Block floating point compression of signal data

• US 8,301,803 B2
• Filed: 10/23/2009
• Issued: 10/30/2012
• Est. Priority Date: 10/23/2009
• Status: Active Grant

• Current Assignee: Altera Corporation (Intel Corporation)
• Original Assignee: Samplify Systems Incorporated
• Inventors: Wegener, Albert W.
• Primary Examiner(s): Serrao, Ranodhi
• Application Number:
  US12/605,245
• Publication Number:
  US 20110099295A1
• Time in Patent Office: 1,103 Days
• Field of Search: 709/247, 708/203, 713/172, 380/229
• US Class Current: 709/247
• CPC Class Codes: G06F 7/483 Computations with numbers represented by a non-linear combination of denominational numbers, e.g. rational numbers, logarithmic number system or floating-point numbers G06F7/4806, G06F7/4824, G06F7/49, G06F7/491, G06F7/544 take precedence
• CPC Class Codes: H03M 7/24 Conversion to or from floating-point codes
• CPC Class Codes: H03M 7/40 Conversion to or from variable length codes, e.g. Shannon-Fano code, Huffman code, Morse code
• CPC Class Codes: H03M 7/46 Conversion to or from run-length codes, i.e. by representing the number of consecutive digits, or groups of digits, of the same kind by a code word and a digit indicative of that kind

Abstract

A method and apparatus for compressing signal samples uses block floating point representations where the number of bits per mantissa is determined by the maximum magnitude sample in the group. The compressor defines groups of signal samples having a fixed number of samples per group. The maximum magnitude sample in the group determines an exponent value corresponding to the number of bits for representing the maximum sample value. The exponent values are encoded to form exponent tokens. Exponent differences between consecutive exponent values may be encoded individually or jointly. The samples in the group are mapped to corresponding mantissas, each mantissa having a number of bits based on the exponent value. Removing LSBs depending on the exponent value produces mantissas having fewer bits. Feedback control monitors the compressed bit rate and/or a quality metric. This abstract does not limit the scope of the invention as described in the claims.

Claims

1. A method for compressing a plurality of signal samples, where the signal samples consist of a number of bits per sample, comprising:

grouping the plurality of signal samples received in a processor into a sequence of encoding groups;

for an encoding group in the sequence of encoding groups, determining a block exponent value for the encoding group, and determining a mantissa for each signal sample in the encoding group, the mantissa having a number of bits based on said block exponent value for the encoding group;

for the sequence of encoding groups, encoding the block exponent values for the encoding groups to determine exponent tokens for the encoding groups in the sequence, wherein an exponent token represents one or more block exponent values for one or more encoding groups in the sequence, including determining differences between block exponent values for consecutive encoding groups in the sequence;

encoding the determined differences to produce the exponent tokens for at least a portion of the consecutive encoding groups;

providing a predetermined set of exponent token codes, the exponent token codes in the predetermined set encoding respective differences between block exponent values in more than two consecutive encoding groups;

if the determined differences in the block exponent values for more than two consecutive encoding groups in the sequence match one of the exponent token codes in the predetermined set, then representing the block exponent values for more than one of the encoding groups in the consecutive encoding groups using said one of the exponent token codes; and

encoding the plurality of signal samples using the exponent tokens and the mantissas for the sequence of encoding groups to form compressed data for storage in a storage device or transmission from the processor.

2. The method of claim 1, including forming a compressed data set by arranging the exponent tokens for individual encoding groups adjacent to the mantissas arranged in series for the signal samples in the individual encoding groups.

3. The method of claim 1, including forming a compressed data set by arranging the exponent tokens in series for a plurality of encoding groups in the sequence and arranging the mantissas in series for the signal samples in the plurality of encoding groups.

4. The method of claim 1, wherein said block exponent value is based on a place value of a most significant non-zero bit in a signal sample having a maximum magnitude within the encoding group.

5. The method of claim 1, wherein the predetermined set of exponent token codes further includes exponent token codes to encode respective differences between block exponent values in two consecutive encoding groups, wherein said encoding the block exponent values further comprises:

if the determined difference between the block exponent values for two consecutive encoding groups in the sequence matches one of the exponent token codes in the predetermined set, then representing the block exponent value for one of the encoding groups in the two consecutive encoding groups using said one of the exponent token codes.

6. The method of claim 1, wherein said determining a mantissa includes specifying a number of least significant bits (LSBs) for removal from signal samples in the encoding group, and wherein the mantissas for signal samples in the encoding group comprise a number of bits equal to the block exponent value for the encoding group less the specified number of least significant bits.

7. The method of claim 6, wherein said specifying a number of LSBs is specified based on the block exponent value for the encoding group.

8. The method of claim 6, wherein said specifying a number of least significant bits is specified based on a desired bit rate for the compressed data.

9. The method of claim 6, wherein said specifying a number of least significant bits is specified based on a desired quality metric for the compressed data.

10. The method of claim 1, wherein each signal sample includes a sign bit, wherein said determining a mantissa includes appending the sign bit of each signal sample with the determined mantissa.

11. The method of claim 1, including using the processor in a signal processing system to perform the steps of a claim, the signal processing system including an analog to digital converter that receives an analog signal, the method further comprising:

converting the analog signal to the plurality of signal samples provided for the step of grouping.

12. The method of claim 1, including receiving the plurality of signal samples using a receiver in a data communication channel, the receiver providing the signal samples to the processor.

13. The method of claim 1, wherein the plurality of signal samples is stored in a data storage system in communication with the processor, the processor reading the plurality of signal samples from the data storage system.

14. A method for decompressing an input signal conveying compressed data to a processor representing a plurality of encoded original signal samples, the method comprising:

disassembling the compressed data received at the processor to obtain a plurality of exponent tokens and a plurality of compressed groups of mantissas, where the plurality of the encoded original signal samples are represented by the exponent tokens and the mantissas;

decoding the plurality of exponent tokens to form a plurality of block exponent values, wherein each block exponent value is associated with one of the plurality of compressed groups, wherein a portion of the plurality of exponent tokens encodes differences between block exponent values associated with consecutive compressed groups, the decoding the plurality of exponent tokens further comprising;

providing a predetermined set of exponent token codes, the exponent token codes in the predetermined set encoding respective differences between block exponent values associated with consecutive compressed groups;

for a given exponent token, using the predetermined set to decode one or more differences represented by the given exponent token;

for each decoded difference, adding the decoded difference to a previous block exponent value to determine the block exponent value associated with a corresponding one of the consecutive compressed groups;

for each of the compressed groups of mantissas, determining a number of bits representing each of the mantissas in the compressed group using the associated block exponent value, and mapping the number of bits of each mantissa to a corresponding decompressed sample, to form a group of decompressed samples; and

applying said determining a number of bits and said mapping the number of bits of each mantissa to the plurality of compressed groups of mantissas to generate a plurality of decompressed samples.

15. The method of claim 14, wherein said determining a number of mantissa bits further comprises:

determining a number of least significant bits (LSBs) based on the associated block exponent value, wherein said mapping includes inserting a bit value of zero or a dithered bit value for the number of LSBs of the corresponding decompressed sample.

16. The method of claim 15, wherein the input signal includes a compression control parameter, the step of determining a number of LSBs further comprises:

receiving the compression control parameter from the input signal;

selecting a quantization profile in accordance with the compression control parameter from a plurality of quantization profiles stored in a local memory in communication with the processor, wherein each quantization profile comprises a look-up table or a formula relating respective block exponent values to respective numbers of LSBs; and

using the selected quantization profile to determine the number of LSBs based on the associated block exponent value.

17. The method of claim 14, including using a data processor to perform said disassembling, decoding, determining, mapping and applying.

18. The method of claim 14, including receiving the input signal using a receiver in a data communication channel, the receiver providing the input signal to the processor.

19. The method of claim 14, including retrieving the input signal from a data storage system in communication with the processor.

20. An apparatus for compressing a plurality of signal samples, where the signal samples consist of a number of bits per sample, comprising a processor and a memory in communication with the processor, the processor including:

logic for grouping the signal samples into a sequence of encoding groups, to form a sequence of encoding groups;

logic determining a block exponent value for each encoding group;

a mantissa encoder that receives the signal samples in each encoding group and forms a mantissa for each signal sample, wherein each mantissa of a particular encoding group has a number of bits based on the block exponent value for the particular encoding group;

an exponent encoder that encodes the block exponent values for the sequence of encoding groups to produce exponent tokens, wherein an exponent token represents one or more block exponent values for one or more encoding groups, the exponent encoder further comprising;

a difference calculator determining differences between the block exponent values associated with consecutive compressed groups, the exponent encoder encoding the determined differences to produce corresponding exponent tokens;

a predetermined set of exponent token codes stored in the memory, the exponent token codes in the predetermined set encoding respective differences between block exponent values for more than two consecutive encoding groups;

logic determining if the determined differences between block exponent values for more than two consecutive encoding groups in the sequence match one of the exponent token codes in the predetermined set, and if so, selecting said one of the exponent token codes to represent the block exponent values for more than one of the encoding groups in the consecutive encoding groups; and

a bit packer arranging the exponent tokens and the mantissas for the sequence of encoding groups to form compressed data for storage in a storage device or transmission from the processor.

21. The apparatus of claim 20, wherein the bit packer further comprises:

a multiplexer that arranges the exponent tokens for individual encoding groups adjacent to the mantissas arranged in series for the signal samples in the individual encoding groups to form a compressed data set.

22. The apparatus of claim 20, wherein the bit packer further comprises:

a multiplexer that arranges the exponent tokens in series for a plurality of encoding groups in the sequence and arranges the mantissas in series for the signal samples in the plurality of encoding groups to form a compressed data set.

23. The apparatus of claim 20, wherein the block exponent value is based on a place value of a most significant non-zero bit in a signal sample having a maximum magnitude within the encoding group.

24. The apparatus of claim 20, wherein the predetermined set of exponent token codes further includes exponent token codes to encode respective differences between block exponent values for two consecutive encoding groups, wherein the exponent encoder further comprises logic determining if the determined difference between the block exponent values for two consecutive encoding groups in the sequence matches one of the exponent token codes in the predetermined set, and if so, selecting said one of the exponent token codes to represent the block exponent value for one of the encoding groups in the two consecutive encoding groups.

25. The apparatus of claim 20, wherein the mantissa encoder specifies a number of least significant bits (LSBs) for removal from signal samples in each encoding group, wherein each mantissa of a particular encoding group has a number of bits equal to the block exponent value less the specified number of least significant bits.

26. The apparatus of claim 25, the mantissa encoder further including a quantization profile stored in the memory, the quantization profile comprising a look-up table or a formula relating respective block exponent values to respective numbers of LSBs, the mantissa encoder using said quantization profile to specify the number of LSBs based on the block exponent value for the encoding group.

27. The apparatus of claim 26, further comprising a compression controller providing a compression control parameter to the mantissa encoder, wherein a plurality of quantization profiles are stored in the memory, the mantissa encoder selecting one of the quantization profiles in accordance with the compression control parameter.

28. The apparatus of claim 27, wherein the compression controller determines the compression control parameter based on a desired bit rate for the compressed data.

29. The apparatus of claim 27, wherein the compression controller determines the compression control parameter based on a desired quality metric for the compressed data.

30. The apparatus of claim 20, further comprising:

an analog to digital converter coupled to receive an analog signal and converting the analog signal to the plurality of signal samples provided to the processor.

31. An apparatus for decompressing an input signal conveying compressed data representing a plurality of encoded original signal samples, comprising a processor and a memory in communication with the processor, the processor including:

a buffer receiving the compressed data, wherein the plurality of encoded original signal samples are represented by a plurality of exponent tokens and a plurality of compressed groups of mantissas;

logic coupled to the buffer for disassembling the compressed data to obtain the plurality of exponent tokens and the plurality of compressed groups of mantissas;

an exponent decoder receiving the plurality of exponent tokens and determining a plurality of block exponent values, wherein each block exponent value is associated with one of the compressed groups of mantissas, wherein a portion of the plurality of exponent tokens encode differences between block exponent values associated with consecutive compressed groups, the exponent decoder further comprising;

a predetermined set of exponent token codes stored in the memory, the exponent token codes in the predetermined set to encode respective differences between block exponent values for consecutive compressed groups, the exponent decoder using the predetermined set to decode one or more differences represented by a received exponent token to obtain one or more decoded differences;

an adder receiving each decoded difference and a respective previous block exponent value and producing the block exponent value for a corresponding one of the consecutive compressed groups; and

a mantissa decoder receiving each compressed group of mantissas and the associated block exponent value, wherein a number of bits representing each of the mantissas in the compressed group is based on the associated block exponent value, the mantissa decoder mapping the number of bits of each mantissa to a corresponding decompressed sample to form a group of decompressed samples, the mantissa decoder decoding the plurality of compressed groups to produce a plurality of decompressed samples.

32. The apparatus of claim 31, wherein the mantissa decoder determines a number of least significant bits (LSBs) based on the associated block exponent value and inserts a bit value of zero or a dithered bit value for the number of LSBs of the corresponding decompressed sample.

33. The apparatus of claim 32, wherein the input signal includes a compression control parameter, the apparatus further comprising:

a decompression controller receiving the compression control parameter from the input signal and providing the compression control parameter to the mantissa decoder; and

a plurality of quantization profiles stored in the memory coupled to the mantissa decoder, wherein each quantization profile comprises a look-up table or a formula relating respective block exponent values to respective numbers of LSBs, the mantissa decoder selecting one of the quantization profiles in accordance with the compression control parameter, the mantissa decoder using the selected quantization profile to determine the number of LSBs based on the associated block exponent value.

34. The apparatus of claim 31, further including:

a receiver coupled to receive the input signal from a data communication channel and providing the compressed data to the buffer.

35. The apparatus of claim 31, further including:

a data storage system storing the input signal and coupled to provide the compressed data to the buffer.