Method for improving the voice quality in low-rate dynamic bit allocation sub-band coding

• US 5,842,160 A
• Filed: 07/18/1997
• Issued: 11/24/1998
• Est. Priority Date: 01/15/1992
• Status: Expired due to Term

• Current Assignee: Ericsson, Inc. (Telefonaktiebolaget LM Ericsson)
• Original Assignee: Ericsson, Inc. (Telefonaktiebolaget LM Ericsson)
• Inventors: Zinser, Richard L
• Primary Examiner(s): Knepper, David D.
• Application Number:
  US08/897,022
• Time in Patent Office: 494 Days
• Field of Search: 704/205-209, 704/226-227, 704/500-504, 704/265, 704/229
• US Class Current: 704/229
• CPC Class Codes: G10L 19/012 Comfort noise or silence coding
• CPC Class Codes: H04B 1/665 using psychoacoustic properties of the ear, e.g. masking effect

Abstract

Voice quality is improved in low-rate dynamic bit allocation sub-band coding (DBASC) by inserting synthetic filler signals for non-transmitted signal bands. Large spectral holes may exist in an output voice speech spectrum of DBASC with a low transmission rate. The holes are caused by low energy signal sub-bands that are not transmitted and create noticeable artifacts in the received voice signal. To avoid these artifacts, the spectral holes in the received signal are filled with synthetic signals generated from the received signal. The synthetic filler signals are signals from transmitted energy bands that are scaled to the energy level of the non-transmitted band.

Claims

1. A method for improving a multi-band signal in a receiver having a sub-frequency band combining filter bank, said method comprising the steps of:

a. decoding energy levels for each transmitted band of a received signal,

b. inputting a received sub-band signal sample into the filter bank for each transmitted band;

c. generating a filler sub-band signal sample for each non-transmitted band of the received signal using at least one but not all received sub-band signal samples from the transmitted bands, where the at least one of the received sub-band signals has a frequency different than does the non-transmitted band, and inputting the filler sub-band signal into the filter bank, and

d. combining the received sub- band signal samples and the filler sub-band signal sample in the filter bank to form a multi-band signal.

2. A method as in claim 1 wherein the filler sub-band signal of step (c) is a normalized received sub-band signal sample multiplied by a quantizer scaling factor for the non-transmitted band.

3. A method as in claim 2 wherein the generated filler sub-band signal sample is further multiplied by a value of approximately 0.5.

4. A method for improving a multi-band signal in a receiver having a sub-band combining filter bank, said method comprising the steps of:

a. decoding energy levels for each transmitted band of a received signal,

b. inputting a received sub-band signal sample into the filter bank for each transmitted band; and

c. generating a filler sub-band signal sample for each non-transmitted band of the received signal using a received sub-band signal sample from a transmitted band, wherein the transmitted band is offset by four band levels from the corresponding non-transmitted band.

5. A method for improving a low-rate digital signal in a dynamic bit allocation sub-band receiver having quadrature mirror filters, said method comprising the steps of:

a. separating a received bit stream into bits representing sub-band energy levels and corresponding sub-band signal samples;

b. allocating bits to each band using the bits from the received bit stream representing sub-band energy levels;

c. inputting sub-band signal samples into the quadrature mirror filters for each band for which bits are allocated; and

d. generating a filler signal for a band for which no bits are allocated using a selected sub-band signal sample from a band to which bits are allocated, where the selected sub-band signal is of a frequency sub-band different than a sub-band of the filler signal, and inputting the filler signal to the quadrature mirror filters.

6. A method as in claim 5 wherein in step (d) the filler signal is a signal substantially proportional to the selected sub-band signal sample.

7. A method as in claim 6 wherein the band for the selected signal sample is offset from the band with no allocated bits by a predetermined number of bands.

8. A method for improving a low-rate digital signal in a dynamic bit allocation sub-band receiver having quadrature mirror filters, said method comprising the steps of:

a. separating a received bit stream into bits representing sub-band energy levels and corresponding sub-band signal samples;

b. allocating bits to each band of the digital signal using the bits from the received bit stream representing sub-band energy levels;

c. inputting sub-band received signal samples into the quadrature mirror filters for each band for which bits are allocated; and

d. generating a sub-band filler signal for a band for which no bits are allocated using a selected sub-band signal sample from a band to which bits are allocated, and inputting the filler signal to the quadrature mirror filters, wherein the filler signal is a signal substantially proportional to the selected sub-band signal sample where the selected sub-band signal sample is multiplied by a quantizer scaling factor for the band having no allocated bits and by the reciprocal of the quantizer scaling factor for the band for the selected signal sample.

9. A method for improving a low-rate digital signal in a dynamic bit allocation sub-band receiver having quadrature mirror filters, said method comprising the steps of:

a. separating a received bit stream into bits representing sub-band energy levels and corresponding sub-band signal samples;

b. allocating bits to each band of the digital signal using the bits from the received bit stream representing sub-band energy levels;

c. inputting sub-band received signal samples into the quadrature mirror filters for each band for which bits are allocated; and

d. generating a filler sub-band signal for a band for which no bits are allocated using a selected sub-band signal sample from a band to which bits are allocated, and inputting the filler signal to the quadrature mirror filters, wherein the filler signal is a signal substantially proportional to the selected sub-band signal sample and wherein the energy band for the selected signal sample is offset from the band with no allocated bits by four bands.