Speech sampling and companding device

• US 4,550,425 A
• Filed: 09/20/1982
• Issued: 10/29/1985
• Est. Priority Date: 09/20/1982
• Status: Expired due to Term

• Current Assignee: Sperry Corporation
• Original Assignee: Sperry Corporation
• Inventors: Hedin, Raymond C., Andersen, David P., Siebenand, John F.
• Primary Examiner(s): Kemeny, E. S. Matt
• Application Number:
  US06/420,545
• Time in Patent Office: 1,135 Days
• Field of Search: 381/29-35, 381/106, 340/347 DD, 358/261
• US Class Current: 704/212
• 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: H03M 7/3053 Block-companding PCM systems
• CPC Class Codes: H04B 14/048 Non linear compression or expansion

Abstract

An analog speech signal is sampled of a nominal rate of 6 kilohertz and digitized in a Mu-Law Encoder. The digital output of the Mu-Law Encoder is converted by a microprocessor performing table look-up to linearized pulse code modulation (PCM) samples nominally of eight bits per sample. Using a BSPCM (Block Scaled Pulse Code Modulation) method, in each block of nominally 246 eight-bit PCM samples (representing approximately 41 milliseconds), the maximum and minimum sample values are found and used to calculate a scale factor equal to the maximum sample value minus the minimum sample value, with the difference being then divided by a constant number nominally equaling 16. Then the BSPCM samples are generated from the PCM samples each as a corresponding one PCM sample minus the minimum PCM sample value, the difference being then divided by the scale factor. In effect, the bit rate is reduced by adjusting the step size to follow the local block dynamic range. The BSPCM samples so created are susceptible to signal processing operations like as to PCM samples. When the BSPCM encoded words plus the minimum, PCM encoded, sample plus the range increment scale factor are stored as a data block, then such data block can, at a later time, be decoded, or reconstituted, into linear PCM data. A silence interval is encoded as zero amplitude using run length coding of the number of blocks. Such digital PCM data can be converted to an analog audio signal for voice output across a telephone system.

Claims

1. A method for the digital encoding of an analog electrical signal representing speech comprising:

digitalizing said analog electrical signal at a sampling rate in order to produce upon an interval of time a multiplicity, or block, of linear pulse code modulated samples;

computing a scale factor for said multiplicity of pulse code modulated samples as a first quantity divided by a second quantity, wherein siad first quantity equals the maximum valued one of said multiplicity of samples minus the minimum valued one of said multiplicity of samples while said second quantity equals a constant number; and

generating, as said digital encoding of said analog electrical signal, a multiplicity of successive ones of digital code words wherein each successive one digital code word is formulated as a third quantity divided by said scale factor, wherein said third quantity equals a corresponding one of said multiplicity of pulse code modulated samples minus said minimum valued one of said multiplicity of samples;

whereby said method of digital encoding is called Block-Scaled Pulse Code Modulation because a multiplicity, or Block, of Pulse Code Modulated samples are Scaled by a factor, a scale factor, which scale factor is computed in consideration of the dynamic range of said multiplicity, or Block, of Pulse Code Modulated Samples.

2. The digital encoding method according to claim 1 wherein said digitalizing step further comprises:

analog-to-digital converting said analog signal at said sampling rate in a (Greek letter) "mu" law companding CODer-DECoder (CODEC) in order to produce upon an interval of time a multiplicity, or block, of digital samples;

plus

digital-to-digital converting said multiplicity of digital samples into a corresponding multiplicity, or block, of linearized digital samples, wherein said linearized digital samples by being correspondent to said digital samples which by action of said "mu" law companding CODer-DECoder are proportional to said analog electrical signal are thusly linear pulse code modulated samples.

3. The digital encoding method according to claim 2 wherein said digital-to-digital converting step further comprises:

converting through a table lookup said digital output of said (Greek letter) "mu" law companding CODer-DECoder (CODEC) ino said multiplicity of pulse code modulated samples.

4. The digital encoding method according to claim 1 wherein said digitalizing step is further defined as:

digitalizing said analog electrical signal at a sampling rate of approximately 6 kilohertz in order to produce, during an interval of approximately 41 milliseconds, a multiplicity of approximately 246 eight-bit linear pulse code modulated samples;

and wherein said computing step is further defined as;

computing a scale factor for said multiplicity of pulse code modulated samples as the rounded first quantity equaling the maximum valued one of said multiplicity of samples minus the minimum valued one of said multiplicity of samples divided by a constant number second quantity equaling, or approximately equaling, 16;

whereby if said constant number second quantity equals 16, then said rounded scale factor developed in consideration of eight-bit pulse code modulated samples is thusly four binary bits in length.

5. The digital encoding method according to claim 1 which further comprises after said generating step:

formulating, as a block of digital data words communicatable to other digital devices, a record consisting of said multiplicity of successive ones of said digital code words, plus said computed scale factor, plus said minimum valued one of said multiplicity of linear pulse code modulated samples;

whereby said record so formulated contains all information which is necessary to reconstruct said multiplicity of linear pulse code modulated samples from said multiplicity of said Block-Scaled Pulse Code Modulated digital code words, plus from said computed scale factor, plus from said minimum valued one of said multiplicity of linear pulse code modulated samples.

6. The digital encoding method according to claim 5 wherein said formulating step further comprises:

formulating said record consisting of said multiplicity of said successive ones of said digital code words, plus said computed scale factor, plus said minimum valued one of said multiplicity of linear pulse code modulated samples save that, when all one said multiplicity of said linear pulse code modulated samples are below an energy threshold thusly indicating silence upon said analog electrical signal representing speech, then said record consists of the number of said multiplicities of linear pulse code modulated samples which are below said energy threshold, plus said computed scale factor, plus said minimum valued one of said multiplicity of linear pulse code modulated samples

whereby since said record consisting, in part, of said number of said multiplicities of linear pulse code modulated samples which are below said energy threshold is not formulated save when all one said multiplicity of said linear pulse code modulated samples are below said energy threshold, then said number, for the alternative record-formulating condition of silence upon said analog electrical signal representing speech, must equal at least one;

whereby, if all said linear pulse code modulated samples are below an energy threshold then necessarily said minimum valued one of said multiplicity of said linear pulse code modulated samples is essentialy equal to zero;

whereby since said multiplicity of said linear pulse code modulated samples were produced by said digitalizing of said analog electrical signal at said sampling rate, said record type consisting, in part, of said number of said multiplicities of pulse code modulated samples which are below said energy threshold does indicate thereby said number the length of an interval of silence upon said analog electrical signal representing speech.

7. A method for decoding Block-Scaled Pulse Code Modulated (BSPCM) digitalized information consisting of a multiplicity of BSPCM digital code words, plus a minimum valued one of that like multiplicity of linear pulse code modulated (PCM) sample quantities from which said BSPCM digital code words were originally encoded, plus a range increment sale factor, in order to thereby said decoding reconstitute a multiplicity of linear PCM digital sample words, said decoding method comprising:

receiving, all as information input into a first buffer, a multiplicity of BSPCM digital code words, plus a minimum valued one of that like multiplicity of linear pulse code modulated (PCM) sample quantities from which said BSPCM digital code words were originally encoded, plus a range increment scale factor which is that quantity equaling the maximum valued one minus the minimum valued one of said multiplicity of linear PCM sample quantities from which said BSPCM digital code words were originally encoded entirely divided by a constant number;

computing from and upon the contents of said buffer a multiplicity of linear PCM digital sample words wherein each successive one is formulated as ((the corresponding successive one of said multiplicity of BSPCM digital code words) times (said range increment scale factor)) plus (said minimum valued one of said multiplicity of linear PCM sample quantities); and

storing said computed multiplicity of linear PCM digital sample words in a second buffer;

whereby said receiving, said computing, and said storing, digitalized information of a BSPCM nature which, most notably, is both more compact than PCM information but which BSPCM information is not directly susceptible to digitalized signal processing has been decoded into PCM digitalized information which is, most notably, susceptible to digitalized signal processing operations.

8. The BSPCM digitalized information decoding method of claim 7 extended to incorporate digitalized signal processing, said decoding and signal processing method further comprising:

modifying each said PCM digital sample words in place within said second buffer by shiftng, specifically by either right-shifting each said digital sample words either two places or one place, or by either left-shifting each said digital sample words either one place or two places while limiting, to a maximum value threshold, the value which is by said left-shifting obtained;

whereby said modifying by right-shifting two places, by right-shifting one place, by left-shifting one place, or by left-shifting two places a change in the PCM encoded loudness, from softest to soft to loud to loudest, of said PCM digital samples words is obtained;

whereby said shifting is a method of digitalized signal processing for the control of the loudness represented by linear PCM digital encoded information.

9. The BSPCM digitalized information decoding method of claim 7 extended to incorporate digitalized signal processing, said decoding and signal processing method further comprising:

modifying a succession of PCM digital samples words within said multiplicity of PCM digital samples words within said second buffer by the respective successive repetitive addition of a plurality of digital values proportional to that amplitude, both positive and negative, which is obtained upon each of a like plurality of parts of one cycle of a sine wave;

whereby said modifying accomplishes the mixing of a digitalized sine wave signal, or tone, with said PCM digital sample words.

10. The BSPCM digitalized information decoding method of claim 7 further comprising:

converting said computed multiplicity of linear PCM digital sample words stored in said second buffer into an audio analog speech output signal.

11. The BSPCM digitalized information decoding method of claim 10 when, additionally to said multiplicity of BSPCM digital code words plus said minimum valued one of said PCM sample quantities plus said range increment scale factor, further data concerning the pause, or silence interval, duration between successive intervals of BSPCM encoded speech is also part of said information, wherein said receiving step further comprises:

receiving said data concerning said pause, or silence interval, duration between successive intervals of BSPCM encoded speech into said first buffer;

and wherein said converting step further comprises;

converting said data concerning said pause, or silence interval, duration between successive intervals of BSPCM encoded speech into a null, or silent, audio analog output signal equal in duration to the time value of said data.

12. The BSPCM digitalized information decoding method of claim 11 extended to incorporate signal processing, said decodiing and signal processing method further comprising before said converting step:

modifying said data concerning said pause, or both said data concerning said pause plus said computed multiplicity of linear PCM digital sample words, by either increasing the magnitude, or the duration, of all said data concerning said pause, or by both increasing the magnitude of all said data concerning said pause plus causing each said computed multiplicity of linear PCM digital sample words, which constitute a block as stored within said second buffer, to be twice successively rendered to said converting step;

whereby said increase in said magnitude, or time duration, of all said data concerning said pause does slow said audio analog speech output signal;

whereby said increase in said magnitude, or time duration, of all said data concerning said pause plus said twice successive rendering, and a resultant twice successive said converting, of each said computed multiplicity of linear PCM sample words does maximally slow said audio analog speech output signal;

whereby the digitalized signal processing obtained by said modifying is thusly the slowing of speech rate.

13. The BSPCM digitalized information decoding method of claim 11 extended to incorporate signal processing, said decoding and signal processing method further comprising before said converting step:

modifying said data concerning said pause, or both said data concerning said pause plus said computed multiplicity of linear PCM digital sample words by either decreasing the magnitude, or time duration, of all said data concerning said pause, or by both decreasing the magnitude of all said data concerning said pause plus causing each said computed multiplicity of linear PCM digital sample words, which constitute a block as stored within said second buffer, to be truncated, with some sample words discarded, upon being rendered to said converting step;

whereby said decrease in said magnitude, or time duration, of all said data concerning said pause does speed up said audio analog speech output signal;

whereby said decrease in said magnitude, or time duration, of all said data concerning said pause plus said truncation of each said computed multiplicity of linear PCM sample words does maximally speed up said audio analog speech output signal;

whereby the digitalized signal processing obtained by said modifying is thusly the speeding up of speech rate.