Method of compressing waveform data with differential entropy based compression

• US 7,429,938 B1
• Filed: 03/26/2007
• Issued: 09/30/2008
• Est. Priority Date: 03/26/2007
• Status: Active Grant

• Current Assignee: Medtronic Incorporated (Medtronic Plc)
• Original Assignee: Medtronic Incorporated (Medtronic Plc)
• Inventors: Corndorf, Eric D.
• Primary Examiner(s): YOUNG, BRIAN K
• Application Number:
  US11/691,082
• Publication Number:
  US 20080238749A1
• Time in Patent Office: 554 Days
• Field of Search: 341/67, 341/106, 341/107, 341/155
• US Class Current: 341/67
• CPC Class Codes: H03M 7/40 Conversion to or from variable length codes, e.g. Shannon-Fano code, Huffman code, Morse code

Abstract

Waveforms are digitally sampled and compressed for storage in memory. The compression of the data includes generating a truncated entropy encoding map and using the values within the map to obtain good compression. An encoder further sub-selects values to be encoded and values to remain unencoded to provide an overall compression of the data.

Claims

1. A method comprising:

sampling an analog waveform and converting the waveform into a digitized data set;

generating an encoded waveform data point for each of an equivalent data point from a subset of the digitized data with a predictive encoding process; and

substituting a codeword for an encoded waveform data point using an entropy encoding process if a predetermined condition is met, wherein the codeword is stored into a memory when the predetermined condition is met or the encoded waveform data point is stored into the memory when the predetermined condition is not met.

2. The method claim 1, wherein the predictive encoder is a linear predictive encoder.

3. The method of claim 2, wherein the linear predictive encoder calculates a difference between a predicted value and a measured value and stores the difference as the encoded waveform data point.

4. The method of claim 3, wherein the expected value for a given data point is the measured value of an immediately preceding data point.

5. The method of claim 1, wherein the subset does not include an initial digitized data point of a series and the initial digitized data point is stored in memory as an unencoded value.

6. The method of claim 5, wherein digitized data points are selected at a predetermined interval and stored in an unencoded form.

7. The method of claim 1, further comprising a truncated entropy probability tree data set stored in memory and including the codewords for selected encoded waveform data point values.

8. The method of claim 7, wherein the data set further includes an exit code codeword so that the entropy encoding process stores the exit code codeword and the encoded digital value when the predetermined condition is met and the encoded digital value does not have a codeword stored within the entropy probability data set.

9. The method of claim 8, wherein the entropy probability data set consists of codewords for the exit code codeword, a value of zero, and twice as many values as a bit resolution value.

10. The method of claim 1, further comprising a rule processor configured to determine whether the predetermined condition has been met for a given data point by evaluating the encoded value of an immediately prior encoded data point.

11. The method of claim 10, wherein the predetermined condition is whether the immediately prior encoded data point should have been encoded.

12. The method of claim 11, wherein a given data point should have been encoded if its value is within an optimized waveform band.

13. The method of claim 12, wherein the optimized waveform band includes all values within the truncated entropy probability data set.

14. A method comprising:

sampling an analog signal in an implantable medical device (IMD) and generating digitized data;

generating encoded waveform data using a predictive encoding process for each of an equivalent data point from a subset of the digitized data;

replacing an encoded waveform data point with a codeword using an entropy encoding when a predetermined condition is met, wherein the codeword is stored into a device memory when the predetermined condition is met or the encoded waveform data point is stored into the device memory when the predetermined condition is not met;

transmitting data from the IMD to an external receiver having a decoder; and

decoding the data received in the external receiver by converting any codeword into an encoded value and converting encoded values into the digitized data.

15. The method claim 14, wherein the predictive encoding process further includes calculating a difference between a predicted value and a measured value and storing the difference as the encoded waveform data point and the decoder adds the difference to the predicted value to decode the data.

16. The method of claim 15, wherein the expected value for a given data point is the measured value of an immediately preceding data point.

17. The method of claim 14, wherein the subset does not include an initial digitized data point of a series and the initial digitized data point is stored in memory as an unencoded value and the decoder decodes subsequent data points based upon the initial digitized data point.

18. The method of claim 14, further comprising a truncated entropy probability tree data set stored in the IMD memory and in the device memory and including the codewords for selected encoded waveform data point values.

19. The method of claim 18, wherein the data set further includes an exit code codeword so that the entropy encoding process stores the exit code codeword and the encoded digital value when the predetermined condition is met and the encoded digital value does not have a codeword stored within the entropy probability data set and the decoder recognizes the exit code codeword and process the data immediately following the exit code codeword as the encoded data.

20. The method claim 14, further comprising a first rule processor operating within the IMD and a second rule processor operating within the encoder, wherein each rule processor is configured to determine whether the predetermined condition has been met for a given data point by evaluating the encoded value of an immediately prior encoded data point.