Low delay low complexity lossless compression system

• US 9,542,839 B2
• Filed: 06/25/2013
• Issued: 01/10/2017
• Est. Priority Date: 06/26/2012
• Status: Active Grant

• Current Assignee: BTS Software Solutions
• Original Assignee: BTS Software Solutions LLC (Berico Tailored Systems LLC)
• Inventors: Li, Dunling
• Primary Examiner(s): Tran, Khanh C
• Application Number:
  US13/927,043
• Publication Number:
  US 20130342357A1
• Time in Patent Office: 1,295 Days
• Field of Search: 375/240.01, 375/240.02, 375/240.08, 375/240.12, 375/240.23, 375/240.24, 375241-242, 375/246
• US Class Current: 1/1
• CPC Class Codes: G08C 19/00 Electric signal transmission systems G08C17/00 takes precedence
• CPC Class Codes: H03M 7/30 Compression speech analysis-synthesis for redundancy reduction G10L19/00; for image communication H04N; Expansion; Suppression of unnecessary data, e.g. redundancy reduction
• CPC Class Codes: H03M 7/3059 Digital compression and data reduction techniques where the original information is represented by a subset or similar information, e.g. lossy compression
• CPC Class Codes: H03M 7/6064 Selection of Compressor

Abstract

A method for compressing data is disclosed. The method may include classifying the data from a plurality of data sources into classifications including at least PCM-encoded data and data structure information. The method may also include determining data states associated with the PCM-encoded data and the data structure information. The method may further include compressing the PCM-encoded data and the data structure information into compressed bit-streams based on the determined data states. The method may also include shuffling the compressed bit-streams to introduce randomness.

Claims

1. A method for compressing data, comprising:

classifying the data from a plurality of data sources into classifications including at least PCM-encoded data and data structure information;

determining data states associated with the PCM-encoded data and the data structure information; and

compressing the PCM-encoded data and the data structure information into compressed bit-streams based on the determined data states.

2. The method of claim 1, wherein determining the data states associated with the PCM-encoded data and the data structure information comprises:

grouping similar PCM-encoded data and data structure information together;

dividing the PCM-encoded data and the data structure information into non-overlapping segments; and

classifying a current segment of the PCM-encoded group and the data structure information into one of five data states;

transition state, constant, periodical, slow-varying, and noise.

3. The method of claim 2, wherein classifying a current segment into one of five states comprises:

determining first order derivatives of the current segment;

determining cross-correlation coefficients of the current segment;

determining short-term and long-term mean, maximum, minimum, and variance values; and

determining a zero-crossing rate of the current segment.

4. The method of claim 3, wherein classifying a current segment into one of five states further comprises:

classifying the current segment as constant if the number of first order derivatives is greater than a pre-defined threshold;

classifying the current segment as periodic if a first peak of the cross-correlation coefficients is greater than a pre-defined threshold;

classifying the current segment as transition state based on the short- and long-term mean, maximum, minimum, and variance values;

classifying the current segment as noise if the zero-crossing rate is greater than a pre-defined threshold; and

classifying the current segment as slow-varying if the zero-crossing rate is less than the pre-defined threshold used for classifying the segment as noise.

5. The method of claim 2, further including, encoding a current segment by an adaptive waveform encoder if the current segment is classified as either periodic, slow-varying, or transition state.

6. The method of claim 5, wherein encoding a current segment by the adaptive waveform encoder comprises extracting a theme skeleton from the current segment by:

setting a skeleton length to either a period exhibited by the current segment if the current segment is periodical or to a 1/N length of the current segment if the current segment is slow-varying or transition state;

setting a threshold based on an average of the maximum and minimum value of the current segment;

associating data points of the current segment as S_high if the data point is great than or equal to the threshold, and associating data points of the current segment as S_low if the data point is lesser than the threshold;

averaging S_high data points and averaging S_low data points;

determining switching positions in the current segment that correspond to the values of averaged S_high and averaged S_low; and

generating a skeleton based on S_low, S_high, and the switching positions.

7. The method of claim 5, further including, encoding the current segment by an adaptive waveform encoder, when the current segment is classified as periodical, by:

dividing the current segment into non-overlapping periods;

extracting the theme skeleton based on a first period of the current segment;

obtaining a skeleton for a period by fine-tuning a skeleton for a previous period;

determining a residual signal for a best skeleton obtained by fine-tuning; and

encoding the period based on the best skeleton and a residual signal corresponding to the difference between the best skeleton and an actual signal of the period.

8. The method of claim 5, further including, encoding the current segment by an adaptive waveform encoder, when the current segment is classified as slow-varying, by:

calculating 2^nd order, 4^th order, and 6^th order autoregressive predictions and corresponding residuals for the current segment; and

determining a skeleton model and corresponding residual for the current segment;

choosing the 2^nd order autoregressive prediction, 4^th order autoregressive prediction, 6^th order autoregressive prediction, or skeleton model that requires the least number of bits as representation of the current segment; and

encoding the current segment using corresponding parameters and residual of the chosen representation.

9. The method of claim 5, further including encoding the current segment by an adaptive waveform encoder, when the current segment is classified as transition, by:

dividing the current segment into 1/N length sub-segments;

extracting a skeleton and corresponding residual for each sub-segment;

determining a number of bits B_SKEL-RESID required for representing each sub-segment with a skeleton;

determining skeleton differences for each sub-segment based on a previous sub-segment;

determining a number of bits B_DIFF required for representing sub-segments based on only the difference between the sub-segment and the previous sub-segment; and

encoding the current segment based on skeleton differences if B_SKEL-RESID is greater than B_DIFF, and encoding the current segment based on the skeleton and corresponding residual for each sub-segment if B_SKEL-RESID is not greater than B_DIFF.

10. The method of claim 1, further comprising:

shuffling the compressed bit-streams to introduce randomness,

wherein shuffling the compressed bit-streams comprises alternating big-endian and little-endian representation every n bytes in the shuffled bit-stream; and

at least one of;

randomly alternating data in the shuffled bit-stream based on the classification of the data;

randomly allocating the order of the data in the shuffled bit-stream based on the data source;

or

randomly assigning bit allocation in individual encoding processes based on the state of segments of the data.

11. A data transmission system, comprising:

a local unit comprising a receiver,

a remote unit, the remote unit comprising transducers for generating data, a processor, and a transmitter;

the processor configured to compress the data by;

classifying the data into classifications including at least PCM-encoded data and data structure information,

determining data states associated with the PCM-encoded data and the data structure information, and

compressing the PCM-encoded data and the data structure information into compressed bit-streams based on the determined data states;

the transmitter transmitting the first and second compressed bit-streams to the local unit.