue.

  Accordingly, the present invention has been made to solve such a conventional problem, and can reversibly compress biometric data with a high compression rate and a small amount of calculation, and can flexibly adapt to specification changes and various biometric information. An object of the present invention is to provide a method and apparatus for compressing biometric information that can be handled and that consumes less power.

  As a means for achieving the above object, the biometric information compression method and data compression apparatus of the present invention comprise the following arrangement.

によって前記差分値ｄの変換値ｌを得、
前記ステップ３において、

となるｉを求め、ｉの値だけ１を接頭語として並べ、セパレータとして０を挿入し、末尾に、ｋ＋ｉビットで

をバイナリ符号化したものを付け加えることによって前記符号語を得る。 The biometric information compression method is a method for reversibly compressing time series data of biometric information by a data processing device, and calculating a temporal difference value d of the time series data by the data processing device. step 1, by the data processing apparatus, with respect to the difference value d, and step 2 Ru obtain a conversion to convert values l into positive negative value of the difference value d, by the data processing apparatus, wherein said step The conversion value l obtained by the conversion of 2 is encoded using the exponent kolom encoding parameter k (k ≧ 0) determined based on the distribution of the difference values in the exponent golomb encoding. Obtaining a codeword, and in step 2,

To obtain a converted value l of the difference value d,
In step 3 above,

I is obtained, and only the value of i is arranged with 1 as a prefix, 0 is inserted as a separator, and k + i bits at the end

The codeword is obtained by adding a binary encoded version of.

  As a result, the biometric information data is reversibly compressed with a high compression rate and a small amount of calculation, and the biometric information processing that can flexibly cope with the specification change and has high performance and low power consumption can be performed.

によって前記差分値ｄの負値を正値に変換して変換値ｌを得るための専用命令、前記変換で得られた変換値ｌに対して、指数ゴロム符号化を利用するにおいて前記差分値ｄの分布に基づいて決められた指数ゴロム符号化のパラメータｋ（ｋ≧０）を用いて

となるｉを求め、ｉの値だけ１を接頭語として並べ、セパレータとして０を挿入し、末尾に、ｋ＋ｉビットで

をバイナリ符号化したものを付け加える符号化によって符号語を得るための専用命令、前記符号化で得られた符号語を前記データメモリへ格納するための専用命令、及び前記時系列データの未ロード部分があるか否かを判断し、あれば前記時系列データのロード処理へ戻り、なければ終了処理へ分岐するための専用命令を保持する命令メモリと、前記プロセッサ、前記圧縮処理回路及び前記データメモリ間のデータ転送を行うバスとを備え、前記圧縮処理回路は前記プロセッサが発行する前記命令メモリに保持された前記専用命令を含む命令に従って前記時系列データを基に符号化し前記符号語を出力する。 The data compression apparatus is a data compression apparatus for executing the biometric information compression method, and includes a dedicated processor including a processor and a compression processing circuit, time series data of the biometric information, and the time series. A data memory for holding a code word obtained by compressing data, and a dedicated instruction issued by the processor when executing the compression method of the biological information, for loading the time-series data from the data memory Dedicated instruction, and the time difference value d of the time series data,

The dedicated instruction because the negative value of the difference value d is converted to positive values to obtain a conversion value l, the converted value l obtained by the conversion, the difference value in utilizing the Exponential Golomb code by Using a parameter k (k ≧ 0) for exponential Golomb coding determined based on the distribution of d

I is obtained, and only the value of i is arranged with 1 as a prefix, 0 is inserted as a separator, and k + i bits at the end

A dedicated instruction order to obtain the code word by encoding to add a material obtained by binary coding, dedicated instructions for storing the code words obtained by the encoding to the data memory, and unloaded of said time-series data It is determined whether or not there is a part, and if there is, the process returns to the time-series data load process, and if not, the instruction memory holding a dedicated instruction for branching to the end process, the processor, the compression processing circuit, and the data A bus for transferring data between memories, and the compression processing circuit encodes based on the time series data according to an instruction including the dedicated instruction held in the instruction memory issued by the processor and outputs the code word To do.

  As a result, the data of biometric information can be reversibly compressed with a high compression rate and a small amount of calculation, and a data compression apparatus that can flexibly cope with specification changes and has high performance and low power consumption.

によって前記差分値の負値を正値に変換し、

となるｉを求め、ｉの値だけ１を接頭語として並べ、セパレータとして０を挿入し、末尾に、ｋ＋ｉビットで

をバイナリ符号化したものを付け加える符号化によって符号語を得る計算を１命令で実行するための専用命令を保持する命令メモリと、前記プロセッサ、前記圧縮処理回路及び前記データメモリ間のデータ転送を行うバスと、を備え、前記圧縮処理回路は前記プロセッサが発行する前記命令メモリに保持された前記専用命令を含む命令に従って前記時系列データを基に符号化し前記符号化で得られた符号語を出力する。 The data compression device is a data compression device for executing the biometric information compression method, and includes a dedicated processor including a processor and a compression processing circuit, time-series data of the biometric information, and the biometric information. A data memory for storing a codeword obtained by compression, and a dedicated instruction issued by the processor when executing the compression method of the biological information, and calculating a temporal difference value d of the time-series data , to the difference value d, to obtain a conversion value l by converting negative values of the difference values d positive value, the converted value l obtained by the conversion, in utilizing the exponential Golomb coding Using a parameter k (k ≧ 0) of exponent Golomb coding determined based on the distribution of the difference value d

The negative value of the difference value is converted into a positive value by

I is obtained, and only the value of i is arranged with 1 as a prefix, 0 is inserted as a separator, and k + i bits at the end

An instruction memory for holding private instructions for execution one instruction to give Ru calculating a codeword by a coding to add a material obtained by binary coding, the processor, the data transfer between the compression circuit and the data memory The compression processing circuit encodes based on the time-series data according to an instruction including the dedicated instruction held in the instruction memory issued by the processor, and obtains a code word obtained by the encoding Output.

  Thereby, the data of biometric information can be reversibly compressed with a high compression rate and a smaller calculation amount, and a data compression apparatus that can flexibly cope with specification changes and has high performance and low power consumption.

Further, in the data memory, the is held exponential Golomb encoding parameter k, the parameter k in accordance with data attributes that are expressed in at least one type and the sampling interval of the biometric information may be changed .

  Thereby, the compression according to the kind of biological information, the sampling interval of time series data, etc. is attained by changing the said parameter from the outside.

  As a conventional technique, as in Non-Patent Document 3, lexicographic coding is used for compression of electrocardiographic data and the like. However, lexical coding is computationally intensive and takes time to execute. The biometric information compression method using the difference and exponential Golomb coding proposed in the present invention requires a small amount of calculation because one code is assigned to one difference as long as parameters are determined in advance. Less time is required for encoding. Since the biological information sensing system requires processing in real time, it can be said that the method of the present invention is more realistic by implementation.

  In addition, the implementation by the processor to which a dedicated instruction is added proposed in the present invention is that (1) an instruction for obtaining a code word, (2) an instruction for performing positive / negative processing, (3) By adding an instruction for determining a code word and (4) an instruction for assisting storage in a memory, encoding and composite processing are efficiently executed. The features of this device are the same level of processing performance (number of execution cycles) as the conventional method (B) and the ability to execute with a small amount of power consumption. To be achieved.

(A) The figure which shows the time series data of the intravesical pressure, (b) The figure which shows the appearance frequency of the difference of the time series data of the intravesical pressure. (A) The figure which shows the time series data of the rectal pressure, (b) The figure which shows the appearance frequency of the difference of the time series data of the rectal pressure. The figure which shows an example of the code word of exponential Golomb coding. The figure which shows the compression result which performed exponential Golomb encoding with the parameter k = 0-9 about the data of the intravesical pressure. (A) The figure which shows the data of an electrocardiogram, (b) The figure which shows the appearance probability of the difference of the data of an electrocardiogram. The figure which shows the comparison result of the compression rate of exponent Golomb coding and the other lossless compression method about the data of an intravesical pressure. The figure which shows the periodicity of the data of an electrocardiogram. The figure of the data compression device concerning the present invention. The figure which shows the processing flow of the part which concerns on the compression method of biometric information based on this invention. The figure which shows the processing flow of Example 1 which mounted the part which concerns on the compression method of biometric information based on this invention in the data compression apparatus. The figure which shows the processing flow of Example 2 which mounted the part which concerns on the compression method of biometric information based on this invention in the data compression apparatus. The figure which shows the block diagram which comprises the dedicated command of Example 2 which mounted the part which concerns on the compression method of biometric information based on this invention in the data compression apparatus. The figure which shows another block diagram which comprises the dedicated command of Example 2 which mounted the part which concerns on the compression method of biometric information based on this invention in the data compression apparatus. The figure of the conventional biological information processing apparatus by the software on a general purpose processor. The figure of the conventional biometric information processing apparatus using a general purpose processor and a dedicated peripheral circuit (ASIC).

  An embodiment of the present invention will be described with reference to FIGS.

  Biometric data such as intravesical pressure data, rectal pressure data, pulse waveform, respiratory waveform, myoelectricity, and electrocardiogram are time-series data and change with time. The inventor of the present application is the time series data of the biological information is data that changes drastically as shown in FIGS. 1 (a) and 2 (a), but looking at the change from the immediately preceding value, We noticed that there are few large changes, and taking the difference before encoding results in an increase in the effect of compression encoding.

  In fact, as shown in FIGS. 1 (b) and 2 (b), in the data of intravesical pressure and intrarectal pressure, those measured at a sampling interval of 120 [ms] and accuracy of 10 [bit] The frequency at which the value becomes 0 is about 50%, and since the data bias is larger than when the acquired data itself is handled, the data can be effectively compressed. In other words, the exponent Golomb code can be compressed to a smaller extent when data with a large bias is encoded, so that the application of the exponent Golomb coding is effective. A procedure for applying exponential Golomb coding as a lossless compression method will be described below.

  First, since a positive number and a negative number appear, the difference value d is converted by the calculation of the following (formula 1), and the value l to be exponential Golomb coding is set to all positive numbers or 0. It is necessary to obtain.

  The exponent Golomb-encoded value is obtained by the following (Step 1) to (Step 4).

となるｉを求める。 (Step 1)

I is obtained.

  (Step 2) 1 is arranged as a prefix only for the value of i.

  (Step 3) 0 is inserted as a separator.

をバイナリ符号化したものを付け加える。ここでパラメータｋは、生体情報に応じて圧縮率を最適化する指数ゴロム符号のパラメータである。 (Step 4) At the end, with k + i bits

Add the binary encoded version. Here, the parameter k is a parameter of an exponential Golomb code that optimizes the compression rate according to the biological information.

  As an example, when the difference is 1, a procedure for encoding with an exponential Golomb code (k = 0) will be described. The value l to be exponential Golomb encoded is 1. It can be seen that i is 1 in encoding (Step 1), and the upper 2 bits of the code word are 10 in (Step 2) and (Step 3). In the encoding (Step 4), adding a binary encoded 0 to 1 at the end adds 100. Therefore, the code word of difference = 1 is 100.

  FIG. 3 shows an example of a code word for exponential Golomb coding for l = 0 to 9 and k = 0, 1, 2 obtained in the same manner.

  In addition, by changing the parameter k, compression corresponding to any change in biological information can be performed. For example, as shown in FIG. 1 and FIG. 2, the data of the intravesical pressure and the intrarectal pressure measured with a sampling interval of 120 [ms] and an accuracy of 10 [bit] has a value of about 50% of the immediately preceding difference. The appearance frequency is 0. Furthermore, most of the difference values other than 0 are distributed before and after the difference value. Therefore, a high compression ratio can be obtained by using an exponential Golomb code k = 0 for the difference.

  FIG. 4 shows a compression result obtained by performing exponential Golomb coding with the parameter k = 0 to 9 for the data of the intravesical pressure. It can be confirmed that the data of the intravesical pressure can be minimized when k = 0. For compressing data of intravesical pressure / intrarectal pressure, if the parameter of the exponent Golomb code is set to k = 0, the amount of data can be reduced to about 70%. Similarly, from the measured ECG data shown in FIG. 5 and the appearance frequency of the immediately preceding difference, it is confirmed that the highest compression rate is obtained when k = 3.

  Exponential Golomb coding by taking the difference from the data immediately before the biometric information can be compressed smaller than miniLZO or the like actually used in medical applications, and can be executed earlier than lexicographic coding. FIG. 6 shows a comparison result of the compression rate with other lossless compression encoding methods for the bladder pressure data. The average value is represented by a bar graph, and the standard deviation from the average value is represented by a vertical bar centered on the average value. The result of using the exponential Golomb coding method is the highest compression ratio with an average compression ratio of 74.3%.

In the above, an example of taking a difference from the immediately preceding value has been given, but periodic waveform data such as an electrocardiogram (for example, data of the region 251 and the region 252) has autocorrelation as shown in FIG. It is also possible to take a difference for each period at the interval of the wave-R wave. Since the position of the R wave (for example, circles R ₁ , R ₂ , R _{3 in} FIG. 7) can be easily found, one cycle from the R wave to the next R wave is stored, and the next R wave is stored. When appears, the difference from the previous cycle is taken.

For example, the region 251 and region 252, expressed from R-wave difference values for the n-th sample of the data _(the difference between the _R 2 + n), _(the difference between the _{R 2 + n) = (R} 2 + n samples values ) − (R ₁ + n sample value). However, it is _{0 ≦ n ≦ R 3 -R 2} . Thereafter, by performing exponential Golomb coding with appropriate parameters, compression can be made smaller.

  Dictionary compression is used for ECG data compression and the like. However, lexical coding is computationally intensive and takes time to execute. The biometric information compression method using the difference and exponential Golomb coding proposed in the present invention requires a small amount of calculation because one code is assigned to one difference as long as parameters are determined in advance. Less time is required for encoding. Since the biological information sensing system requires processing in real time, it can be said that the method of the present invention is more realistic by implementation.

  The above-described biometric information compression method using differential and exponential Golomb coding can be implemented, for example, by software in a conventional general-purpose processor, or by dedicated hardware (ASIC), and calculation of biometric information compression processing. Can be implemented by a processor for a specific application field (ASIP: Application Specific Instruction-set Processor) having an instruction that can be executed with a small number of execution cycles.

  A data compression apparatus using a processor for specific application fields (ASIP) will be described below as one embodiment. Thereby, the compression process of biometric information can be performed efficiently and with a small amount of power consumption.

  FIG. 8 shows a main part related to the compression processing of the data compression apparatus 1.

  In FIG. 8, a dedicated processor 10 which is a processor for specific application fields (ASIP) includes a general-purpose processor 11 and a compression processing circuit 12, which are connected by an internal bus 15. An instruction memory 13 is provided outside the dedicated processor 10, and instructions are read from the general-purpose processor 11 and executed. Data is stored in the data memory 14.

  The data read from the data memory 14 through the data bus 16 is input to the dedicated processor 10 as required for the instruction to be executed and used for the operation of the compression processing circuit 12, etc. Save in the memory 14. That is, the compression processing circuit 12 may encode based on the time-series data according to an instruction including a dedicated instruction held in the instruction memory 13 issued by the general-purpose processor 11 and output the code word.

  Data that is written to or read from the data memory 14 from / to the outside is exchanged with the outside of the data compression apparatus 1 via the data bus 16 through an input / output interface (not shown).

  Examples of the general-purpose processor 11 include a RISC (Reduced Instruction Set Computer) and a CISC (Complex Instruction Set Computer).

  FIG. 9 shows a processing flow diagram of encoding including (Step 1) to (Step 4) in the compression processing. The processing flow of FIG. 9 shows a procedure independent of the implementation of the biometric information compression method using difference and exponential Golomb coding. The processing flow of FIG. 9 is executed by any data processing device such as a general-purpose processor, dedicated hardware (ASIC), or application specific processor (ASIP).

  In step 260, time-series data is pre-processed. In step 261, sample values constituting the pre-processed time-series data are loaded. In step 262, the difference between the sample values, that is, the time-series data Processing for calculating a temporal difference value is performed. In step 263, positive / negative processing of the difference value is performed according to the above equation 1 to obtain a value l to be exponential Golomb encoded, and exponent Golomb encoding is performed in Step 264. Processing for obtaining i in (Step 1) from the power value l is performed, processing for obtaining the code words of (Step 2) to (Step 4) is performed in Step 265, and the code word obtained in Step 266 is stored in the buffer register. In step 267, it is determined whether there is an unloaded sample value. A branch processing to go to end the process of step 268 if Kere.

  Prior to the processing of FIG. 9, the optimum parameter k for compression is determined according to the data attributes such as the type of biological information and the sampling interval of the time series data. For example, the value of the parameter k may be stored in the data memory 14 so that it can be changed from the outside.

  When the processing flow of FIG. 9 is implemented in the data compression apparatus 1, a dedicated processor 10 that is a processor for a specific application field (ASIP) is used to reduce the number of execution cycles to a general-purpose processor having a simple instruction set. A dedicated instruction may be added to perform the encoding process. Such Example 1 is shown in FIG. In FIG. 10, the steps for processing with the added dedicated instruction are as follows.

In step 271, loading a preprocessed sample value by executing the dedicated instruction LDSAM,
In step 273, processing of the positive / negative difference value by execution of the dedicated instruction SIGN,
In step 275, codeword calculation by execution of dedicated instructions CODE1, CODE2,
In step 276, the dedicated instructions RES1 and RES2 are executed and stored in the buffer register.
In step 277, branch processing with end determination by execution of the dedicated instruction BRENC.

  Here, the reason why the two dedicated instructions CODE1 and CODE2 are required in step 275 is that the number of registers that can be accessed at one time is limited because they are implemented as extended instructions in the RISC processor. The reason why two dedicated instructions RES1 and RES2 are required in step 276 is because processing is required when the buffer space is smaller than the codeword length and processing when the codeword length space is in the buffer. .

  This makes it possible to reduce the total number of execution cycles by approximately 28% and the power consumption by approximately 27%, compared with the conventional RISC processor in which dedicated instructions are not added to the above-mentioned intravesical pressure and intrarectal pressure data. It was.

  FIG. 11 shows a second embodiment in which steps 262, 263, 264, and 265 shown in FIG. 9 are realized by one dedicated instruction. In this case, since the steps can be executed together, the number of execution cycles is further reduced and the power consumption is also reduced.

  In FIG. 11, the steps for processing with the added dedicated instruction are as follows.

In step 271, loading a preprocessed sample value by executing the dedicated instruction LDSAM,
In step 282, execution of processing equivalent to steps 262, 263, 264, 265 by execution of the added one dedicated instruction;
A process of storing the obtained code word in a buffer register in step 266;
In step 277, branch processing with end determination by execution of the dedicated instruction BRENC.

  FIG. 12 is a block diagram when configuring this dedicated instruction. By using the circuit of FIG. 12, it is possible to convert a numerical value into an exponential Golomb code with one instruction.

  FIG. 13 is a block diagram showing the decoding process. By using a dedicated instruction that employs this circuit, it is possible to return data that has been exponentially Golomb-encoded with one instruction to the original value.

  The biometric information compression method using differential and exponential Golomb coding and the implementation thereof have been described above. The characteristics of each implementation are compared and shown in the table below. Implementation (C) achieves the same flexibility and expandability as software processing on a general-purpose processor while having the same processing performance (number of execution cycles) and execution capacity with less power consumption as implementation (B). Is done.

  As described above, the biometric information compression method and the data compression apparatus for executing the method according to a plurality of aspects of the present invention have been described based on the embodiment. However, the present invention is limited to this embodiment. It is not a thing. Unless departing from the spirit of the present invention, one or more of the present invention may be implemented by various modifications conceived by those skilled in the art in this embodiment or by combining components in different embodiments. It may be included within the scope of the embodiments.

  The present invention is useful for a biological information sensing system that requires downsizing and long-time use.

DESCRIPTION OF SYMBOLS 1 Data compression apparatus 10 Dedicated processor 11, 100, 110 General-purpose processor 12, 113 Compression processing circuit 13, 101, 111 Instruction memory 14, 102, 112 Data memory 15 Internal bus 16, 103, 114 Data bus

Claims (4)

A method for reversibly compressing time-series data of biological information by a data processing device,
Calculating a temporal difference value d of the time-series data by the data processing device; and
Wherein the data processing apparatus, with respect to the difference value d, and Step 2 Ru obtain a conversion value l by converting negative values of the difference value d to positive,
Exponential Golomb coding parameter k (k) determined based on the distribution of the difference values when exponential Golomb coding is used for the transformation value l obtained by the transformation in Step 2 by the data processing device. Step 3 to obtain a codeword by encoding using ≧ 0) ;
Only including,
In step 2,

To obtain a converted value l of the difference value d,
In step 3 above,

I is obtained, and only the value of i is arranged with 1 as a prefix, 0 is inserted as a separator, and k + i bits at the end

To obtain the codeword by adding the binary encoding of
A method for compressing biological information. A data compression apparatus for executing the biometric information compression method according to claim 1,
A dedicated processor having a processor and a compression processing circuit;
A data memory for holding the time series data of the biological information and a code word obtained by compressing the time series data;
A dedicated instruction issued by the processor when the biometric information compression method is executed,
A dedicated instruction for loading the time series data from the data memory;
For the temporal difference value d of the time series data,

Dedicated instructions order to obtain a conversion value l are converted to positive negative value of the difference value d by,
The exponent Golomb coding parameter k (k ≧ 0) determined based on the distribution of the difference value d in using the exponent Golomb coding for the transformation value l obtained by the transformation is used.

I is obtained, and only the value of i is arranged with 1 as a prefix, 0 is inserted as a separator, and k + i bits at the end

Special instruction of order to obtain a code word by the encoding to add a material obtained by binary encoding,
A dedicated instruction for storing the code word obtained by the encoding in the data memory, and
Determining whether or not there is an unloaded portion of the time-series data, and if there is, return to the time-series data load process, and if not, a dedicated instruction for branching to the end process
An instruction memory that holds
A bus for transferring data between the processor, the compression processing circuit, and the data memory;
The data compression apparatus, wherein the compression processing circuit encodes based on the time-series data according to an instruction including the dedicated instruction held in the instruction memory issued by the processor and outputs the code word. A data compression apparatus for executing the biometric information compression method according to claim 1,
A dedicated processor having a processor and a compression processing circuit;
A data memory for holding time-series data of the biological information and a code word obtained by compressing the biological information;
A dedicated instruction issued by the processor when executing the biometric information compression method, calculating a temporal difference value d of the time-series data, and calculating the difference value d with respect to the difference value d converts negative values to positive to obtain a conversion value l, the converted value l obtained by the conversion, the index was determined based on the distribution of the difference values d in using exponential Golomb coding Golomb Using the encoding parameter k (k ≧ 0)

The negative value of the difference value is converted into a positive value by

I is obtained, and only the value of i is arranged with 1 as a prefix, 0 is inserted as a separator, and k + i bits at the end

An instruction memory for holding private instructions for execution one instruction to give Ru calculating a codeword by a coding to add a material obtained by binary coding,
A bus for transferring data between the processor, the compression processing circuit, and the data memory;
The data compression apparatus, wherein the compression processing circuit encodes based on the time-series data according to an instruction including the dedicated instruction held in the instruction memory issued by the processor and outputs a code word obtained by the encoding. Wherein the data memory, the is held parameter k exponential Golomb coding, the parameter k can be changed in accordance with at least one data attribute represented by the type and the sampling interval of the biometric information according to claim 2, wherein Item 4. The data compression device according to Item 3.

Images