Method and apparatus employing audio frequency offset extraction and floating-point conversion for digitally encoding and decoding high-fidelity audio signals

• US 4,922,537 A
• Filed: 06/02/1987
• Issued: 05/01/1990
• Est. Priority Date: 06/02/1987
• Status: Expired due to Fees

• Current Assignee: Frederiksen & SHU Laboratories Incorporated
• Original Assignee: Frederiksen & SHU Laboratories Incorporated
• Inventors: Frederiksen, Jeffery E.
• Primary Examiner(s): KEMENY, EMANUEL
• Application Number:
  US07/057,370
• Time in Patent Office: 1,064 Days
• Field of Search: 381/30, 381/31-36, 381/106, 375/122, 340/347 DD
• US Class Current: 704/212
• CPC Class Codes: H03M 7/24 Conversion to or from floating-point codes
• CPC Class Codes: H03M 7/50 Conversion to or from non-linear codes, e.g. companding

Abstract

An audio signal is initially represented by a series of high-resolution pulse code modulated (PCM) data. A lower rate series of representative values are extracted from the initial series of PCM data. Half of the lower rate is at an intermediate audio frequency so that the lower rate series encodes low frequency components of the audio signal. The PCM data are adjusted by offsetting in accordance with corresponding representative values and are then converted to a floating-point representation by extracting scale factor or exponents. The combination of the series of representative values and the floating-point data provides a rate-compressed representation of the audio signal which is capable of being decoded after transmission or storage to reproduce the audio signal without substantial noise, distortion or loss of dynamic range. The splitting of the audio information between the lower rate series and the adjusted floating-point PCM limits the normally destructive effect that low frequency components of high amplitude have upon high frequency components of relatively low amplitude. In a preferred embodiment, a common offset is determined for each block by computing the arithmetic mean of the maximum and minimum PCM data values for the block and truncating the result, the PCM data are adjusted by subtracting their corresponding common offsets, and a common exponent is determined for the block of adjusted PCM data. For encoding high-fidelity audio, preferably the audio signal is initially represented by a series of 16-bit PCM samples at a rate of at least 36 kilohertz, the block size is chosen to be 16 audio samples, and the encoded and compressed data for each block includes a 160 bit frame consisting of an 8-bit block offset, a 3-bit block exponent, a 5-bit error correction code, and sixteen floating-point values each including eight data bits and one parity bit. This format permits 9 stereo audio channels and frame synchronization to be readily transmitted over a conventional video channel.

Claims

1. A method for digitally encoding an audio signal represented by an initial series of pulse-code modulated (PCM) data values occurring at a first rate, said audio signal including low frequency components of high amplitude and high frequency components of relatively low amplitude, said method comprising the steps of:

(a) extracting from said PCM data values a series of representative values occurring at a second rate substantially lower than said first rate, half of said second rate being at an intermediate frequency in the audio spectrum,

(b) offsetting said PCM data values in accordance with corresponding values in said series of representative values to obtain a series of adjusted PCM data values, and

(c) converting the adjusted PCM data values to a series of floating-point data values by extracting exponents, so that the combination of said series of representative values and said series of floating-point data values encode said audio signal at a substantially lower data rate than said initial series of PCM data values, and thereby preventing the low frequency components of high amplitude from having a destructive effect upon the high frequency components of relatively low amplitude.

2. The method as claimed in claim 1, wherein said step (a) of extracting is performed by grouping a predetermined number of consecutive PCM data values into blocks, and computing a respective one of said representative values from the PCM data values in each block, so that said predetermined number is the ratio of said first and second rates.

3. The method as claimed in claim 2, wherein the representative value for each block is computed by selecting the maximum and minimum PCM data values in the block, and calculating said representative value as the mean value of said maximum and minimum values.

4. The method as claimed in claim 2, wherein said PCM data values are offset by subtracting from them corresponding values in said series of representative values.

5. The method as claimed in claim 1, wherein said half of said second rate is approximately one kilohertz.

7. The method as claimed in claim 2, wherein said first predetermined rate is approximately 36 kilohertz to accommodate an audio bandwidth of a 18 kilohertz, and said predetermined number of consecutive PCM data values is about 16.

8. The method as claimed in claim 1, further comprising the step of transmitting the floating-point data values along with said series of representative values over a video-bandwidth channel to a remote location where an audio signal is decoded from the transmitted series of values.

10. The method as claimed in claim 7, wherein said half of said second rate is approximately one kilohertz.

11. The method as claimed in claim 10, wherein said step (e) of combining includes the operation of adding said representative values to said fixed-point data values.

12. The method as claimed in claim 11, where said operation of adding is performed by adding each representative value to each of a predetermined number of consecutive fixed-point data values.

13. The method as claimed in claim 12, wherein said first rate is approximately 36 kilohertz to accommodate an audio bandwidth of about 18 kilohertz, and said predetermined number of consecutive adjusted PCM data values is 16.

17. The method as claimed in claim 12, wherein said conversion of the companded PCM data values from floating-point representative to fixed-point representation is performed by arithmetically right-shifting each of said predetermined number of consecutive ones of said floating-point data values by a selected number of binary places.

6. A decoder for decoding an audio signal having been represented by an initial series of pulse-code modulated (PCM) data values occurring at a first rate, said audio signal including low frequency components of high amplitude and high frequency components of relatively low amplitude, said audio signal having been encoded by:

(a) extracting from said PCM data a series of representative values occurring at a second rate substantially lower than said first rate, half of said second rate being at an intermediate frequency in the audio spectrum, said intermediate frequency lying between the frequencies of said low frequency components and said high frequency components;

(b) offsetting said PCM values in accordance with corresponding values in said series of representative values to obtain a series of adjusted PCM data values;

(c) extracting a series of scale factors from said series of adjusted PCM data values, said scale factors occurring at a rate substantially less than said first rate, said series of scale factors being selected in accordance with the magnitudes of said adjusted PCM data values; and

(d) scaling said adjusted PCM data values by corresponding ones of said scale factors, to obtain a scaled series of PCM data values, so that the combination of said series of representative values, said series of scale factors and said scaled series of PCM data values encode said audio signal at a substantially lower data rate than said initial series of PCM data values;

said decoder comprising;

(a) means for receiving said series of representative values, said series of scale factors, and said series of PCM data values;

(b) means for translating the scaled PCM data values in accordance with said scale factors to obtain a series of translated PCM data values; and

(h) means for combining corresponding ones of said representative values with said translated PCM data values to obtain a series of PCM data values approximating said initial data values, and thereby preventing the low frequency components of high amplitude from having a destructive effect upon the high frequency components of relatively low amplitude.

9. A method of decoding an audio signal which has been digitally encoded from an initial series of pulse-code modulated (PCM) data values occurring at a first rate, said audio signal including low frequency components of high amplitude and high frequency components of relatively low amplitude, said method including the steps of:

(a) extracting from said PCM data values a series of representative values occurring at a second rate substantially lower than said first rate, half of said second rate being at an intermediate frequency in the audio spectrum,

(b) offsetting said PCM data values in accordance with corresponding values in said series of representative values to obtain a series of adjusted PCM data values, and

(c) converting the adjusted PCM data values to a series of floating-point data values by extracting exponents, so that the combination of said series of representative values and said series of floating-point data values encode said audio signal at a substantially lower-data rate than said initial series of PCM data values,

said method of decoding comprising the steps of;

(d) translating the series of floating-point data values to obtain a series of fixed-point data values, and

(e) combining corresponding ones of said representative values with said fixed-point data values to obtain a series of PCM data values approximately said initial series, and thereby preventing the low frequency components of high amplitude from having a destructive effect upon the high frequency components of relatively low amplitude.

14. A method for digitally encoding an audio signal represented by an initial series of fixed-point pulse code modulate (PCM) data values occurring at a predetermined sampling rate, each value being represented by a predetermined number of bits, said audio signal including low frequency components of high amplitude and high frequency components of relatively low amplitude, said method comprising the steps of:

dividing said series of fixed-point PCM data values into blocks, each block comprising a plurality of consecutive PCM data values, each block including a predetermined number of said consecutive PCM data values so that said blocks occur at a determined block rate being less than said sampling rate, half of said block rate being at an intermediate frequency in the audio spectrum,

centering the fixed-point PCM data values within each block about a zero reference level by extracting a common offset value K for the data values within the block,

selecting for each block a respective one of a plurality of predetermined scale factors, the respective scale factor for each block being selected in accordance with the centered fixed-point PCM data value of maximum magnitude in the block,

scaling the centered fixed-point PCM data values in each block by the respective scale factor selected for the block to obtain a series of scaled PCM data values that are represented by a predetermined number of bits less than the number of bits of the PCM data values in the initial series, so that the combination of the series of floating-point PCM data values, the common offsets K for the blocks, and the common scale factors for the blocks encode said audio signal at a substantially lower bit rate than said initial series of fixed-point PCM data values, and thereby preventing the low frequency components of high amplitude from having a destructive effect upon the high frequency components of relatively low amplitude.

15. The method of claim 14, wherein said half of said block rate is approximately one kilohertz.

16. The method of claim 14, wherein the offset value K is determined as the median value between the minimum and maximum data values within each block.

18. A method for digitally encoding an audio signal represented by an initial series of fixed-point pulse code modulated (PCM) data values occurring at a predetermined sampling rate, each value being represented by a predetermined number of bits, said audio signal including low frequency components of high amplitude and high frequency components of relatively low amplitude, said method comprising the steps of:

dividing said series of fixed-point PCM data values into blocks, each block comprising a plurality of consecutive PCM data values, each block including a predetermined number of said consecutive PCM data values so that said blocks occur at a predetermined block rate being less than said sampling rate, half of said block rate being at an intermediate frequency in the audio spectrum,

centering the fixed-point PCM data values within each block about a zero reference level by extracting a common offset value K for the data values within the block,

transforming the centered fixed-point PCM data values within each block to a floating-point format in which the centered PCM data values are represented by a predetermined number of bits less than the number of bits of the PCM data values in said initial series, and in which a common exponent is determined for the block corresponding to a common scale factor for the floating-point conversion, so that the combination of the series of floating-point PCM data values, the common offsets K for the blocks, and the common exponents for the blocks encode said audio signal at a substantially lower bit rate than said initial series of fixed-point PCM data values, and thereby preventing the low frequency components of high amplitude from having a destructive effect upon the high frequency components of relatively low amplitude.

19. The method of claim 18 wherein each of the blocks consist of 16 PCM data values.

20. The method of claim 18 wherein said offset value K is determined as the median value between the minimum and maximum data values within each block.

21. The method of claim 18 wherein said predetermined rate is about 36 kilohertz to accommodate an audio bandwidth of about 18 kilohertz.

22. The method of claim 18 further comprising the step of transmitting, for each block, the common offset value K, the common exponent, and the floating-point PCM values, said transmission being performed over a video channel to a plurality of decoders.

23. The method as claimed in claim 22 wherein said video channel is one of a number of channels in a conventional cable television network transmitting a standard chrominance component, and said predetermined sampling rate is 37.879 KHz so as to be most compatible with said chrominance component.

24. The method as claimed in claim 18, further comprising the steps of truncating the common offsets K to 8 bits, truncating the common exponents to 3 bits, and truncating the floating-point PCM values to 8 bits.

25. The method as claimed in claim 24, further comprising the step of transmitting the truncated values to at least one decoder.

26. The method as claimed in claim 24 wherein said predetermined number of bits for representing each initial fixed-point PCM data value is 16.

27. The method as claimed in claim 18, wherein said half of said block rate is approximately one kilohertz.

28. A method for decoding an audio signal which has been encoded from an initial series of fixed-point pulse code modulated (PCM) data values occurring at a predetermined sampling rate and representing samples of said audio signal, each value being represented by predetermined number of bits, said audio signal including low frequency components of high amplitude and high frequency components of relatively low amplitude, said audio signal having been encoded by:

dividing said series of fixed-point PCM data values into blocks, each block comprising a plurality of consecutive PCM data values, each block including a predetermined number of said consecutive PCM data values so that said blocks occur at a predetermined block rate being less than said sampling rate, half of said block rate being at an intermediate frequency in the audio spectrum,

centering the fixed-point PCM data values within each block about a zero reference level by extracting a common offset value K for the data values within the block, and

transforming the centered fixed-point PCM data values within each block to a floating-point format in which the centered PCM data values are represented by a predetermined number of bits less than the number of bits of the PCM data values in said initial series, and in which a common exponent is determined for the block corresponding to a common scale factor for the floating-point conversion, so that the combination of the series of floating-point PCM data values, the common offsets K for the blocks, and the common exponents for the blocks encode said audio signal at a substantially lower bit rate than said initial series of fixed-point PCM data values,

said method of decoding comprising the steps of;

receiving said common offset values K, said common exponent values, and said floating-point PCM data values,

transforming the received floating-point PCM data values in accordance with their respective received exponent values to recover fixed-point PCM data values,

de-centering the fixed-point PCM data values by adding to them their respective common offset values, and

converting the de-centered fixed-point PCM data values to analog form, thereby preventing the low frequency components of high amplitude from having a destructive effect upon the high frequency components of relatively low amplitude.

29. The method as claimed in claim 28, wherein each of the blocks consists of 16 PCM data values.

30. The method as claimed in claim 28, wherein said step of receiving receives 8-bit common offset values K, 3-bit common exponent values, and 8-bit floating-point PCM data values.

31. The method as claimed in claim 30 wherein said predetermined sampling rate is approximately 36 KHz to provide an audio bandwidth of about 18 KHz, and wherein the de-centered fixed-point PCM data values converted to analog form are each represented by 16 bits to reproduce said audio signal with high fidelity.

32. The method as claimed in claim 28, wherein said half of said block rate is approximately one kilohertz.

33. A decoder for decoding an audio signal which has been encoded from an initial series of fixed-point pulse code modulated (PCM) data values occurring at a predetermined sampling rate and representing samples of said audio signal, each value being represented by a predetermined number of bits, said audio signal including low frequency components of high amplitude and high frequency components of relatively low amplitude, said audio signal having been encoded by:

dividing said series of fixed-point PCM data values into blocks, each block comprising a plurality of consecutive PCM data values, each block including a predetermined number of said consecutive PCM data values so that said blocks occur at a predetermined block rate being less than said sampling rate, half of said block rate being at an intermediate frequency in the audio spectrum,

centering the fixed-point PCM data values within each block about a zero reference level by extracting a common offset value K for the data values within the block, and

transforming the centered fixed-point PCM data values within each block to a floating-point format in which the centered PCM data values are represented by a predetermined number of bits less than the number of bits of the centered PCM data values, and in which a common exponent is determined for the block corresponding to a common scale factor for the floating-point conversion, so that the combination of the series of floating-point PCM data values, the common offsets K for the blocks, and the common exponents for the blocks encode said analog audio signal at a substantially lower bit rate than said initial series of fixed-point PCM data values, and thereby limiting the normally destructive effect that the low frequency components of high amplitude have upon the high frequency components of low amplitude,

said decoder comprising, in combination,

means for receiving said common offset values K, said common exponent values, and the floating-point PCM data values,

means for transforming the received floating-point PCM data values in accordance with their respective received exponent values to recover fixed-point PCM data values,

means for de-centering the fixed-point PCM data values by adding their respective common offset values, and

means for converting the de-centered fixed-point PCM data values to analog form.

34. The decoder as claimed in claim 33, wherein said means for receiving includes means for receiving encoded data blocks consisting of 16 PCM data values, and a common exponent and offset value K for each block.

35. The decoder as claimed in claim 33, wherein said means for receiving includes means for receiving 8-bit common offset values K, a 3-bit common exponent values, and 8-bit floating-point PCM data values.

36. The decoder as claimed in claim 33, wherein said means for transforming includes a shift register for performing an arithmetic right-shift by a selected number of binary places indicated by the exponent values.

37. The decoder as claimed in claim 33, wherein said predetermined sampling rate is approximately 36 KHz to provide an audio bandwidth of about 18 KHz, the de-centered fixed-point PCM data values converted to analog form are each represented by 16 bits, and said means for converting the de-centered fixed-point PCM data values comprises a 16-bit digital-to-analog converter to reproduce said audio signal with high fidelity.

38. The decoder as claimed in claim 33, wherein said means for receiving further comprises a cable television tuner and demodulator.

39. The decoder as claimed in claim 33, wherein said decoder consists essentially of hard-wired digital logic.

40. The decoder as claimed in claim 33, wherein said half of said block rate is approximately one kilohertz.

41. A method of digitally encoding, transmitting, and decoding an audio signal represented by an initial series of pulse-code modulated (PCM) data values occurring at a first rate, said audio signal including low frequency components of high amplitude and high frequency components of relatively low amplitude, said method comprising the steps of:

(a) extracting from said PCM data a series of representative values occurring at a second rate substantially lower than said first rate, half of said second rate being at an intermediate frequency in the audio spectrum, said intermediate frequency lying between the frequencies of said low frequency components and said high frequency components,

(b) offsetting said PCM values in accordance with corresponding values in said series of representative values to obtain a series of adjusted PCM data values,

(c) extracting a series of scale factors from said series of adjusted PCM data values, said scale factors occurring at a rate substantially less than said first rate, said series of scale factors being selected in accordance with the magnitudes of said adjusted PCM data values,

(d) scaling said adjusted PCM data values by corresponding ones of said scale factors, to obtain a scaled series of PCM data values, so that the combination of said series of representative values, said series of scale factors and said scaled series of PCM data values encode said audio signal at a substantially lower data rate than said initial series of PCM data values,

(e) transmitting said series of scaled PCM data values, said series of representative values and said series of scale factors over a band-limited channel,

(f) receiving said series of scaled PCM values, said series of representative values and said series of scale factors from said band-limited channel,

(g) translating the scaled PCM data values in accordance with said scale factors to obtain a series of translated PCM data values, and

(h) combining corresponding ones of said representative values with said translated PCM data values to obtain a series of PCM data values approximately said initial series, and thereby preventing the low frequency components of high amplitude from having a destructive effect upon the high frequency components of relatively low Amplitude.

42. The method of claim 41, wherein said half of said second rate is approximately one kilohertz.

43. The method of claim 41, wherein said scale factors are exponents and said adjusted fixed-point PCM data values are scaled exponentially by said exponents.

44. The method of claim 41, wherein said step (a) of extracting the representative values is performed by grouping a predetermined number of consecutive PCM data values into blocks, and computing a respective one of said representative values from the PCM data values in each block, so that said predetermined number is the ratio of said first and second rates.

46. The method as claimed in claim 44, wherein said first predetermined rate is approximately 36 kilohertz to accommodate an audio bandwidth of about 18 kilohertz, and said predetermined number of consecutive PCM data values is 16.

45. The method of claim 49, wherein one scale factor is extracted from each block.