Patent Application
20080025528
These and other features, aspects, and advantages of the present invention will become better understood with regard to the following description, appended claims, and accompanying drawings where:
Reference will now be made in detail to the present preferred embodiments of the invention, examples of which are illustrated in the accompanying drawings. Wherever possible, the same reference numbers are used in the drawings and the description to refer to the same or like parts.
This invention offers a noise reduction system and an audio processing unit used in the BTSC system to reduce noise during an encoding process or a decoding process in the approach of the digital processing. The filter of the noise reduction system and the audio processing unit uses a new transfer function with fewer parameters (coefficients of the transfer function) to reduce required memory capacity. Using this device, parameters are more economically stored in the memory.
When the filter of the noise reduction system is used for an encoding process, the transfer function is:
S(f,b)=[1+(f/20.1[kHz])(b+51)/(b+1)]/[1+(f/20.1[kHz])(1+51 b)/(b+1)] (2a)
When the filter of the noise reduction system is used for a decoding process, the transfer function is:
S
−1(f,b)=[1+(f/20.1[kHz])(1+51 b)/(b+1)]/[1+(f/20.1[kHz])(b+51)/(b+1)] (2b)
In order to apply the transfer functions (2a) and (2b) in a digital audio processor, the S(f,b) and S−1(f,b) have to be bilinear transformed into Z domain. Therefore set b=10(d/20), i.e. d=20 log (b), and the transfer functions (2a) and (2b) respectively become:
S(Z,b)=[2πf(Z+1)(b+1)+2fs(Z−1)(b+51)]/[2πf(Z+1)(b+1)+2fs(Z−1)(1+51 b)] (3a)
S
−1(Z,b)=[2πf(Z+1)(1+b)+2fs(Z−1)(1+51b)]/[2πf(Z+1)(1+b)+2fs(Z−1)(b+51)] (3b)
where f=20.1 kHz, fs is the sampling frequency.
In the transfer function (3b), S−1(Z,b) is equal to S (Z,b−1). Thus, the transfer functions (3a) and (3b) are set to be:
When d>0,S(Z,b)=H(Z)=(b0+b1Z−1)/(a0+a1Z−1) (4a)
When d<0,S(Z,b−1)=H−1(Z)=(a0+a1Z−1)/(b0+b1Z−1) (4b)
In the transfer function (1) of the prior art, the memory needs to store 6 parameters. In order to reduce the amount of the parameters, the transfer functions (4a) and (4b) are transformed to be:
when the variable d is greater than 0, i.e. d>0:
when the variable d is less than 0, i.e. d<0:
when the variable d is greater than 0, i.e. d>0:
when the variable d is less than 0, i.e. d<0:
The memory 240 is arranged to store the parameters, when the variable d is greater than 0, i.e. d>0, the memory outputs the parameters b0/a0, b1/b0 and a1/a0 to the filter 230; when the variable d is less than 0, i.e. d<0, the memory outputs the parameters a0/b0, b1/b0 and a1/a0 to the filter 230.
From this transfer function, the memory just needs to store 4 parameters b0/a0, a0/b0, b1/b0 and a1/a0. Furthermore, the parameter a0/b0 can be generated from the parameter b0/a0 by using hardware (such as a circuit). Therefore, compared with the conventional memory that stores 6 parameters in the noise reduction system, this memory just needs to store 3˜4 parameters. This memory needs only ½˜⅔ capacity of conventional memory.
The variable d is an address of the memory, and the variable d is equal to 20 μg (the time-weighted root mean square of the encoded audio signal). In order to get equal filter frequency response of the parameter d>0 and d<0, the range of the variable d is about ±35[decibel ERMS] to about ±45[decibel ERMS].
When the noise reduction system 210 is used in the encoding process, a wideband compression unit 250a coupled to the memory 240 and the filter 230 in the noise reduction system compresses the encoded audio signal into a wideband compression signal. The noise reduction system 210 further has a multiplier 260 coupled to the wideband compression unit 250a and the filter 230. The multiplier 260 generates the input signal by multiplying the audio signal with the wideband compression signal.
In real products, the memory 240 in the noise reduction system 210 is conventionally implemented with a ROM table, such as a look up ROM table.
The filter 230 of the audio spectral expansion unit 220b is arranged to filter the encoded signal and generate an output signal according to an inverse of the transfer function, the variable d, and the parameters b0/a0, a0/b0, b1/b0 and a1/a0 as described in
When the noise reduction system 210 is used in the decoding process, a wideband expansion unit 250b coupled to the memory 240 in the noise reduction system expands the encoded audio signal to be a wideband expansion signal. The noise reduction system 210 further has a multiplier 260. The multiplier 260 coupled to the wideband expansion unit 250b and the filter 230 is arranged to multiply the output signal with the wideband expansion signal to be the audio signal.
The memory 340 coupled to the multiplexer is arranged to receive the parameter addresses and output several parameters b0/a0, a0/b0, b1/b0 and a1/a0. When the variable d is greater than 0, i.e. d>0, the memory outputs the parameters b0/a0, b1/b0 and a1/a0 to the filter 330. When the variable d is less than 0, i.e. d<0, the memory outputs the parameters a0/b0, b1/b0 and a1/a0 to the filter 330.
The filter 330 is coupled to the memory. When the audio processing unit 330 is used in the encoding process, the filter 330 is arranged to filter an input signal according to a transfer function, the variable d, and the parameters b0/a0, a0/b0, b1/b0 and a1/a0. The transfer functions are equations (5a) and (5b) as described above.
In the digital audio processing unit 310a, the multiplexer 320 can be configured in the memory, and the variable d is an address of the memory. Furthermore, the variable d is 20 log (the time-weighted root mean square of the encoded audio signal). In order to get equal filter frequency response for the parameter d>0 and the parameter d<0, the range of the variable d is about 35[decibel ERMS] to about ±45[decibel ERMS].
The digital audio processing unit 310a further has a gain device 370, a spectral bandpass filter 380, and an energy level detecting device 390. The gain device 370 is coupled to the filter 330 to receive and increase the gain of the encoded audio signal. The spectral bandpass filter 380 is coupled to the gain device 370 to generate a spectral signal according to the encoded audio signal with increasing gain. The energy level detecting device 390 is coupled to the spectral bandpass filter 380 and the multiplexer 320 to generate the variable d according to the spectral signal.
When the digital audio processing unit 310a is used in the encoding process, a wideband compression unit 350a coupled to the gain device and the filter 330 compresses the encoded audio signal into a wideband compression signal. The digital audio processing unit 310a further has a multiplier 360 coupled to the wideband compression unit 350a and the filter 330. The multiplier 360 generates the input signal by multiplying the audio signal with the wideband compression signal.
In real products, the memory 340 in the digital audio processing unit 310a is conventionally implemented by a ROM table, such as a look up ROM table.
When the digital audio processing unit 310b is used in the decoding process, the filter 330 is arranged to filter the encoded signal and generate an output signal according to an inverse of the transfer function, the variable d, and the parameters b0/a0, a0/b0, b1/b0 and a1/a0.
When the digital audio processing unit 310b is used in the decoding process, the digital audio processing unit 310b further has a wideband expansion unit 350b coupled to the gain device 370 to expand the encoded audio signal into a wideband expansion signal. The digital audio processing unit 310b further has a multiplier 360. The multiplier 360 coupled to the wideband expansion unit 350b and the filter 330 generates the audio signal by multiplying the output signal with the wideband expansion signal.
Using the noise reduction system or the audio processing unit described above, this memory needs only ½˜⅔ capacity of a conventional memory. The audio processing data of the multimedia in real life is very huge, the noise reduction system and the audio processing unit can reduce the necessary memory capacity.
It will be apparent to those skilled in the art that various modifications and variations can be made to the structure of the present invention without departing from the scope or spirit of the invention. In view of the foregoing, it is intended that the present invention cover modifications and variations of this invention provided they fall within the scope of the following claims and their equivalents.
