Method and apparatus of communication
The invention pertains to a method and apparatus of efficient encoding and decoding of vector quantized data. The method and system explores and implements sub-division of a quantization vector space comprising class-leader vectors and representation of the class-leader vectors by a set of class-leader root-vectors facilitating faster encoding and decoding, and reduced storage requirements.
1. A method of representing signal samples with quantized samples for communication over a communications channel characterized in that comprising:
quantizing, using a processor, the signal samples to signals corresponding to points of an E-lattice, wherein each quantized signal has a class-leader root-vector representative as determined including a sign distribution, a quantized signal being identified by the class-leader root vector representative and the sign distribution,
wherein the sign distribution is communicated over a communications channel;
wherein the quantization is minimizing a distortion criterion as determined from class-leader root-vectors of the E-lattice;
wherein the minimum is transformed as need be to form a minimum over the set of class leaders; and
wherein the distortion as determined from class-leader root-vectors of the E-lattice is conditionally transformed depending on a sign distribution of components of a vector representation of a signal sample to be quantized.
2. The method according to claim 1 , comprising a conditional sign-parity update.
3. The method according to claim 2 , wherein the signal sample representation includes inversion of a sign bit when sign parity of a signal sample and sign parity of a class-leader root-vector differ.
4. The method according to claim 1 , wherein each signal sample is represented by a vector, for which each component comprises a magnitude and a sign bit and distortion determined from the one or more class-leader root-vectors is transformed when the sign parity of the vector representation differs from the sign parity of the class-leader root-vector.
5. The method according to claim 4 , wherein the determined distortion is additively increased proportional to a product of the magnitude of the signal sample representation and a class-leader root-vector selected for providing minimum transformed distortion for a component for which the sign bits of an ordered signal sample and the class-leader root-vector differ.
6. The method according to claim 4 , wherein an index is determined for a vector comprising the magnitude components.
7. The method according to claim 1 , wherein the cardinality of a set of said class-leader root-vectors of the E-lattice is smaller than the cardinality of a set of class leaders for said points of the E-lattice.
8. The method according to claim 1 , wherein the distortion criterion comprises squared Euclidean distance.
9. The method according to claim 1 , wherein an index is determined representing a vector comprising the sign distribution.
10. The method according to claim 9 , wherein the index is determined for an adjusted vector, wherein a vector component has been adjusted corresponding to a sign inversion.
11. The method according to claim 1 , wherein the quantized samples are communicated over the communications channel by means of offset codewords determined by, for each codeword, offsetting an index achieved for a sequentially ordering of components of a data block of magnitudes according to size, and wherein the offset represents a particular equivalence class corresponding to the root class representative of the quantized signal.
12. The method according to claim 1 , wherein the E-lattice fulfills the requirement
E L v =D L v ∪( D L v +v ): v=[ ½ . . . 1 / 2 ],
where
D
L
v
:
y
∈
{
[
y
0
y
1
…
y
L
v
-
1
]
:
y
j
∈
ℤ
,
∑
j
=
0
L
v
-
1
y
j
≡
0
(
modulo
2
)
}
is the checkerboard lattice, or D-lattice, of the same number of dimensions, L v , as the E-lattice for integers .