Audio codec using the Fast Fourier Transform, the partial overlap and a decomposition in two plans based on the energy.

• US 20070094015A1
• Filed: 09/18/2006
• Published: 04/26/2007
• Est. Priority Date: 09/22/2005
• Status: Abandoned Application

• Current Assignee: Georges Samake
• Original Assignee: Georges Samake
• Inventors: SAMAKE, GEORGES
• Application Number:
  US11/532,563
• Publication Number:
  US 20070094015A1
• Time in Patent Office: Days
• US Class Current: 704/212
• CPC Class Codes: G10L 19/0212 using orthogonal transformation
• 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 7/3053 Block-companding PCM systems

Abstract

Audio codec using the Fast Fourier Transform, the partial overlap and a decomposition in two plans based on the energy. The present invention concerns a method of audio compression and decompression, simple, of high quality, not requiring a lot of computations and allowing to obtain very high compression ratios. This codec is optimized for both the voice and the music. The most spread methods nowadays use the Linear Predictive Coding (LPC) in the time domain for the voice and the Modified Discrete Cosine Transform (MDCT) in the frequency domain for the music. The present codec uses the Fast Fourier Transform (FFT). The Fast Fourier Transform buffers are split into a forward plan (composed only of the biggest points) and a backward plan (composed of the most energetic bands). The non null points in the bands are composed only of points not taken into account in the forward plan. For the voice, this codec uses only the magnitudes of the local peaks (without the laterals points) and only the imaginary part in decompression. For the music and all audio signals, it uses the magnitudes and the phases of the points of the forward and backward plans, in compression and in decompression. It can also use only the local peaks with the phases. The edge effects are canceled with the help of a partial overlap method (50% or less) allowing a perfect reconstruction. Efficient methods of coding of magnitudes and phases are used. This codec is intended for all vocal bi-directional communications (voice over IP or mobile phones for instance), for the audio streaming (radios on Internet for instance) as well as for the stocking of audio data (files on hard disk for instance).

Claims

1. ) A method of audio compression and decompression comprising

the use of the Fast Fourier Transform (FFT) for the voice and the music and

a decomposition in two plans based on the energy.

2. ) The method of audio compression and decompression of claim 1, wherein said every frame is split into a forward plan composed of the N biggest points and a backward plan composed of the M most energetic bands.

3. ) The method of audio compression and decompression of claim 1 or 2, wherein said for the voice, one uses only the magnitudes of the local peaks (without the lateral points) and only the imaginary part in decompression.

4. ) The method of audio compression and decompression of claim 1 or 2, wherein said for the edge effects canceling with the music, one uses a method of partial overlap allowing a perfect reconstruction, with 50% or less than 50% of overlapping.

5. ) The method of audio compression and decompression of claim 1, 2 or 3 wherein said to reduce the size of data in the backward bands, one can apply the Adaptive Differential Pulse Code Modulation (ADPCM).

6. ) The method of audio compression and decompression of claim 1, 2 or 3, wherein said to reduce the size of data in the backward bands, one can apply the simple decimation which leads to a light quality loss, or the double decimation which leads to a bigger quality loss.

7. ) The method of audio compression and decompression of claim 1 or 2, wherein said for the backward plan, one can apply a coefficient of reduction between 1 and 2 to reduce the size of bands.

8. ) The method of audio compression and decompression of claim 1, wherein said the frame size (the FFT buffer size) depends on the sampling rate.

9. ) The method of audio compression and decompression of claim 1, 2 or 3, wherein said two methods of coding of magnitudes are presented:

base-10 logarithm and base-2 corrected scale which allows to obtain a great precision.

10. ) The method of audio compression and decompression of claim 1, 2 or 3, wherein said to diminish the number of bits in the coding of positions, one can use the relative coding.