# US 9,519,619 B2 · Data processing method and device for processing speech signal or audio signal

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 9,519,619 B2
- Document ID: US9519619B2
- Status: Granted
- Kind: B2
- Application number: 13/938,834
- Filed: 2013-07-10
- Published: 2013-11-14
- Granted: 2016-12-13
- Earliest claimed: 2011-01-10
- Assignees: Huawei Technologies Co., Ltd.
- Applicants: Huawei Technologies Co., Ltd.
- Inventors: Deming Zhang; Haiting Li; Anisse Taleb; Jianfeng Xu
- Web page: https://patentgrep.com/patent/US9519619B2

## CPC

- G06F 17/142: {Fast Fourier transforms, e.g. using a Cooley-Tukey type algorithm}
- G06F 17/141: {Discrete Fourier transforms}
- G06F 17/147: {Discrete orthonormal transforms, e.g. discrete cosine transform, discrete sine transform, and variations therefrom, e.g. modified discrete cosine transform, integer transforms approximating the discrete cosine transform (G06F17/145 takes precedence)}
- G10L 19/0212: {using orthogonal transformation}
- G10L 19/022: Blocking, i.e. grouping of samples in time; Choice of analysis windows; Overlap factoring
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G10L: SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING

## Abstract

A data processing method is disclosed, including: twiddling input data, so as to obtain twiddled data; pre-rotating the twiddled data by using a symmetric rotate factor, where the rotate factor is a·W4L2p+1, p=0, . . . , L/2−1, and a is a constant; performing a Fast Fourier (Fast Fourier Transform, FFT) transform of L/2 point on the pre-rotated data, where L is the length of the input data; post-rotating the data that has undergone the FFT transform by using a symmetric rotate factor, where the rotate factor is b·W4L2q+1, q=0, . . . , L/2−1, and b is a constant; and obtaining output data.

## Claims

### Claim 1

A method for processing a raw audio signal to transform the raw audio signal from time domain to frequency or from frequency domain to time domain and performed by an audio signal processing device including a processor, comprising:
obtaining an input signal corresponding to the raw audio signal;
obtaining a twiddled signal by twiddling the input signal;
pre-rotating the twiddled signal by using a first symmetric rotate factor to generate a pre-rotated signal, wherein the first symmetric rotate factor is (a·W4L2p+1), wherein p=0, . . . , L/2−1, and (a) is a constant;
performing a Fast Fourier Transform (FFT) of a L/2 point on the pre-rotated signal to generate a FFT transformed signal, wherein L is the length of the input signal;
post-rotating the FFT transformed signal using a second symmetric rotate factor to generate a post-rotated signal, wherein the second symmetric rotate factor is (b·W4L2q+1), wherein q=0, . . . , L/2−1, and wherein (b) is a constant; and
writing a processed audio signal based on the post-rotated signal into a bitstream for transmitting or storing.

### Claim 2 (depends on 1)

The method according to claim 1, before the post-rotating step, the method further comprises performing fixed rotate compensation by using a fixed rotate compensation factor.

### Claim 3 (depends on 2)

The method according to claim 2, wherein performing the fixed rotate compensation by using the fixed rotate compensation factor comprises:
performing fixed rotate compensation at least one time, wherein a product of a rotate compensation factor of the at least one time fixed rotate compensation is W8L−3.

### Claim 4 (depends on 2)

The method according to claim 2, wherein performing the fixed rotate compensation by using the fixed rotate compensation factor comprises:
performing fixed rotate compensation at least one time, wherein a rotate compensation factor of the at least one time fixed rotate compensation is a first order Taylor series expansion of at least one factor a product of which is W8L−3.

### Claim 5 (depends on 1)

The method according to claim 1, wherein a product of (a) and (b) is equal to
2
2
L
2
.

### Claim 6

A time-domain to frequency-domain signal processing method for processing a raw audio signal performed by an audio signal processing device comprising a processor, comprising:
obtaining a time domain signal corresponding to the raw audio signal;
obtaining a pre-processed signal by pre-processing the time domain signal;
pre-rotating the pre-processed signal by using a rotate factor (a·WNn+0.5) to generate a pre-rotated signal;
performing a Fast Fourier Transform (FFT) of N/4 point on the pre-rotated signal to generate a FFT transformed signal;
post-rotating the FFT transformed signal using a rotate factor (b·WNk+0.5), so as to obtain a frequency domain signal;
writing a processed audio signal based on the frequency domain signal into a bitstream for transmitting or storing;
wherein, before the obtaining the frequency domain signal, the method further comprises performing fixed rotate compensation by using a fixed rotate compensation factor; (a) and (b) are constants, the N is the length of the time domain signal, and
W
N
=
ⅇ
-
j
⁢
⁢
2
⁢
π
N
.

### Claim 7 (depends on 6)

The method according to claim 6, wherein performing the fixed rotate compensation by using the fixed rotate compensation factor comprises:
performing fixed rotate compensation at least one time, wherein a product of a rotate compensation factor of the at least one time fixed rotate compensation is WN−0.75.

### Claim 8 (depends on 6)

The method according to claim 6, wherein performing the fixed rotate compensation by using the fixed rotate compensation factor comprises:
performing fixed rotate compensation at least one time, wherein a rotate compensation factor of the at least one time fixed rotate compensation is a first order Taylor series expansion of at least one factor a product of which is WN−0.75.

### Claim 9 (depends on 1)

The method according to claim 1, wherein a product of (a) and (b) is equal to
2
N
.

### Claim 10

A frequency-domain to time-domain signal processing method for processing a raw audio signal performed by an audio signal processing device including a processor, comprising:
obtaining a frequency domain signal corresponding to the raw audio signal;
obtaining a twiddled signal by twiddling the frequency domain signal;
pre-rotating the twiddled signal by using a rotate factor (c·WNk+0.5) to generate a pre-rotated signal;
performing a Fast Fourier Transform (FFT) of N/4 point on the pre-rotated signal to generate a FFT transformed signal;
post-rotating the FFT transformed signal using a rotate factor (d·WNn+0.5) to generate a post-rotated signal;
post-processing the post-rotated signal, so as to obtain a time domain signal; and
obtaining a processed audio audio signal based on the post-rotated signal;
wherein, before obtaining the time domain signal, the method further comprises performing fixed rotate compensation by using a fixed rotate compensation factor; (c) and (d) are constants, the N is twice the length of the frequency domain signal, and
W
N
=
ⅇ
-
j
⁢
⁢
2
⁢
π
N
.

### Claim 11 (depends on 10)

The method according to claim 10, wherein performing the fixed rotate compensation by using the fixed rotate compensation factor comprises:
performing fixed rotate compensation at least one time, wherein a product of a rotate compensation factor of the at least one time fixed rotate compensation is WN−0.75.

### Claim 12 (depends on 10)

The method according to claim 10, wherein performing the fixed rotate compensation by using the fixed rotate compensation factor comprises:
performing fixed rotate compensation at least one time, wherein a rotate compensation factor of the at least one time fixed rotate compensation is a first order Taylor series expansion of at least one factor a product of which is WN−0.75.

### Claim 13

A signal processing device for processing a raw audio signal to transform the raw audio signal from time domain to frequency or from frequency domain to time domain and including a processor, comprising:
a unit configured to obtain an input signal corresponding to the raw audio signal;
a twiddle unit, configured to twiddle the input signal, so as to obtain a twiddled signal;
a pre-rotate unit, configured to pre-rotate the twiddled signal by using a first symmetric rotate factor to generate a pre-rotated signal, wherein the first symmetric rotate factor is (a·W4L2p+1), wherein p=0, . . . , L/2−1, and (a) is a constant;
a transform unit, configured to perform a Fast Fourier Transform (FFT) of L/2 point on the pre-rotated signal to generate a FFT transformed signal, wherein L is the length of the input signal;
a post-rotate unit, configured to post-rotate the FFT transformed signal using a second symmetric rotate factor to generate a post-rotated signal, wherein the second symmetric rotate factor is (b·W42q+1), wherein q=0, . . . , L/2−1, and wherein (b) is a constant; and
a unit configured to write a processed audio signal based on the post-rotated signal into a bitstream for transmitting or storing.

### Claim 14 (depends on 13)

The device according to claim 13, further comprising:
a fixed rotate compensation unit, configured to perform fixed rotate compensation by using a fixed rotate compensation factor.

### Claim 15 (depends on 14)

The device according to claim 14, wherein the fixed rotate compensation unit is configured to perform fixed rotate compensation at least one time, and a product of a rotate compensation factor of the at least one time fixed rotate compensation is W8L−3.

### Claim 16 (depends on 14)

The device according to claim 14, wherein the fixed rotate compensation unit is configured to perform fixed rotate compensation at least one time, and a rotate compensation factor of the at least one time fixed rotate compensation is a first order Taylor series expansion of at least one factor a product of which is W8L−3.

### Claim 17 (depends on 14)

The device according to claim 14, wherein a product of (a) and (b) is equal to
2
2
L
2
.

### Claim 18

A time-domain to frequency-domain signal processing device for processing a raw audio signal and including a processor, comprising:
a unit configured to obtain a time domain signal corresponding to the raw audio signal;
a pre-processing unit, configured to pre-process the time domain signal, so as to obtain a pre-processed signal;
a pre-rotate unit, configured to pre-rotate the pre-processed signal by using a rotate factor (a·Nn+0.5) to generate a pre-rotated signal;
a transform unit, configured to perform a Fast Fourier Transform (FFT) of N/4 point on the pre-rotated signal to generate a FFT transformed signal;
a post-rotate unit, configured to post-rotate the FFT transformed signal using a rotate factor (b·WNk+0.5), so as to obtain a frequency domain signal;
a unit configured to write a processed audio signal based on the frequency domain signal into a bitstream for transmitting or storing;
wherein, the device further comprises:
a fixed compensation unit, configured to perform fixed rotate compensation by using a fixed rotate compensation factor; the (a) and (b) are constants, the N is the length of the time domain signal, and
W
N
=
ⅇ
-
j
⁢
⁢
2
⁢
π
N
.

### Claim 19 (depends on 18)

The device according to claim 18, wherein the fixed rotate compensation unit is configured to perform fixed rotate compensation at least one time, and a product of a rotate compensation factor of the at least one time fixed rotate compensation is WN−0.75.

### Claim 20 (depends on 18)

The device according to claim 18, wherein the fixed rotate compensation unit is configured to perform fixed rotate compensation at least one time, and a rotate compensation factor of the at least one time fixed rotate compensation is a first order Taylor series expansion of at least one factor a product of which is WN−0.75.

### Claim 21

A frequency-domain to time-domain signal processing device for processing a raw audio signal and including a processor, comprising:
a unit configured to obtain a frequency domain signal corresponding to the raw audio signal;
a twiddle unit, configured to twiddle the frequency domain signal, so as to obtain a twiddled signal;
a pre-rotate unit, configured to pre-rotate the twiddled signal by using a rotate factor (c·WNk+0.5) to generate a pre-rotated signal;
a transform unit, configured to perform a Fast Fourier Transform (FFT) of N/4 point on the pre-rotated signal to generate a FFT transformed signal;
a post-rotate unit, configured to post-rotate the FFT transformed signal using a rotate factor (d·WNn+0.5);
a unit configured to obtain a processed audio signal based on the post-rotated signal;
wherein, the device further comprises:
a fixed compensation unit, configured to perform fixed rotate compensation by using a fixed rotate compensation factor; the (c) and (d) are constants, the N is twice the length of the frequency domain signal, and
W
N
=
ⅇ
-
j
⁢
2
⁢
⁢
π
N
.

### Claim 22 (depends on 21)

The device according to claim 21, wherein the fixed rotate compensation unit is configured to perform fixed rotate compensation at least one time, and a product of a rotate compensation factor of the at least one time fixed rotate compensation is WN−0.75.

### Claim 23 (depends on 21)

The device according to claim 21, wherein the fixed rotate compensation unit is configured to perform fixed rotate compensation at least one time, and a rotate compensation factor of the at least one time fixed rotate compensation is a first order Taylor series expansion of at least one factor a product of which is WN−0.75.

### Claim 24

A time-domain to frequency-domain audio signal processing method, with the audio signal processing being performed by an audio signal processing device including a processor, comprising:
obtaining a time domain signal corresponding to a raw audio signal;
applying at least one twiddle factor to the time domain signal to obtain a twiddled signal;
pre-rotating, by using a symmetric rotation factor (a·W4L2p+1), the twiddled signal to obtain a pre-rotated signal, wherein p=0, . . . , L/2−1, and (a) is a constant;
performing a Fast Fourier Transform (FFT) of L/2 points on the pre-rotated signal to obtain a FFT transformed signal;
performing a fixed rotate compensation to the FFT transformed signal by multiplying the FFT transformed signal with a fixed rotate compensation factor;
post-rotating the FFT transformed signal using a symmetric rotate factor (b·W4L2q+1) to obtain a post-rotated signal, wherein q=0, . . . , L/2−1, and (b) is a constant;
writing a processed audio signal based on the post-rotated signal into a bitstream for transmitting or storing.

### Claim 25

A time-domain to frequency-domain audio signal processing apparatus, comprising:
a processor; and
memory coupled to the processor and including non-transitory computer-readable instructions that, when executed by the processor:
obtain a time domain signal corresponding to a raw audio signal;
apply twiddle factors to the time domain signal to obtain a twiddled signal;
pre-rotate, by using a symmetric rotation factor (a·W4L2p+1), the twiddled signal to obtain a pre-rotated signal, wherein p=0, . . . , L/2−1, and (a) is a constant;
perform a Fast Fourier Transform (FFT) of L/2 points on the pre-rotated signal to obtain a FFT transformed signal;
perform a fixed rotate compensation to the FFT transformed signal by multiplying the FFT transformed signal with a fixed rotate compensation factor;
post-rotate the FFT transformed signal using a symmetric rotate factor (b·W4L2q+1) to obtain a post-rotated signal, wherein q=0, . . . , L/2−1, and (b) is a constant;
write a processed audio signal based on the post-rotated signal into a bitstream for transmitting or storing.

### Claim 26 (depends on 1)

The method according to claim 1, wherein W4L2p+1 in the first symmetric rotate factor is expressed in the following form:
W
4
⁢
⁢
L
2
⁢
⁢
p
+
1
=
cos
⁢
2
⁢
π
⁡
(
2
⁢
⁢
p
+
1
)
4
⁢
⁢
L
-
j
⁢
⁢
sin
⁢
2
⁢
π
⁡
(
2
⁢
⁢
p
+
1
)
4
⁢
⁢
L
.

### Claim 27 (depends on 1)

The method according to claim 1, wherein W4L2q+1 in the second symmetric rotate factor is expressed in the following form:
W
4
⁢
⁢
L
2
⁢
⁢
p
+
1
=
cos
⁢
2
⁢
π
⁡
(
2
⁢
⁢
p
+
1
)
4
⁢
⁢
L
-
j
⁢
⁢
sin
⁢
2
⁢
π
⁡
(
2
⁢
⁢
p
+
1
)
4
⁢
⁢
L
.

### Claim 28 (depends on 13)

The method according to claim 13, wherein W4L2p+1 in the first symmetric rotate factor is expressed in the following form:
W
4
⁢
⁢
L
2
⁢
⁢
p
+
1
=
cos
⁢
2
⁢
π
⁡
(
2
⁢
⁢
p
+
1
)
4
⁢
⁢
L
-
j
⁢
⁢
sin
⁢
2
⁢
π
⁡
(
2
⁢
⁢
p
+
1
)
4
⁢
⁢
L
.

### Claim 29 (depends on 13)

The method according to claim 13, wherein W4L2q+1 in the second symmetric rotate factor is expressed in the following form:
W
4
⁢
⁢
L
2
⁢
⁢
p
+
1
=
cos
⁢
2
⁢
π
⁡
(
2
⁢
⁢
p
+
1
)
4
⁢
⁢
L
-
j
⁢
⁢
sin
⁢
2
⁢
π
⁡
(
2
⁢
⁢
p
+
1
)
4
⁢
⁢
L
.

### Claim 30 (depends on 24)

The method according to claim 24, wherein W4L2p+1 in the first symmetric rotate factor is expressed in the following form:
W
4
⁢
⁢
L
2
⁢
⁢
p
+
1
=
cos
⁢
2
⁢
π
⁡
(
2
⁢
⁢
p
+
1
)
4
⁢
⁢
L
-
j
⁢
⁢
sin
⁢
2
⁢
π
⁡
(
2
⁢
⁢
p
+
1
)
4
⁢
⁢
L
.

### Claim 31 (depends on 24)

The method according to claim 24, wherein W4L2q+1 the second symmetric rotate factor is expressed in the following form:
W
4
⁢
⁢
L
2
⁢
⁢
p
+
1
=
cos
⁢
2
⁢
π
⁡
(
2
⁢
⁢
p
+
1
)
4
⁢
⁢
L
-
j
⁢
⁢
sin
⁢
2
⁢
π
⁡
(
2
⁢
⁢
p
+
1
)
4
⁢
⁢
L
.

### Claim 32 (depends on 25)

The method according to claim 25, wherein W4L2p+1 in the first symmetric rotate factor is expressed in the following form:
W
4
⁢
⁢
L
2
⁢
⁢
p
+
1
=
cos
⁢
2
⁢
π
⁡
(
2
⁢
⁢
p
+
1
)
4
⁢
⁢
L
-
j
⁢
⁢
sin
⁢
2
⁢
π
⁡
(
2
⁢
⁢
p
+
1
)
4
⁢
⁢
L
.

### Claim 33 (depends on 25)

The method according to claim 25, wherein W4L2q+1 in the second symmetric rotate factor is expressed in the following form:
W
4
⁢
⁢
L
2
⁢
⁢
p
+
1
=
cos
⁢
2
⁢
π
⁡
(
2
⁢
⁢
p
+
1
)
4
⁢
⁢
L
-
j
⁢
⁢
sin
⁢
2
⁢
π
⁡
(
2
⁢
⁢
p
+
1
)
4
⁢
⁢
L
.