Audio signal coding system and method using speech signal rotation prior to lattice vector quantization
View Patent ↗An apparatus configured to model an encoded signal to estimate at least one distribution of the signal, rotate the signal with respect to a lattice, for lattice quantization of the signal, dependent on the at least one distribution of the signal, and quantize the signal rotated with respect to the lattice.
1. An apparatus comprising:
at least one processor; and
at least one memory including computer program code, where the at least one memory and the computer program code are configured, with the at least one processor, to cause the apparatus to at least:
model a vector of coefficients of a speech signal to estimate at least one distribution of the vector of coefficients of the speech signal;
rotate the vector of coefficients of the speech signal with respect to a lattice, for lattice quantization of the vector of coefficients of the speech signal, dependent on the at least one distribution, wherein the rotation is selected to align a densest direction of the lattice with a denser direction of the vector of coefficients of the speech signal; said apparatus configured to determine the denser direction of the vector of coefficients of the speech signal by applying a mixture of Gaussian model to the vector of coefficients of the speech signal; and
quantize the vector of coefficients of the speech signal rotated with respect to the lattice.
2. The apparatus as claimed in claim 1 , wherein the at least one memory including the computer program code is configured with the at least one processor to cause the apparatus to entropy encode the quantized vector of coefficients of the speech signal.
3. The apparatus as claimed in claim 1 , wherein the at least one memory including the computer program code is configured with the at least one processor to cause the apparatus to model the vector of coefficients of the speech signal to estimate at least two distributions, wherein for each quantization the at least one memory including the computer program code is configured with the at least one processor to cause the apparatus to:
estimate a distortion parameter associated with the quantized vector of coefficients of the speech signal; and
select one of the at least two distributions with the smallest associated distortion parameter.
4. The apparatus as claimed in claim 1 , wherein the modeling comprises at least one of:
a mixture of Gaussian modeling;
a clustering analysis modeling; and
a correlation analysis.
5. The apparatus as claimed in claim 1 , wherein the at least one memory including the computer program code is configured with the at least one processor to cause the apparatus to rotate the vector of coefficients of the speech signal with respect to a lattice by applying a transform to the vector of coefficients of the speech signal to rotate the vector of coefficients of the speech signal by a first angle about a first axis.
6. The apparatus as claimed in claim 5 , wherein the at least one memory including the computer program code is configured with the at least one processor to cause the apparatus to apply at least one further transform to the rotated vector of coefficients of the speech signal to further rotate the rotated vector of coefficients of the speech signal by a further angle about a further axis.
7. A method comprising:
modeling, in a processor, a vector of coefficients of a speech signal to estimate at least one distribution of the vector of coefficients of the speech signal;
rotating, in the processor, the vector of coefficients of the speech signal with respect to a lattice, for lattice quantization of the vector of coefficients of the speech signal, dependent on the at least one distribution, wherein the rotation is selected to align a densest direction of the lattice with a denser direction of the vector of coefficients of the speech signal;
determining the denser direction of the vector of coefficients of the speech signal by applying a mixture of Gaussian model to the vector of coefficients of the speech signal; and
quantizing, in the processor, the vector of coefficients of the speech signal rotated with respect to the lattice.
8. The method as claimed in claim 7 , wherein modelling the vector of coefficients of the speech signal comprises modeling the vector of coefficients of the speech signal to estimate at least two distributions, and rotating and quantizing for each distribution, wherein the method further comprises:
estimating a distortion parameter associated with the quantized vector of coefficients of the speech signal; and
selecting one of the at least two distributions with the smallest associated distortion parameter.
9. The method as claimed in claim 7 , wherein modeling comprises at least one of:
a mixture of Gaussian modeling;
a clustering analysis modeling; and
a correlation analysis.
10. The method as claimed in claim 7 , wherein the rotating the vector of coefficients of the speech signal with respect to the lattice comprises applying a transform to the vector of coefficients of the speech signal to rotate the vector of coefficients of the speech signal by a first angle about a first axis.
11. The method as claimed in claim 10 , wherein the rotating the vector of coefficients of the speech signal with respect to the lattice further comprises applying at least one further transform to the rotated vector of coefficients of the speech signal to further rotate the rotated vector of coefficients of the speech signal by a further angle about a further axis.
12. A non-transitory computer readable medium embodying computer program code, the computer program code executed by a processor to perform operations comprising:
modeling, in the processor, a vector of coefficients of a speech signal to estimate at least one distribution of the vector of coefficients of the speech signal;
rotating, in the processor, the vector of coefficients of the speech signal with respect to a lattice, for lattice quantization of the vector of coefficients of the speech signal, dependent on the at least one distribution, wherein the rotation is selected to align a densest direction of the lattice with a denser direction of the vector of coefficients of the speech signal;
determining the denser direction of the vector of coefficients of the speech signal by applying a mixture of Gaussian model to the vector of coefficients of the speech signal; and
quantizing, in the processor, the vector of coefficients of the speech signal rotated with respect to the lattice.
13. The non-transitory computer readable medium as claimed in claim 12 , wherein modeling the vector of coefficients of the speech signal comprises modeling the vector of coefficients of the speech signal to estimate at least two distributions, and rotating and quantizing for each distribution, further comprising:
estimating a distortion parameter associated with the quantized vector of coefficients of the speech signal; and
selecting one of the at least two distributions with the smallest associated distortion parameter.
14. The non-transitory computer readable medium as claimed in claim 12 , wherein modeling comprises at least one of:
a mixture of Gaussian modeling;
a clustering analysis modeling; and
a correlation analysis.
15. The non-transitory computer readable medium as claimed in claim 12 , wherein the rotating the vector of coefficients of the speech signal with respect to the lattice comprises applying a transform to the vector of coefficients of the speech signal to rotate the vector of coefficients of the speech signal by a first angle about a first axis.
16. The non-transitory computer readable medium as claimed in claim 15 , wherein the rotating the vector of coefficients of the speech signal with respect to the lattice further comprises applying at least one further transform to the rotated vector of coefficients of the speech signal to further rotate the rotated vector of coefficients of the speech signal by a further angle about a further axis.