Apparatus and method for low complexity combinatorial coding of signals
View Patent ↗During operation of an encoder, a signal vector (x) is received. A first multi-precision operand (Ψ′ k ) will be generated based on the signal vector to be encoded. A mantissa operand and an exponent operand are generated. Both the mantissa operand and the exponent operand are representative of a second multi-precision operand that is based on the signal vector to be encoded. A portion of Ψ′ k is selected to be modified based on the exponent operand. A part of Ψ′ k is modified based on the mantissa operand to produce a modified multi-precision operand (Ψ′ k+1 ). Finally, a multi-precision codeword is generated for use in a corresponding decoder.
1. A method for operating an encoder that produces a codeword (C) from an input vector (x), the method comprising the steps of:
receiving, by a vector generator of the encoder, an input signal;
creating, by the vector generator, the input vector;
receiving, by a combinational coding circuitry of the encoder, the input vector to be encoded, wherein the input vector comprises speech, audio, image, or video signals;
generating, by the combinational coding circuitry of the encoder, a first multi-precision operand (Ψ′ k ) based on the input vector;
generating, by the combinational coding circuitry, a mantissa operand (M) and an exponent operand (h), wherein the mantissa operand and the exponent operand are representative of a second multi-precision operand (F′(p k ,k)) that is based on the signal vector to be encoded;
selecting, by the combinational coding circuitry, a portion of the first multi-precision operand to be modified based on the exponent operand;
modifying, by the combinational coding circuitry, the portion of the first multi-precision operand based on the mantissa operand to produce a modified multi-precision operand (Ψ′ k+1 ); and
generating, by the combinational coding circuitry, the codeword (C) based on the modified multi-precision operand; wherein the codeword is for use by the encoder to encode the speech, audio, image, or video signals and by a corresponding decoder to decode the speech, audio, image, or video signals.
2. The method of claim 1 wherein Ψ′ k is based on a set of mantissa and exponent operands that are representative of F′(p i ,i); 1≦i<k, where p i are the positions of the non-zero elements of the input vector x.
3. The method of claim 1 wherein M≈└2 k f ·2 min(k i ,l g) ┘.
4. The method of claim 1 wherein h is a function k i .
5. The method of claim 1 wherein the portion of the first multi-precision operand Ψ′ k to be modified is represented by a location indicator defined by integers w l and w u .
6. The method of claim 5 wherein
w
l
=
⌊
h
16
⌋
,
w
u
=
⌊
h
+
l
R
16
⌋
+
1
,
where 16 is a word length and l R is a length of the mantissa.
7. The method of claim 1 wherein Ψ′ k+1 =Ψ′ k +F′(p k ,k).
8. The method of claim 1 wherein C π =Ψ′ v+1 .
9. An encoder comprising:
a vector generator creating an input vector x from a received input signal;
combinatorial coding circuitry to generate a codeword (C) wherein the combinational coding circuitry performing the steps of:
receiving the input vector x to be encoded, wherein the vector x comprises speech, audio, image, or video signals;
generating a first multi-precision operand (Ψ′ k ) based on x;
generating a mantissa operand (M) and an exponent operand (h), wherein the mantissa operand and the exponent operand are representative of a second multi-precision operand (F′(p k ,k)) that is based on the signal vector to be encoded;
selecting a portion of Ψ′ k to be modified based on the exponent operand;
modifying the portion of Ψ′ k based on the mantissa operand and location indicator to produce a modified multi-precision operand (Ψ′ k+1 ); and
generating a codeword (C) based on Ψ′ k+1 ; wherein the codeword is for use by the encoder to encode speech, audio, image, or video signals and by a corresponding decoder to decode the speech, audio, image, or video signals.
10. The encoder of claim 9 wherein
Ψ
k
′
=
∑
i
=
1
k
-
1
F
′
(
p
i
,
i
)
,
where p i are the positions of the non-zero elements of vector x.
11. The encoder of claim 9 wherein M≈└2 k f ·2 min(k i ,l R) ┘.
12. The encoder of claim 9 wherein h=min(0,k i −l R ).
13. The encoder of claim 9 wherein the portion of Ψ′ k to be modified is represented by a location indicator defined by integers w l and w u that identifies the words in the multi-precision codeword which are to be modified.
14. The encoder of claim 13 wherein
w
l
=
⌊
h
16
⌋
,
w
u
=
⌊
h
+
l
R
16
⌋
+
1
,
where 16 is a word length and l R =16 is chosen as the length of the mantissa.
15. The encoder of claim 9 where Ψ′ k+1 =Ψ′ k +F′(p k ,k).
16. The encoder of claim 9 wherein C π =Ψ′ v+1 .
17. A method for operating a decoder that generates a vector (x) from a codeword (C), the method comprising the steps of:
receiving, by combinatorial decoding circuitry of the decoder, the codeword (C π );
generating, by the combinatorial decoding circuitry, a first multi-precision operand (Ψ′ k+1 ) based on C π;
generating, by the combinatorial decoding circuitry, a mantissa operand (M) and an exponent operand (h), wherein the mantissa operand and the exponent operand are representative of a second multi-precision operand (F′(p k ,k));
selecting, by the combinatorial decoding circuitry, a portion of Ψ′ k+1 to be modified based on the exponent operand;
modifying, by the combinatorial decoding circuitry, the portion of Ψ′ k+1 based on the mantissa operand and location indicator to produce a modified multi-precision operand (Ψ′ k );
decoding, by the combinatorial decoding circuitry, a position p k−1 of the (k−1)th non-zero elements of vector x;
generating the vector (x) based on the position p k−1 ;
passing the vector x to signal reconstruction circuitry of the decoder, and
creating, by the signal reconstruction circuitry, an output signal comprising speech, audio, image, or video based on the generated vector (x).
18. The method of claim 17 wherein
Ψ
k
′
=
∑
i
=
1
k
-
1
F
′
(
p
i
,
i
)
,
where p i are the positions of the non-zero elements of vector x.
19. The method of claim 17 wherein M≈└2 k f ·2 min(k i ,l R) ∅.
20. A decoder comprising:
combinatorial coding circuitry performing the steps of:
receiving the codeword (C π );
generating a first multi-precision operand (Ψ′ k+1 ) based on C π;
generating a mantissa operand (M) and an exponent operand (h), wherein the mantissa operand and the exponent operand are representative of a second multi-precision operand (F′(p k ,k)) that is based on the signal vector to be encoded;
selecting a portion of Ψ′ k+1 to be modified based on the exponent operand;
modifying the portion of Ψ′ +1 based on the mantissa operand and location indicator to produce a modified multi-precision operand (Ψ′ k );
decoding a position p k−1 of the (k− 1 )th non-zero elements of vector x; and
generating the vector (x) based on the position p k−1 and
signal reconstruction circuitry to create an output signal of speech, audio, image, or video signals using the vector (x).