Digital system and method for compressing speech signals for storage and transmission

• US 4,802,221 A
• Filed: 07/21/1986
• Issued: 01/31/1989
• Est. Priority Date: 07/21/1986
• Status: Expired due to Term

• Current Assignee: Magnachip Semiconductor Limited (Magnachip Semiconductor)
• Original Assignee: NCR Corporation
• Inventors: Jibbe, Mahmoud K.
• Primary Examiner(s): Kemeny, Emanuel S.
• Application Number:
  US06/887,492
• Time in Patent Office: 925 Days
• Field of Search: 381/29-53
• US Class Current: 704/208
• CPC Class Codes: G10L 19/00 Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis in musical instruments G10H
• CPC Class Codes: G10L 25/00 Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 muting semiconductor-based amplifiers when some special characteristics of a signal are sensed by a speech detector, e.g. sensing when no signal is present, H03G3/34
• CPC Class Codes: G10L 25/78 Detection of presence or absence of voice signals switching of direction of transmission by voice frequency in two-way loud-speaking telephone systems H04M9/10
• CPC Class Codes: G10L 25/90 Pitch determination of speech signals
• CPC Class Codes: G10L 25/93 Discriminating between voiced and unvoiced parts of speech signals G10L25/90 takes precedence
• 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

The system and method of the invention operates upon a serial bit stream of speech encoded data by reading the data in fixed length frames. Each frame is classified as to whether the frame represents voiced, unvoiced, or a silence condition. For a voiced frame, the pitches are detected. The bits of the voiced frame are then compared with the bits of prior received voiced frames, on a pitch basis, so as to detect repetition. When repetition is detected, the repetition count, the pitch period, and the one set of the repetitive pitch bits are stored ready for transmission. The compressed data is reconstructed by utilizing a decoder which interprets the stored information and causes reinsertion of the repetitions.

Claims

1. A method for compressing a serial bit stream of speech encoded data for storage and transmission comprising the steps of:

(a) storing the bits of said serial bit stream;

(b) reading the stored bits in fixed length frames;

(c) processing each frame of read bits to classify the frame as voiced, unvoiced or silence;

(d) determining the pitch for a frame classified as voiced;

(e) comparing, on a pitch basis, the bits of the present voiced frame with the bits of the preceding voiced frames to detect repetition;

(f) temporarily storing the repetition count, the pitch period, and the repetitive pitch bits; and

(g) repeating steps (b) through (f) to compress the serial bit stream that is temporarily stored.

2. The method according to claim 1 and further comprising the step of:

reconstructing the serial bit stream from the temporarily stored repetition count pitch period, repetitive pitch bits and the bits of the unvoiced and silence frames.

3. The method according to claim 1 and further comprising the steps of:

(a) reading the temporarily stored repetition count, the pitch period, and the repetitive pitch bits; and

(b) repeating the repetitive pitch bits the number of times corresponding to the repetition count to reconstruct said serial bit stream

4. The method according to claim 1 wherein the pitch is determined by the steps of:

(a) calculating the width of one pulses;

(b) calculating the average width of one pulses in a fixed length frame;

(c) calculating the location of one pulses;

(d) calculating the number of one pulses;

(e) selecting the large one pulses whose width is greater than the average width of one pulses;

(f) determining the distance between the large one pulses,

(g) sorting, in ascending order, the large one pulses in pairs according to the distance between the large one pulses;

(h) forming classes of pitch periods by placing in each class all large one pulses that are within a fixed threshold of each other; and

(i) selecting the distance of the closest spaced pair of large one pulses from the class having the greatest number of large one pulses as the determined pitch.

5. A method for compressing a serial bit stream of speech encoded data for storage and transmission comprising the steps of:

(a) storing the bits of said serial bit stream;

(b) reading the stored bits in fixed length frames;

(c) processing each frame of read bits to classify the frame as voiced, unvoiced or silence;

(d) compressing the bits of a silent frame by storing one of the repeating bit patterns and a number corresponding to the number of times the bit pattern repeats in a frame;

(e) compressing the bits of a voiced frame;

(f) comparing, on a pitch basis, the bits of the present voiced frame with the bits of the preceding voiced frames to detect repetition;

(g) temporarily storing the repetition count, the pitch period, and the repetitive pitch bits; and

(h) repeating steps (b) through (g) to compress the serial bit stream that is temporarily stored.

6. A system for compressing a serial bit stream of speech encoded data comprising:

storage means for temporarily storing the serial bit stream of speech encoded data;

receiver means operatively coupled to said storage means for reading the serial bit stream from said storage means in fixed length frames;

compression means coupled to said receiver means for classifying each frame as voiced, unvoiced or silence and for compressing a voiced frame by recording the number of times that a pitch occurs in each preceding frame and the pattern of the repetitive bits;

second storage means for temporarily storing the compressed frames from said compression means; and

decoding means coupled to said second storage means for decompressing the compressed frames to reconstruct said serial bit stream of speech encoded data.

7. The system according to claim 6 wherein said compression means is comprised of:

detector means for classifying each frame of said serial bit stream as voice, unvoiced or silence;

pitch detector means for determining the pitch of a voiced frame by comparing the pitch of a present frame against the pitch of preceding voiced frames and providing a compressed data frame of bits indicative of the number of times the data repeats the pitch period of the data and the data itself and the pitch of the unvoiced and silence frames; and

a pitch interval repetition means coupled to said detector means and said pitch detector means for organizing the storage of compressed data in serial frames and for outputting said frames in serial order to said second storage means.

8. The system according to claim 7 wherein said detector means is comprised of:

means for detecting the number of bit sequences 0000, 0101 or 1010 in a fixed length frame and for classifying the frame as silent if the number of sequences exceeds a predetermined threshold; and

means for calculating the number of one pulses and the number of one bits in the frames not classed as silent such that if the number of one bits divided by the number of one pulses is less than a second predetermined threshold the frame is classed as unvoiced and if greater the frame is classed as voiced.

9. The system according to claim 6 wherein said decoding means is comprised of:

repetition count means coupled to said second storage means for reading the number of times that a pitch occurs in a frame and the number of bits in a frame;

data field reader means coupled to said second storage means and to said repetition count means to provide the sequence of data bits sorted in said second storage means as a function of the number of times that a pitch occurs from the repetition count means; and

repetition reinsertion means responsive to said data field reader means and said repetition count means for reconstructing the encoded serial bit stream of said speech encoded data.

10. The system according to claim 6 and further comprising:

output drive means coupled to said decoding means for driving said serial bit stream encoded data signal; and

a bit serial decoder for receiving the signal from said output drive means and for providing a reconstructed analog output signal corresponding to the speech encoded data.