1. The present invention relates broadly to a method of processing a digital signal such as an audio signal. The invention also relates broadly to a computer or device-readable medium for processing a digital signal or filter, and a computer system for processing a digital signal or filter. The invention extends to other digital processing including processing images and other signals including signals associated with digital communications.
2. In digital recording and playback an analog signal representative of audio is converted into a digital signal which lends itself to manipulation and storage. The conversion is performed in an analog to digital converter (ADC). The stored digital signal can be converted back to an analog signal in a digital to analog converter (DAC). The analog signal is played back using conventional audio equipment such as amplifiers and speakers. The digital signal can be manipulated prior to the DAC to improve its quality before playback. This manipulation includes audio EQ where selected parts of the frequency spectrum of the audio are filtered to, for example, compensate for irregularities in the frequency response. The audio may also be filtered to resolve problems from its conversion into a digital signal or back to an analog signal. This manipulation in the digital domain also includes audio emulation, compression, time-stretching and pitch-shifting.
3. According to a first aspect of the present invention there is provided a method of processing a digital signal, said method comprising the steps of:
providing a digital filter including a plurality of neighbouring sample points;
performing a sample rate increase on the digital filter to provide a plurality of intermediate sample points between adjacent of the neighbouring sample points, said intermediate points being populated dependent on a weighted influence determined in the time domain of a predetermined number of the neighbouring sample points;
applying the digital filter to the signal where:
4. Preferably the method comprises an initial step of selecting an initial digital filter having a sample resolution substantially the same as the signal, and expanding the selected initial filter to effect offset of the offset sample points of the signal relative to the respective neighbouring sample points of the filter.
5. Preferably the step of applying the digital filter to the signal involves applying said filter at an adjusted sampling rate proportional to the offset of the sample points of the signal relative to the respective neighbouring sample points of the filter.
6. Preferably the filter is at least in part represented by an impulse response produced by an impulse fed to said filter and wherein the neighbouring sample points of the impulse response are offset in the time domain relative to the respective offset sample points of the signal. More preferably the impulse response is in the time domain represented at least in part by a mathematical function.
7. According to a second aspect of the invention there is provided a method of processing a digital signal, said method comprising the steps of:
providing another digital signal including a plurality of neighbouring sample points;
performing a sample rate increase on the other digital signal to provide a plurality of intermediate sample points between adjacent of the neighbouring sample points, said intermediate points being populated dependent on a weighted influence determined in the time domain of a predetermined number of the neighbouring sample points;
applying the other signal to the signal where:
8. Preferably the method comprises an initial step of expanding the other signal in its time domain to effect offset of the offset sample points relative to the respective neighbouring sample points of the signal.
9. Preferably the method comprises the step of representing the other signal in the time domain by at least its cosine and sine components. More preferably the cosine components are replaced at least in part with square-shaped wave components for valve-type emulation on application to the signal. Alternatively the sine components are replaced at least in part with triangular-shaped wave components for transistor-type emulation on application to the signal.
10. Preferably the other signal includes random noise. More preferably the random noise is represented by discrete values of a mathematical function to be applied to the signal. Even more preferably one or more of the discrete values of the random noise is replaced with either a zero (0) or one 1).
11. Preferably the step of applying the other signal to the signal involves applying said other signal at an adjusted sampling rate proportional to the offset of the offset sample points.
12. Preferably the method also comprises a step of deriving or constructing a modulation envelope to be applied as the other signal. More preferably the modulation envelope is derived from an initial envelope representing value changes in the other signal. Even more preferably the initial envelope is manipulated by a mathematical function to obtain the modulation envelope.
13. Preferably the signal and/or the other signal includes an audio signal.
14. According to a third aspect of the invention there is provided a method of processing a digital audio filter, said method comprising the steps of:
providing an audio signal including a plurality of neighbouring sample points;
performing a sample rate increase on the audio signal to provide a plurality of intermediate sample points between adjacent of the neighbouring sample points, said intermediate points being populated dependent on a weighted influence determined in the time domain of a predetermined number of the neighbouring sample points;
applying the audio signal to the digital filter where:
15. Preferably the method comprises an initial step of including additional sample points in the audio signal between the neighbouring sample points to effect time-stretching of the audio signal prior to its application to the filter. Alternatively or additionally the method involves applying the signal to the filter at an adjusted sampling rate to effect pitch-shifting of another audio signal to be filtered by the digital filter.
16. According to a fourth aspect of the invention there is provided a computer or device-readable medium including instructions for processing a digital signal or filter, said instructions when executed by a processor cause said processor to:
provide a digital filter or signal including a plurality of neighbouring sample points;
perform a sample rate increase on the digital filter or signal to provide a plurality of intermediate sample points between adjacent of the neighbouring sample points, said intermediate points being populated dependent on a weighted influence determined in the time domain of a predetermined number of the neighbouring sample points;
apply the digital filter to the signal or vice versa where
17. According to a fifth aspect of the invention there is provided a computer or device-readable medium including instructions for processing a digital signal or filter, said instructions when executed by a processor cause said processor to:
provide another digital signal or filter including a plurality of neighbouring samples either side of the mid-point sample;
perform a sample rate increase on the digital filter or signal to provide a plurality of intermediate sample points between adjacent of the neighbouring sample points, said intermediate points being populated dependent on a weighted influence determined in the time domain of a predetermined number of the neighbouring sample points;
apply the other signal to the signal, or the other filter to the filter where:
18. According to a sixth aspect of the invention there is provided a computer system for processing a digital signal or filter, said computer system comprising:
a digital filter or signal including a plurality of neighbouring sample points;
a processor configured to:
perform a sample rate increase on the digital filter or signal to provide a plurality of intermediate sample points between adjacent of the neighbouring sample points, said intermediate points being populated dependent on a weighted influence determined in the time domain of a predetermined number of the neighbouring sample points;
apply the digital filter to the signal or vice versa where:
19. Accordingly to a seventh aspect of the invention there is provided a computer system for processing a digital signal or filter, said computer system comprising:
another digital signal or filter including a mid-point sample and a plurality of neighbouring samples either side of the mid-point sample;
a processer configured to:
perform a sample rate increase on the digital filter or signal to provide a plurality of intermediate sample points between adjacent of the neighbouring sample points, said intermediate points being populated dependent on a weighted influence determined in the time domain of a predetermined number of the neighbouring sample points;
apply the other signal to the signal, or the other filter to the filter where:
20. In order to achieve a better understanding of the nature of the present invention a preferred embodiment of a method of digitally processing a signal will now be described, by way of example only, with reference to the accompanying drawings in which:
21.
(1). The digital signal is processed to provide either valve-type or transistor-type emulation in the signal;
(2). The digital signal undergoes various steps in signal processing to effect signal compression;
(3). The digital signal is processed for time-stretching or pitch-shifting;
(4). Filters are constructed for processing of the digital signal in order to influence its frequency response.
22. The invention in its preferred form is embodied in computer program code or software such as plugin software. The digital filter of the digital signal processor, or the digital signal, is represented by a particular frequency response. The particular frequency response is generally dependent on the impulse response of the filter or the signal which is characterised by the software or techniques of the various embodiment of this invention. The invention in its preferred embodiments is intended to cover the basic types of frequency response by which digital filters are classified including lowpass, highpass, bandpass and bandreject or notch filters. The digital filters are broadly categorised as Finite Impulse Response (FIR) or Infinite Impulse Response (IIR) filters.
23. In each of the applications described in the preceding paragraphs the method involves application of digital signals (or filters) to one another by adopting a modified convolution technique. In the case of filtering a digital signal, the general steps involved are as follows:
(1) Providing a digital filter including a mid-point sample and a plurality of neighbouring sample points either side of the mid-point sample;
(2) Performing a sample rate increase on the digital filter to provide a plurality of intermediate sample points between adjacent of the neighbouring sample points, said intermediate points being populated dependent on a weighted influence of a predetermined number of the neighbouring sample points;
(3) Applying the digital filter to the signal where:
24.
(1) Sine functions for absolute time values represented in the time domain;
(2) Sine functions for time values represented in the time domain from zero (0) to positive infinitely only;
(3) Sinc functions for values represented in the time domain from zero (0) to positive infinity only;
(4) Absolute values of a cosine function in the time domain substantially limited to half a waveform cycle at its mid-point.
25. The applicant's co-pending International patent application PCT/AU2014/000317 describes in some detail waveforms constructed from components represented by sine functions for absolute time values. The disclosures of this PCT application are to be considered included herein by nature of this reference.
26. The waveform components may also be adjusted by for example applying a mathematical function to each of the components or the waveforms themselves. In one embodiment the waveforms or their components may be modified by applying an averaging curve having its width in the time domain adjusted proportional to the wavelength of the respective wave-forms. This modification of the waveform components is discussed in the applicant's co-pending International patent application No. PCT/AU2014/000321 the disclosures of which are to be considered included herein by nature of this reference.
27.
28. The filter 10 and the signal 12 have each undergone the same sample rate increase. For simplicity each of the filter 10 and signal 12 have been shown with a single intermediate sample point such as 22a and 22A (shown by hollow dots) respectively between adjacent of the neighbouring sample points, such as 16/18a and 18A/20. In practice it will be understood that depending on the sample rate increase or required resolution there will be many more intermediate sample points. The sample rate increase is performed by populating each of the intermediate sample points depending on a weighted influence determined in the time domain of a predetermined number of the neighbouring sample points. For example 1,024 neighbouring sample points may be taken into account with 512 sample points either side of the intermediate sample point being populated. The weighted influence may be calculated for each of the intermediate sample points by any one of the following exemplary techniques involving:
(1) Representative waveforms at respective of neighbouring sample points where values are combined at the intermediate sample points;
(2) A representative waveform at the intermediate sample point where values are combined at the neighbouring sample points;
(3) Representative waveforms at respective neighbouring sample points shifted midway to the intermediate sample point with values combined at the intermediate sample point;
(4) A representative waveform at the neighbouring sample points shifted midway to the intermediate sample point with values combined from the neighbouring sample points.
29. The filter of this embodiment may be constructed from cosine components represented by absolute values of a cosine function in the time domain substantially limited to half a waveform cycle at its mid-point. The half-cycle cosine components are summed across the relevant frequency range to obtain the relevant waveform, see for example
30.
31. The digital signal S at n[1] is for simplicity represented with three weighted waveform components designated wc[1, 1] and wc[1, 2]. In practice the signal S will be represented by many more waveform components sufficient to cover the frequency content of the signal. Each of the waveform components are in this embodiment represented by absolute values of a cosine function in the time domain substantially limited to half a waveform cycle at its mid-point. The waveform components such as wc[1, 1], wc[1, 2] and wc[1, 3] are combined or in this exampled summed at the relevant neighbouring sample point n[1] to obtain a representative waveform designated rw[1]. With this technique the weighted influence is calculated by combining values for the representative waveforms such as rw[1] and rw[2] at the intermediate sample point by i[1]. The determined values at the intermediate sample point i[1] are in this example designated i[1, 1] and i[1, 2]. These steps are repeated for each of the intermediate sample points in order to populate the digital signal or filter at the increased sample rate.
32.
33.
34.
35. In this embodiment the sample points of the digital signal 12 may be offset relative to the respective sample points of the filter 10 or vice versa by expansion of a filter waveform representative of an initial digital filter (not shown). The sample points of the initial digital filter or waveform align in the time domain with respective and corresponding sample points of the signal 12 and thus is of the same sample resolution. In this example the initial filter may be expanded in its time domain by a factor or multiplier of between two (2) and ten (10). The filter 10 has undergone its weighted sample rate increase to the increased or required resolution prior to its expansion. The expansion factor is in this example proportional to the sample rate increase.
36. In this embodiment the digital filter 10 is applied to the signal 12 at an adjusted sampling rate. This adjusted sampling rate compensates for and is substantially proportional to the offset of the sample points 18A to 18D of the signal 12 relative to the respective neighbouring sample points 18a to 18d of the filter 10. In this example the filter 10 is thus applied to the signal 12 at a fraction of the sample resolution of the initial filter.
37. In this example the application of the filter 10 to the signal 12 involves a modified form of convolution based on dot products of values at neighbouring and intermediate sample points of the filter and the signal, respectively. In considering a limited numbering of sample points this dot product methodology involves:
(1) Multiplying the mid-point sample 16 of the filter 10 with the corresponding sample point 20 of the signal 12;
(2) Multiplying each of the neighbouring sample points 18a to 18d of the filter 10 with their respective offset neighbouring sample points 18A to 18D of the signal 12 for a predetermined number of sample points;
(3) Multiplying each of the intermediate sample points 22a to 22d of the filter 10 with their respective offset intermediate sample points 22A to 22D of the signal 12;
(4) Summing the product results from the multiplications of steps 1 to 3;
(5) Shifting or stepping the digital filter 10 at its adjusted sampling rate wherein the mid-point sample 16 of the filter 10 aligns in the time domain with the offset sample point 18A of the signal 12;
(6) Repeating steps 1 to 4 with the filter 10 at this position;
(7) Continuing this modified convolution at the adjusted sampling rate stepping across a predetermined number of sample points in the signal 12.
38. In another aspect of the invention a digital signal is processed by imposing the frequency response of another digital signal. This is effected by applying the other signal to the signal adopting the modified convolution techniques described in the preceding paragraphs. In this instance the filter of the preceding aspect is replaced with the other signal. In other respects the processing of the digital signal is essentially the same as the previous aspect and has application in audio (such as valve and transistor type) emulation, and signal compression.
39. The other signal may be in the form of a modulation envelope constructed from any one or more of the following:
(1) Proportional representation of changes, such as volume changes, in the signal, see for example the applicant's copending U.S. provisional patent application No. 62/159,027;
(2) Mathematical functions depending on the required processing of the signal;
(3) Trigonometric functions or waveforms representative of the signal, such as half a waveform cycle of absolute values of cosine functions;
(4) Clipping the signal below a threshold value to directly derive the required modulation envelope in the form of a noise gate.
40. The other signal (or modulator) and signal may be reversed where the signal is applied to the modulator by the modified convolution technique. The modulator may be otherwise mathematically manipulated to obtain the required envelope.
41.
42. In an alternative embodiment the signal may be applied to the other signal which has its cosine and/or sine components replaced with square or triangular-shaped waveforms. In this example the frequency response of the signal is imposed on the other signal with both signals being in phase. In each of these alternative embodiments the cosine and/or sine component of the other signal may be at least in part replaced with a band of square-wave components which are summed across the relevant frequency range.
43. In another application random noise may be imposed on an audio signal by applying the modified convolution technique of the previous embodiments. In this case the random noise is applied as the other signal and is represented by discrete values, for example a mathematical function. The random noise signal may be adjusted by replacing discrete values with for example either a zero (0) or one (1). The addition of random zeros is understood to introduce odd or lower harmonics into the signal whereas the addition of random ones promotes even or upper harmonics. The random noise may be applied to the audio signal using the earlier-described convolution technique at an adjusted sampling rate which is a fraction, say substantially one eighth (⅛), the sample resolution. The other signal can thus be processed so that when applied to the signal it modulates or influences the audio signal. This process may be reversed where the signal is applied to the random noise using modified convolution.
44.
(1) The absolute values of the other signal are limited to a threshold value to obtain a limited waveform;
(2) The limited waveform is divided by the other signal to provide another signal or modulation envelope which is to be applied in modified convolution to the signal to effect its compression.
45. In its simplest form the other signal 30 is clipped to the threshold value to provide a clipped waveform 32 (shown bolded). The clipped waveform 32 is divided by the other signal 30 to provide another signal 34 which is to be convolved with the signal for compression. The other signal 34 may alternatively be limited by applying a mathematical function to obtain a desired variant of the limited waveform.
In another embodiment the signal may for a particular frequency range be represented in its cosine and sine components. These cosine and sine components for that frequency range or band are then combined to provide a virtual or mathematical filter for limiting. The filter may be limited to the required modulation envelope by applying a transfer function. The virtual filter may be applied in modified convolution directly to the equivalent cosine and sine components of the signal, or another filter which represents these cosine and sine components in accordance with another aspect of the invention.
46. In another aspect the invention involves modified convolution in the application of an audio signal to a digital filter. This aspect has application in time-stretching or pitch-shifting where:
(1) One or more sample points are inserted in the audio signal between neighbouring sample points to provide time-stretching;
(2) The time-stretched signal may be applied to the filter at an adjusted sampling rate;
(3) The adjusted sampling rate is varied to effect pitch-shifting of another audio signal to be filtered by the resulting digital filter.
The signal may be pitch-shifted without necessarily including additional sample points to time-stretch it. In this variation the signal is applied to the filter at a sampling rate different to its sample resolution to provide the required pitch-shifting.
47. The applicant's co-pending International patent application No. PCT/AU2014/000319 described several techniques for sample rate increases. The disclosures of this PCT application are to be considered included therein by nature of this reference.
48. The signal or filter may be processed before and/or after modified convolution as earlier-described. This processing includes but is not limited to the application of:
(1) One or more filters constructed from cosine and/or sine components having the application of proportional averaging curves in the time domain, see for example the applicant's International patent application no. PCT/AU2014/000321;
(2) Sample rate increases to the signal and/or filter using for example techniques disclosed in the applicant's International patent application no.'s PCT/AU2014/000318 and PCT/AU2014/000319 together with US provisional patent application No. 62/056,349;
(3) Sample zoning and the combination of sample values depending on the frequency of the zone as disclosed in the applicant's International patent application no. PCT/AU2015/000197;
(4) Modulation envelopes such as volume envelopes or frequency envelopes where for example a high pass filter is applied to the signal prior to convolution;
(5) Nyquist or other filters, for example the application of a Nyquist filter to noise prior to its application to a signal in modified convolution.
49. It is to be understood that the methods and techniques described can be implemented as computer-readable instructions stored on a computer-readable medium. The computer-readable instructions can be executed by a processor of practically any computer system including desktop, portable, tablet, hand-held, and/or any other computer device.
50. It is also to be understood that the present invention extends to computer-readable media for carrying or having computer-executable instructions stored thereon. The computer-readable media include RAM, ROM, EEPORM, CD-ROM or other optical disc storage, magnetic disc storages, or any other medium which carries or stores program code means in the form of computer-executable instructions. In the event of information being transferred or provided over a network or another communications connection to a computer, the computer is to be understood as viewing the connection (hardwired, wireless, or a combination thereof) as a computer-readable medium.
51. The contents of the applicant's following co-pending patent applications are to be taken as incorporated herein by these references:
(1) PCT/AU2014/000325 titled “Audio Filtering with Virtual Sample Rate Increases”;
(2) PCT/AU2014/000318 titled “Audio Sample Rate Increases”;
(3) PCT/AU2015/000197 titled “Modified Digital Filtering with Sample Zoning”.
52. Now that several preferred embodiments of the invention have been described it will be apparent to those skilled in the art that the method of processing the digital signal has at least the following advantages:
(1) The various techniques adopted in modified convolution lend themselves to a range of audio processing applications;
(2) The modified convolution techniques are adopted in the time domain with mathematical functions representing the signal or filter or components thereof;
(3) The modified techniques for convolution also lend themselves to processing of signals designed to obtain subsequent audio effects when applied to audio signals.
53. Those skilled in the art will appreciate that the invention described herein is susceptible to variations and modifications other than those specifically described. The processing of audio signals need not be limited to acoustics but extends to other sound applications including ultrasound and sonar. The invention also extends beyond audio signals to other signals including signals derived from a physical displacement such as that obtained from measurement devices, for example a strain gauge or other transducer which generally converts displacement into an electronic signal. The invention also covers digital processing of signals associated with digital communications. The invention in another embodiment is applied to imaging.
54. The filter(s) may be constructed or represented by fast fourier transform (FFT) algorithms rather than the trigonometric functions described in the preferred embodiments, such as the cosine and/or sine components. It is also possible that convolution can be applied in the frequency domain instead of the time domain. For example filtering in the frequency domain can involve application of FFT algorithms.
55. All such variations and modifications are to be considered within the scope of the present invention the nature of which is to be determined from the foregoing description.
Filing Document | Filing Date | Country | Kind |
---|---|---|---|
PCT/AU2015/050576 | 9/26/2015 | WO | 00 |
Number | Date | Country | |
---|---|---|---|
62056343 | Sep 2014 | US | |
62091312 | Dec 2014 | US | |
62165849 | May 2015 | US |