WO2014206265A1 - 谐波分析方法和装置以及确定谐波间杂波的方法和装置 - Google Patents

谐波分析方法和装置以及确定谐波间杂波的方法和装置 Download PDF

Info

Publication number
WO2014206265A1
WO2014206265A1 PCT/CN2014/080562 CN2014080562W WO2014206265A1 WO 2014206265 A1 WO2014206265 A1 WO 2014206265A1 CN 2014080562 W CN2014080562 W CN 2014080562W WO 2014206265 A1 WO2014206265 A1 WO 2014206265A1
Authority
WO
WIPO (PCT)
Prior art keywords
spectrum
filtering
peaks
wave
peak
Prior art date
Application number
PCT/CN2014/080562
Other languages
English (en)
French (fr)
Inventor
王喆
Original Assignee
华为技术有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 华为技术有限公司 filed Critical 华为技术有限公司
Publication of WO2014206265A1 publication Critical patent/WO2014206265A1/zh

Links

Classifications

    • GPHYSICS
    • G01MEASURING; TESTING
    • G01RMEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R23/00Arrangements for measuring frequencies; Arrangements for analysing frequency spectra
    • G01R23/16Spectrum analysis; Fourier analysis
    • G01R23/165Spectrum analysis; Fourier analysis using filters
    • GPHYSICS
    • G10MUSICAL INSTRUMENTS; ACOUSTICS
    • G10LSPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L21/00Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
    • G10L21/02Speech enhancement, e.g. noise reduction or echo cancellation
    • G10L21/0208Noise filtering
    • G10L21/0216Noise filtering characterised by the method used for estimating noise
    • G10L21/0232Processing in the frequency domain

Definitions

  • This invention relates to signal processing, and more particularly to methods and apparatus for harmonic analysis and methods and apparatus for determining inter-harmonic clutter. Background technique
  • Harmonic analysis is a basic method of signal processing and analysis. It has a wide range of applications in signal recognition, classification, coding, and enhancement. Harmonic analysis can cover a variety of different analytical purposes, such as calculating the number of harmonics, calculating the fundamental frequency of harmonics, calculating the strength of harmonics, calculating the intra/inter correlation of harmonics, and calculating the harmonic-to-noise ratio. and many more. Harmonic analysis can be based on time domain signals for analysis or based on frequency domain signals. In general, harmonic analysis is most common on frequency domain signals, and the effect is better.
  • a common method is to use the peak-to-average ratio method in the frequency domain to determine the harmonics.
  • the spectrum of the audio frame is "fuzzy", that is, there is a strong clutter between the harmonics, the energy of the harmonic is weakened.
  • the peak-to-average ratio of the sub-band containing the harmonics is not harmonic.
  • the difference in the peak-to-average ratio of the sub-bands of the wave is not so obvious.
  • a simple error is obtained by comparing the peak-to-average ratio with the threshold to determine whether or not harmonics exist.
  • Another common method is to obtain the peak intensity based on the relative energy (or amplitude) relationship of the peaks and troughs. Finally, compare this intensity with a threshold to determine whether there is a harmonic. However, if the spectrum of the audio frame is "fuzzy", that is, there is a strong clutter between the harmonics, due to the existence of harmonic clutter, it is possible to treat the trough error of the clutter near the harmonic as a harmonic The wave trough of the wave causes the calculated harmonic intensity to be below the true value and is missed.
  • the present invention proposes a harmonic analysis method and apparatus and a method and apparatus for determining interharmonic clutter to solve the problem of obtaining a clear and accurate harmonic spectrum.
  • a method for harmonic analysis comprising: performing time-frequency conversion on an input time domain signal to obtain a frequency spectrum of the time domain signal in a frequency domain; and filtering the spectrum to Determining a target spectrum; performing a process of eliminating harmonics between the harmonics of the target spectrum to determine a structure of harmonics of the target spectrum; and performing harmonic analysis on the harmonics by using a structure of harmonics of the target spectrum.
  • the filtering processing the spectrum to determine a target spectrum includes: performing k filtering processing on the spectrum, and performing the filtering on the spectrum Recording the number of peaks after each filtering process of the k-th filtering process to obtain k peaks, wherein k is a positive integer greater than or equal to 2; determining the minimum number of peaks in the k peaks; determining The spectrum of the filtering process for obtaining the minimum number of peaks is the target spectrum.
  • the performing filtering processing on the spectrum to determine a target spectrum includes: performing a first filtering process on the spectrum, and recording the first a first peak of the filtered spectrum; performing a second filtering process on the spectrum, and recording a second peak of the spectrum subjected to the second filtering; performing a third filtering process on the spectrum, and Recording a third peak number of the spectrum subjected to the third filtering process; when the second peak number is smaller than the first peak number, and the second peak number is smaller than the third peak number And determining that the spectrum obtained by the second filtering process is the target spectrum.
  • the first filtering process is first-order filtering
  • the second filtering process is second-order filtering
  • the third The filtering process is a third-order filtering
  • the filtering processing the spectrum to determine a target spectrum includes: performing first-order filtering on the spectrum, and recording a first peak of the spectrum that passes through the first-order filtering Performing a second-order filtering on the spectrum, and recording a second peak number of the spectrum filtered by the second order; performing third-order filtering on the spectrum, and recording the third-order filtering The number of the third peaks of the spectrum; when the number of the second peaks is less than the number of the first peaks, and the number of the second peaks is less than the number of the third peaks, determining that the second Order filtering frequency
  • the spectrum is the target spectrum.
  • the performing filtering processing on the spectrum to determine a target spectrum includes: performing low-pass filtering on the spectrum by using a low-pass filter, The frequency word processed by the low pass filtering is used as a target spectrum.
  • the performing filtering processing on the spectrum to determine a target spectrum includes: performing fourth-order filtering on the spectrum; determining, after the fourth-order The filtered spectrum is the target spectrum, wherein the fourth order is a fixed value.
  • the processing of the inter-harmonic clutter of the target spectrum is performed, including using at least one of the following manners : canceling the interharmonic clutter of the target spectrum according to the lobe width, wherein the lobe width is a width between two nearest neighboring troughs on both sides of the peak; the target is according to the inclination of the peak
  • the inter-harmonic clutter of the spectrum is subjected to cancellation processing, wherein the inclination of the peak is an angle between a line connecting two nearest troughs on both sides of the peak and a horizontal direction, or two sides of the peak The slope of the line between the nearest neighboring troughs, or the trigonometric angle of the line between the two nearest neighboring troughs on either side of the crest.
  • the canceling processing the interharmonic clutter of the target spectrum according to the lobe width includes: determining all the target spectrum a peak and a trough; determining a lobe width in the target spectrum according to the determined peak and the trough; and determining a peak corresponding to the lobe width when the lobe width is smaller than the first lobe width
  • the crest of the clutter; the clutter of the target spectrum is cancelled by eliminating the lobes of the clutter.
  • the first lobe width is a preset fixed value, or the first lobe width is an average lobe width, where The average lobe width is a ratio of a width of the target spectrum to a number of the second peaks.
  • the removing the clutter of the target spectrum according to the inclination of the peak includes: determining all peaks and troughs of the target spectrum; Determining the inclination of the peak according to the determined peak and the trough; determining that the peak is a clutter peak when the inclination of the peak is greater than the first threshold; removing the wave of the clutter by The lobes eliminate clutter of the target spectrum.
  • the first threshold is a preset fixed value.
  • the lobe of the clutter is located at a frequency point amplitude between the two troughs at two trough amplitudes Interpolation between instead.
  • a method for determining inter-harmonic clutter comprising: determining all peaks and troughs of the spectrum; determining a lobe width in the spectrum according to the determined peak and the trough, wherein The lobe width is a width between two nearest neighboring troughs on both sides of the peak; when the lobe width is smaller than the first lobe width, it is determined that the peak corresponding to the lobe width is a peak of the clutter.
  • the first lobe width is a predetermined fixed value.
  • the first lobe width is an average lobe width, wherein the average lobe width is a width of the spectrum and a number of all peaks The ratio.
  • a method for determining inter-harmonic clutter comprising: determining all peaks and troughs of the target spectrum; determining an inclination of the peak according to the determined peak and the trough; If the slope of the peak is greater than the first threshold, then the peak is determined to be the peak of the clutter.
  • the inclination of the peak is an angle between a line between two nearest neighboring troughs on both sides of the peak and a horizontal direction; or
  • the slope of the peak is the slope of the line between the two nearest troughs on either side of the peak: or the slope of the peak is the line between the two nearest troughs on either side of the peak.
  • the first threshold is a preset fixed value.
  • an apparatus for harmonic analysis comprising: a time-frequency converting unit, configured to perform time-frequency conversion on an input time domain signal to obtain a frequency spectrum of the time domain signal in a frequency domain; And performing filtering processing on the spectrum converted by the time-frequency conversion unit to determine a target spectrum; a clutter cancellation unit, configured to perform a process of eliminating harmonics between the harmonics of the target spectrum determined by the filtering unit, to determine a structure of harmonics of the target spectrum; and a harmonic analysis unit for utilizing The structure of the harmonics of the target spectrum performs harmonic analysis on the harmonics from which the clutter cancellation unit eliminates clutter.
  • the filtering unit is specifically configured to: perform k filtering processing on the spectrum, and perform filtering on the spectrum after the k filtering processing After processing, the number of peaks is recorded to obtain k peaks, wherein k is a positive integer greater than or equal to 2; determining the minimum number of peaks in the k peaks; determining the filter to obtain the minimum number of peaks
  • the processed spectrum is the target spectrum.
  • the filtering unit is specifically configured to: perform a first filtering process on the spectrum, and record a first peak of the spectrum that is processed by the first filtering Performing a second filtering process on the spectrum, and recording a second peak number of the spectrum subjected to the second filtering process; performing third filtering processing on the spectrum, and recording the third filtering process a third peak of the spectrum; when the number of the second peaks is less than the number of the first peaks, and the number of the second peaks is smaller than the number of the third peaks, determining that the second filtering is performed
  • the processed spectrum is the target spectrum.
  • the first filtering process is first-order filtering
  • the second filtering process is second-order filtering
  • the third The filtering process is a third-order filtering
  • the filtering unit is specifically configured to: perform first-order filtering on the spectrum, and record a first peak number of the spectrum that is filtered by the first-order filtering; Second-order filtering, and recording the second peak number of the spectrum after the second-order filtering; performing third-order filtering on the spectrum, and recording the third peak number of the spectrum passing through the third-order filtering;
  • the filtering unit is specifically configured to: perform a low-pass filtering process on the spectrum by using a low-pass filter, and perform the low-pass filtering process The latter frequency is used as the target spectrum.
  • the filtering unit is specifically configured to: perform fourth-order filtering on the spectrum; and determine, by the fourth-order filtered spectrum, a target spectrum; Wherein the fourth order is a fixed value.
  • the clutter cancellation unit is configured to: at least one of the following: locating the target spectrum according to a lobe width Harmonic between harmonics is eliminated, wherein the width of the lobes is the width between two nearest neighboring troughs on both sides of the peak; and the harmonics between the harmonics of the target spectrum are eliminated according to the inclination of the peak, wherein The inclination of the peak is an angle between a line between two nearest neighboring troughs on either side of the peak and a horizontal direction, or a line between two nearest neighboring troughs on either side of the peak The slope, or a trigonometric function of the angle between the lines between the two nearest neighbors on either side of the peak.
  • the clutter cancellation unit is specifically configured to: determine all peaks and troughs of the target spectrum; according to the determined peak and Determining a lobe width, determining a lobe width in the target spectrum; determining that a peak corresponding to the lobe width is a peak of a clutter when the lobe width is smaller than a first lobe width; The lobes eliminate clutter in the target spectrum.
  • the first lobe width is a preset fixed value, or the first lobe width is an average lobe width, where The average lobe width is a ratio of a width of the target spectrum to a number of the second peaks.
  • the clutter cancellation unit is specifically configured to: determine all peaks and troughs of the target spectrum; according to the determined peak and Determining the slope of the peak; determining the peak as a clutter peak when the slope of the peak is greater than the first threshold; eliminating the target spectrum by erasing the lobe of the clutter Clutter.
  • the first threshold value is a preset fixed value.
  • the clutter eliminating unit is specifically configured to: locate the lobe of the clutter between two troughs The frequency amplitude is replaced by an interpolation between the two valley amplitudes.
  • an apparatus for determining inter-harmonic clutter including: a first determining unit, configured to determine all peaks and troughs of the spectrum; and a second determining unit, configured to determine, according to the first determining unit The peak and the trough, determining a lobe width in the spectrum, wherein The lobe width is a width between two nearest neighboring troughs on both sides of the peak; a third determining unit, configured to determine, by the second determining unit, when the lobe width is smaller than the first lobe width The peak corresponding to the lobe width is a peak of a clutter.
  • the first lobe width is a predetermined fixed value.
  • the first lobe width is an average lobe width, wherein the average lobe width is a width of the spectrum and a number of all peaks The ratio.
  • an apparatus for determining inter-harmonic clutter including: a fourth determining unit, configured to determine all peaks and troughs of the target spectrum; and a fifth determining unit, configured to determine, according to the fourth determining unit Determining the peak and the trough, determining the inclination of the peak; a sixth determining unit, configured to determine, when the inclination of the peak is greater than the first threshold, determine the location determined by the fifth determining unit The peak is the peak of the clutter.
  • the inclination of the peak is an angle between a line between two nearest neighboring troughs on both sides of the peak and a horizontal direction; or
  • the slope of the peak is the slope of the line between the two nearest troughs on either side of the peak: or the slope of the peak is the line between the two nearest troughs on either side of the peak.
  • the first threshold is a preset fixed value.
  • the embodiment of the invention determines the target spectrum of the harmonic analysis by filtering, and can effectively eliminate the clutter interference, thereby obtaining a clear and accurate harmonic spectrum, thereby facilitating harmonic analysis.
  • FIG. 1 is a flow chart of a method of harmonic analysis in accordance with an embodiment of the present invention.
  • Figure 2 is a schematic diagram of the elimination of clutter.
  • Figure 3 shows the method of determining clutter by the slope of the peak.
  • FIG. 4 is a flow chart of a method of determining interharmonic clutter in accordance with an embodiment of the present invention.
  • FIG. 5 is a flow chart of a method of determining interharmonic clutter in accordance with another embodiment of the present invention.
  • FIG. 6 is a schematic structural diagram of an apparatus for harmonic analysis according to an embodiment of the present invention.
  • Figure 7 is a block diagram showing the structure of an apparatus for determining interharmonic clutter according to an embodiment of the present invention.
  • FIG. 8 is a block diagram showing the structure of an apparatus for determining inter-harmonic clutter according to another embodiment of the present invention.
  • FIG. 9 is a schematic structural diagram of an encoder according to an embodiment of the present invention. detailed description
  • FFT Fast Fourier Transform
  • DCT Discrete Cosine Transform
  • the frequency word is subjected to k filtering processing, and after the frequency word passes the filtering process of the k times filtering process, the number of peaks is recorded to obtain k peak numbers, wherein k is a positive integer greater than or equal to 2; determining a minimum number of peaks in the number of k peaks; determining a spectrum of the filtering process for obtaining the minimum number of peaks as a target spectrum.
  • a first filtering process is performed on the spectrum, and a first peak number of a spectrum that passes through the first filtering process is recorded; a second filtering process is performed on the spectrum, and a recording process is performed. a second peak of the spectrum of the second filtering process; performing a third filtering on the spectrum Processing, and recording a third peak number of the spectrum subjected to the third filtering process; when the second peak number is smaller than the first peak number, and the second peak number is smaller than the third The number of peaks determines that the spectrum subjected to the second filtering process is the target spectrum.
  • the first filtering process may be the first-order filtering
  • the second filtering process may be the second-order filtering
  • the third filtering process may be the third-order filtering. It should be understood that the filtering order corresponding to the first-order filtering, the second-order filtering, and the third-order filtering, respectively, is different. In this way, the first-order filtering is performed on the spectrum, and the first peak number of the spectrum passing through the first-order filtering is recorded; the second-order filtering is performed on the spectrum, and the second-order filtering is recorded.
  • a second peak of the spectrum performing third-order filtering on the spectrum, and recording a third peak number of the spectrum filtered by the third order; when the second peak is smaller than the first peak If the number of the second peaks is smaller than the number of the third peaks, it is determined that the spectrum filtered by the second order is the target spectrum.
  • the above filtering process can be cyclic. For example, after the first-order filtering, the second-order filtering, and the third-order filtering in the first filtering combination, if the number of the second peaks is not simultaneously smaller than the first peak number and the third peak number, then Going through the first-order filtering, the second-order filtering, and the third-order filtering in the second filtering combination again, until a second filtering combination is found to be less than the first peak number and the third peak number.
  • the number of peaks That is to say, the process of finding the minimum spectral peaks by such cyclic filtering is a process of finding the optimal filtering and determining the target spectrum.
  • the optimal filtering occurs at the inflection point where the peak number of the peaks is decremented to increasing.
  • an early termination judgment is added to the loop process for determining the target spectrum, that is, once the inflection point is detected, it is considered that the optimal filtering has been found, and the remaining loops are not executed.
  • each cycle is repeated to compare whether the number of peaks of the target spectrum after the current cycle is less than or equal to the number of peaks of the target spectrum after the previous cycle. If yes, continue the loop. If not, the loop is terminated.
  • the target frequency after the previous loop is used as the target spectrum of the subsequent step, that is, the filtering of the previous loop is used as the optimal filter.
  • the filtering order is no longer dynamically determined, but each frame is filtered with a fixed order.
  • a large sample Statistics show that for a given sample rate and frame length, the probability of an optimal filter order at a certain fixed order is significantly higher than other orders. Therefore, using the maximum occurrence probability order as a fixed filtering order can maximize the approximation of the optimal filtering result while saving computational complexity.
  • a fourth filtering process is performed on the spectrum; and a spectrum obtained by the fourth filtering process is determined to be a target spectrum.
  • the fourth filtering process is fourth-order filtering, and the fourth-order filtering is performed on the spectrum, wherein the fourth-order is a fixed value; and the spectrum that is filtered by the fourth-order is determined to be a target spectrum.
  • the spectrum may be low pass filtered using any form of low pass filter to determine the target spectrum as the low pass filtered spectrum.
  • the interharmonic clutter of the target spectrum may be eliminated by using at least one of the following ways: removing the interharmonic clutter of the target spectrum according to the lobe width; according to the inclination of the peak The harmonics between the harmonics of the target spectrum are eliminated.
  • the method for eliminating the interharmonic clutter of the target spectrum according to the lobe width includes the following steps: First, determining all peaks and troughs of the target spectrum, wherein the energy of a certain frequency point in the spectrum is higher than its two The energy closest to the frequency point is considered to be a peak of the spectrum at the frequency point, and the trough is defined as the frequency point with the smallest energy between two adjacent peaks; then, according to the determined peak and the said a valley, the lobe width in the target spectrum is determined, wherein the lobe is a set of frequency points between two nearest troughs on either side of a peak, the lobe width being two on both sides of the peak a width between the nearest valleys; then, when the lobe width is smaller than the first lobe width, determining that the peak corresponding to the lobe width is a peak of the clutter; and finally, by eliminating the clutter
  • the lobes eliminate the clutter of the target spectrum, for example by locating the lobes of the clutter
  • the first lobe width is a preset fixed value, or the first lobe width is an average lobe width, wherein the average lobe width is a width of the target spectrum and the second peak number The ratio. Determining all peaks and troughs of the target spectrum; and then determining an inclination of the peak according to the determined peak and the trough; and then, when the inclination of the peak is greater than the first threshold, Determining that the peak is a clutter peak; finally, canceling the clutter of the target spectrum by eliminating the lobe of the clutter, for example, the lobe of the clutter is located at a frequency energy between two troughs Interpolated between the two troughs instead.
  • the inclination of the peak described herein may be that the inclination of the peak is an angle between a line between two nearest neighboring troughs on both sides of the peak and a horizontal direction; or may be both sides of the peak
  • the first threshold may be a preset fixed value or may be related to a filtering order of the target spectrum.
  • the pin performs harmonic analysis on the harmonics using a structure of harmonics of the target spectrum.
  • calculate the number of harmonics calculate the fundamental frequency of the harmonics, calculate the strength of the harmonics, calculate the intra/inter correlation of the harmonics, calculate the harmonic-to-noise ratio, and so on.
  • the embodiment of the present invention determines the target spectrum of the harmonic analysis by filtering, and can effectively eliminate the clutter interference, thereby obtaining a clear and accurate harmonic spectrum, thereby facilitating harmonic analysis.
  • an audio encoder is taken as an example.
  • the time-frequency domain transforms commonly found in the audio field include FFT, DCT, and Modified Discrete Cosine Transform (MDCT).
  • this embodiment uses MDCT transform as an example to illustrate a calculation of a 32 kHz sample audio signal.
  • the harmonic analysis scheme of the number of harmonics in the 8k-16kHz frequency band, the input audio frame is 20ms frame length.
  • the time domain audio frame must be windowed before the MDCT transform. Since it is a well-known technology, it will not be described in detail here.
  • the default s(x) is the windowed time domain audio frame.
  • the number of time domain audio frames s(x) depends on the frame length and sampling rate of the time domain audio frame. In this embodiment, based on the 20 ms frame length and the 32 Hz sampling rate, it can be determined that the number of time domain audio frames s(x) is 640. It is also easy to understand, the frequency domain audio frame after the time-frequency transform The number of points depends on the number of time domain frames and the time-frequency transform method. In this particular embodiment, according to
  • the MDCT transform method can determine that the number of samples of the frequency domain audio frame M(x) is also 640.
  • spectrum S(i) can be determined according to the following formula 1:
  • the specific filtering method is one or a combination of the following formulas (2) to (4):
  • the third step find the number of peaks p [3] in the spectrum S [3] (j). When the energy of a frequency point in the spectrum is higher than the energy of its two nearest neighbors, it is considered to be a peak of the spectrum. After this step is executed, the number of peaks obtained is recorded as 1 [3] .
  • the records mentioned here can be implemented by software, such as being stored in the content after reading, or by hardware, such as in a register.
  • the minimum value ⁇ [ ⁇ ] is found in all recorded p [k] , and its corresponding filtered spectrum is obtained, that is, S [T] (j), which is hereinafter referred to as the target spectrum.
  • the acquisition of the target spectrum S [T] (j) can be obtained directly from the spectrum previously recorded in the loop of P [kI , or it can be re-based on the value of T Due to the different characteristics of the harmonic spectrum per frame, using the same filtering method for each frame does not achieve optimal results in every frame. In general, better filtering should result in fewer spectral peaks in the target spectrum. Therefore, the process of finding the minimum spectral peak number by the above cyclic filtering is a process of finding the optimal filtering and determining the target spectrum.
  • the peaks and troughs of the target spectrum S [T] (j) are searched, and the position and energy of each peak and trough are recorded.
  • the trough is defined as the frequency at which the energy between two adjacent peaks is the smallest. The least frequent energy point between the peak and its endpoint. Usually, peaks and troughs always alternate.
  • the number of troughs is one more than the number of peaks.
  • the clutter of the target spectrum is eliminated according to the lobe width.
  • the process of obtaining the target spectrum by filtering has partially eliminated some of the clutter, but in general there will still be clutter that has not been eliminated.
  • the number of peaks in the target spectrum will be larger than the number of actual harmonics, so that the average lobe width of the target spectrum will be smaller than the harmonic lobe width.
  • the average lobe width is defined as: the ratio of the bandwidth of the target spectrum to the number of peaks.
  • each lobe width Wp(m), where m 0, 1, ... Ll.
  • the lobes are defined as a set of frequency points between two nearest neighboring troughs on either side of a peak.
  • the lobe width is defined as the width between the two troughs of the lobe, ie the frequency range over which the lobe is covered.
  • the lobe width can be determined by the following formula:
  • the clutter of the target spectrum is eliminated according to the inclination of the peak.
  • the target spectrum may still contain clutter, and other methods are needed to further eliminate it.
  • Figure 3 shows a typical clutter spectrum.
  • the peak A is a harmonic peak
  • the peak B and the peak C are clutter peaks.
  • the dotted line M is a straight line passing through the two troughs of the peak B
  • the broken line N is a straight line passing through the two troughs of the peak C
  • the angle ⁇ and the angle ⁇ are the angles of M and N with the horizontal axis, respectively.
  • the crests and crests C are significantly inclined. Specifically, the angles of the angles ⁇ and ⁇ are significantly larger, that is, the tangent of the angle ⁇ and the angle ⁇ is larger, or the slopes of the straight line ⁇ and the straight line ⁇ are larger.
  • the peak By judging the slope of a straight line passing through two valleys of a certain peak, it can be determined to some extent whether the peak is a clutter. Specifically, the obtained slope is compared with a threshold thrl. If it is greater than the threshold, the peak is considered to be a clutter.
  • the sixth step of the clutter cancellation operation has changed the target spectrum, it is necessary to re-search all the peaks and troughs of the target spectrum before eliminating the clutter of the target spectrum based on the slope of the peak.
  • the clutter is eliminated in the same manner as expressed by the formula (6) in the sixth step, and the clutter is eliminated by referring to the method shown in Fig. 2.
  • the second embodiment it is similar to most of the steps in the first embodiment, except that the second to fourth steps of the determination of the target spectrum in the first embodiment are simplified.
  • the determination of the optimum filter in the second to fourth steps of the first embodiment is achieved by a loop process and taking the optimum filter therein.
  • the optimal filtering occurs at the inflection point where the peak number of the peaks is decremented to increasing.
  • the second embodiment adds an early termination judgment to the loop process in the first embodiment, that is, once the inflection point is detected, it is considered that the optimal filtering has been found, and the remaining loop is not executed. .
  • Each cycle compare whether the number of peaks of the target spectrum after the current cycle is less than or equal to the number of peaks of the target spectrum after the previous cycle. If yes, the loop is continued. If not, the loop is terminated.
  • the target spectrum after the previous loop is used as the target spectrum of the subsequent step, that is, the filtering of the previous loop is used as the optimal filter.
  • the third embodiment it is similar to most of the steps in the first embodiment, and differs in the simplification of the second to fourth steps of determining the target spectrum in the first embodiment.
  • the filtering order is no longer dynamically determined, but is filtered at a fixed order for each frame. After counting large samples, it can be found that for a given sampling rate and frame length, the probability of the optimal filtering order at a certain fixed order is significantly higher than other orders. Therefore, using the maximum probability of occurrence as a fixed filtering order can maximize the approximation of the optimal filtering result while saving computational complexity.
  • the fixed filter order used is selected as 5.
  • the fourth embodiment it is similar to most of the steps in the first embodiment, and differs in the simplification of the second to fourth steps of determining the target spectrum in the first embodiment.
  • the input signal of the embodiment of the present invention may not be limited to an audio signal, and may be any signal that can be analyzed in the frequency domain, such as an image signal.
  • the filter used to obtain the target spectrum for filtering the original spectrum may be any form of low pass filter.
  • the following equation (8) gives an embodiment of a first-order low-pass filter.
  • the fifth embodiment it is similar to most of the steps in the first embodiment, except that in the sixth step of determining the target spectrum in the first embodiment, the wave in the process of eliminating the clutter in accordance with the lobe width is eliminated.
  • the method of determining the width of the flap is similar to most of the steps in the first embodiment, except that in the sixth step of determining the target spectrum in the first embodiment, the wave in the process of eliminating the clutter in accordance with the lobe width is eliminated. The method of determining the width of the flap.
  • the clutter is determined by comparing the lobe width of a certain peak with the average lobe width.
  • the lobe width can also be compared to a predetermined fixed value when determining clutter.
  • the threshold can be set to a fixed value of 32. If the lobe width of a peak is less than 32, it is considered as a clutter.
  • the method for determining the interharmonic clutter of the spectrum as shown in FIG. 4 includes the following steps:
  • the first lobe width may be a preset fixed value.
  • the first lobe width may be an average lobe width, wherein the average lobe width is a ratio of the width of the spectrum to the number of all peaks.
  • the method for determining the inter-harmonic clutter of the spectrum in the embodiment of the present invention can effectively remove the inter-harmonic clutter of the spectrum.
  • the method for determining the interharmonic clutter of the spectrum as shown in FIG. 5 includes the following steps:
  • the inclination of the peak is greater than the first threshold, determine that the peak is a peak of the clutter.
  • the inclination of the peak is an angle between the line between the two nearest neighboring troughs on both sides of the peak and the horizontal direction; or the inclination of the peak is;
  • the first threshold may be a preset fixed value.
  • the method for determining the harmonic clutter of the spectrum in the embodiment of the present invention can effectively remove Clutter between harmonics of the spectrum.
  • Fig. 6 is a block diagram showing the structure of an apparatus for harmonic analysis according to an embodiment of the present invention.
  • the harmonic analysis device 60 includes a time-frequency conversion unit 61, a filtering unit 62, a clutter eliminating unit 63, and a harmonic analysis unit 64.
  • the functional modules corresponding to the time-frequency converting unit 61, the filtering unit 62, the clutter eliminating unit 63, and the harmonic analyzing unit 64 are integrated in a digital signal processing (DSP) device.
  • DSP digital signal processing
  • the time-frequency converting unit 61 is configured to perform time-frequency conversion on the input time domain signal to obtain a frequency spectrum of the time domain signal in the frequency domain.
  • the filtering unit 62 is configured to perform filtering processing on the spectrum converted by the time-frequency converting unit 61 to determine a target spectrum.
  • the clutter canceling unit 63 is configured to cancel the interharmonic clutter of the target spectrum determined by the filtering unit 62 to determine the structure of the harmonics of the target spectrum.
  • the harmonic analysis unit 64 is for performing harmonic analysis on the harmonics of the clutter canceled by the clutter canceling unit 63 using the structure of the harmonics of the target spectrum.
  • the filtering unit 62 is specifically configured to: perform k-th filtering processing on the spectrum, and record the number of peaks after each filtering process of the spectrum through the k-th filtering process to obtain k peak numbers, where k is a positive integer greater than or equal to 2; determining a minimum number of peaks in the number of k peaks; determining a spectrum of the filtering process for obtaining the minimum number of peaks as a target spectrum.
  • the filtering unit 62 is configured to: perform a first filtering process on the spectrum, and record a first peak number of the spectrum that is processed by the first filtering; perform a second filtering process on the spectrum, and record the passing a second peak number of the spectrum of the second filtering process; performing a third filtering process on the spectrum, and recording a third peak number of the spectrum subjected to the third filtering process; when the second peak number If the number of the first peaks is less than the number of the first peaks, and the number of the second peaks is smaller than the number of the third peaks, it is determined that the spectrum obtained by the second filtering process is the target spectrum.
  • the filtering unit 62 is configured to: Performing a first-order filtering on the spectrum, and recording a first peak number of the spectrum that passes through the first-order filtering; performing second-order filtering on the spectrum, and recording a second spectrum of the second-stage filtered spectrum The number of peaks; performing third-order filtering on the spectrum, and recording the spectrum of the third-order filtered spectrum a number of three peaks; when the number of the second peaks is smaller than the number of the first peaks, and the number of the second peaks is smaller than the number of the third peaks, determining the spectrum after the second-order filtering For the target spectrum.
  • the filtering unit 62 is configured to: perform low-pass filtering processing on the frequency spectrum by using a low-pass filter of any form, and use the frequency word processed by the low-pass filtering as a target frequency spectrum.
  • the filtering unit 62 is configured to: perform fourth-order filtering on the spectrum, where the fourth-order is a fixed value; and determine that the spectrum that is filtered by the fourth-order is a target spectrum.
  • the clutter canceling unit 63 uses at least one of the following methods: canceling the interharmonic clutter of the target spectrum according to the lobe width, wherein the lobe width is the two most on both sides of the peak The width between adjacent troughs; the interharmonic clutter of the target spectrum is eliminated according to the inclination of the peak, wherein the inclination of the peak is between the two nearest troughs on both sides of the peak The angle between the line and the horizontal direction, or the slope of the line between the two nearest neighboring troughs on either side of the peak, or the line between the two nearest troughs on either side of the peak The trigonometric function of the angle.
  • the clutter canceling unit 63 is configured to: determine all peaks and troughs of the target spectrum; determine a lobe width in the target spectrum according to the determined peak and the trough; when the lobe If the width is smaller than the first lobe width, determining that the peak corresponding to the lobe width is a peak of the clutter; being a preset fixed value, or the first lobe width is an average lobe width, wherein the average wave
  • the flap width is the ratio of the width of the target spectrum to the number of second peaks.
  • the clutter cancellation unit 63 is configured to: determine all peaks and troughs of the target spectrum; determine the inclination of the peak according to the determined peak and the trough; when the inclination of the peak If the value is greater than the first threshold, the peak is determined to be a clutter peak; the clutter of the target spectrum is eliminated by erasing the lobe of the clutter, for example, the lobe of the clutter is located in two troughs The frequency amplitude between the two is replaced by an interpolation between the two trough amplitudes.
  • the first threshold may be a preset fixed value.
  • the embodiment of the present invention determines the target spectrum of the harmonic analysis by filtering, and can effectively eliminate the clutter interference, thereby obtaining a clear and accurate harmonic spectrum, thereby facilitating harmonic analysis.
  • FIG. 7 illustrates an apparatus for determining interharmonic clutter of a spectrum in accordance with an embodiment of the present invention.
  • the means 70 for determining the interharmonic clutter of the spectrum includes a first determining unit 71, a second determining unit 72 and a third determining unit 73.
  • the first determining unit 71 is configured to determine all peaks and troughs of the spectrum.
  • the second determining unit 72 is configured to determine a lobe width in the frequency spectrum according to the peak and the trough determined by the first determining unit 71, wherein the lobe width is two nearest neighbors on both sides of the peak The width between the troughs.
  • the third determining unit 73 is configured to determine that the peak corresponding to the lobe width determined by the second determining unit 72 is a peak of the clutter when the lobe width is smaller than the first lobe width.
  • the first lobe width may be a preset fixed value.
  • the first lobe width may be an average lobe width, wherein the average lobe width is a ratio of the width of the spectrum to the number of all peaks.
  • the embodiment of the present invention can determine the clutter between the harmonics by comparing the width of the lobe with the preset value, so as to obtain a clear harmonic structure by eliminating the clutter.
  • Figure 8 illustrates an apparatus for determining interharmonic clutter of a spectrum in accordance with an embodiment of the present invention.
  • the means 80 for determining inter-harmonic clutter of the spectrum includes a fourth determining unit 81, a fifth determining unit 82, and a sixth determining unit 83.
  • the fourth determining unit 81 is configured to determine all peaks and troughs of the target spectrum.
  • the fifth determining unit 82 is configured to determine the inclination of the peak according to the peak and the trough determined by the fourth determining unit 81.
  • the sixth determining unit 83 is configured to determine that the peak determined by the fifth determining unit 82 is a peak of a clutter when the inclination of the peak is greater than the first threshold.
  • the inclination of the peak is an angle between a line between two nearest neighboring troughs on both sides of the peak and a horizontal direction; or the inclination of the peak is the two most on both sides of the peak
  • the slope of the line between adjacent troughs: or the slope of the peak is a trigonometric function of the angle between the lines between the two nearest troughs on either side of the peak.
  • the first threshold can be a preset fixed value.
  • the clutter between the harmonics is fixed so that the harmonic structure can be obtained by eliminating the clutter.
  • the combination of the two can be used to determine the clutter and eliminate the clutter by means of a device 70 that determines the interharmonics of the harmonics of the spectrum, and then further determine the clutter and eliminate it by means 80 of determining the interharmonics of the harmonics of the spectrum. Clutter, which can eliminate clutter more effectively. It is also possible to integrate the means 70 for determining the inter-harmonic clutter of the spectrum with the means 80 for determining the inter-harmonic clutter of the spectrum to form a A device that determines the interharmonics of harmonics in the spectrum. Thus, the first determining unit 71 has the same function as the fourth determining unit 81, and one can be removed. Thus, the integrated means for determining the interharmonic clutter of the spectrum may include a first determining unit 71, a second determining unit 72, a third determining unit 73, a fifth determining unit 82, and a sixth determining unit 83.
  • Fig. 9 is a block diagram showing the structure of an encoder 90 according to an embodiment of the present invention.
  • the encoder 90 includes a processor 91 and a memory 92.
  • the processor 91 implements a method of harmonic analysis according to an embodiment of the present invention. That is, the processor 91 is configured to perform time-frequency conversion on the input time domain signal to obtain a frequency spectrum of the time domain signal in a frequency domain; filtering the spectrum to determine a target spectrum; and using the target spectrum The interharmonic clutter is subjected to a cancellation process to determine the structure of the harmonics of the target spectrum; the pin performs harmonic analysis on the harmonics using the structure of the harmonics of the target spectrum.
  • the memory 92 is used to store instructions executed by the processor 91.
  • the disclosed systems, devices, and methods may be implemented in other ways.
  • the device embodiments described above are merely illustrative.
  • the division of the unit is only a logical function division.
  • there may be another division manner for example, multiple units or components may be combined or Can be integrated into another system, or some features can be ignored, or not executed.
  • the mutual coupling or direct coupling or communication connection shown or discussed may be indirect coupling through some interfaces, devices or units.
  • a communication connection which can be electrical, mechanical or other form.
  • the components displayed for the unit may or may not be physical units, ie may be located in one place, or may be distributed over multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of the embodiment.
  • each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.
  • the functions, if implemented in the form of software functional units and sold or used as separate products, may be stored in a computer readable storage medium.
  • the technical solution of the present invention which is essential or contributes to the prior art, or a part of the technical solution, may be embodied in the form of a software product, which is stored in a storage medium, including
  • the instructions are used to cause a computer device (which may be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention.
  • the foregoing storage medium includes: a U disk, a removable hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and the like, which can store program codes. .

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Signal Processing (AREA)
  • General Physics & Mathematics (AREA)
  • Computational Linguistics (AREA)
  • Quality & Reliability (AREA)
  • Mathematical Physics (AREA)
  • Health & Medical Sciences (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Human Computer Interaction (AREA)
  • Acoustics & Sound (AREA)
  • Multimedia (AREA)
  • Radar Systems Or Details Thereof (AREA)

Abstract

一种谐波分析方法和装置(60)以及一种确定谐波间杂波的方法和装置(70)。其中,谐波分析方法包括:对输入的时域信号进行时频转换,以获得所述时域信号在频域上的频谱(11);对所述频谱进行滤波处理,以确定目标频谱(12);对所述目标频谱的谐波间杂波做消除处理,以便确定所述目标频谱的谐波的结构(13);利用所述目标频谱的谐波的结构对所述谐波进行谐波分析(14)。通过滤波确定谐波分析的目标频谱,并且能够有效地消除杂波干扰,从而获得清晰和准确的谐波频谱,以利于进行谐波分析。

Description

谐波分析方法和装置以及确定谐波间杂波的方法和装置 技术领域
本发明涉及信号处理, 特别地, 涉及谐波分析方法和装置以及确定谐波 间杂波的方法和装置。 背景技术
谐波分析是一种信号处理与分析的基本手段, 在信号识别、分类、编码、 增强等领域都有着广泛的应用。 谐波分析可以涵盖多种不同的分析目的, 例 如计算谐波的个数、 计算谐波的基频、 计算谐波的强弱、 计算谐波的帧内 / 帧间相关性、 计算谐噪比等等。 谐波分析可以基于时域信号进行分析, 也可 以基于频域信号进行分析。 一般来说, 谐波分析在频域信号上进行最常见, 效果也更好。
在现有的音频编码算法中, 一种常用的方法是在频域上釆用峰均比的方 法来确定谐波。 但是, 如果音频帧的频谱是 "模糊" 的, 即谐波间有较强的 杂波存在, 于是谐波的能量被削弱了, 此时含有谐波的子带的峰均比与不含 谐波的子带的峰均比的区别就会不那么明显, 用简单的将峰均比与门限做比 较的方法判断谐波是否存在时就会引入较大的误差。
另一种常用的方法是根据波峰、 波谷的相对能量(或幅度) 关系得到波 峰的强度, 最后通过将这个强度与一个门限做比较来确定是否存在谐波。 但 是, 如果音频帧的频谱是 "模糊" 的, 即谐波间有较强的杂波存在, 由于谐 波间杂波的存在,而可能把谐波附近的杂波的波谷错误的当作是谐波的波谷, 从而导致计算出的谐波强度低于真实值而被漏掉。
由此可见,在进行频域的谐波分析时,无论谐波分析的具体目的是什么, 一个前提条件是要能得到一个尽量准确的谐波结构 , 或者说一个尽量清晰的 谐波频语。 然而, 实际的时频变换所得到的谐波频谱通常都会包含各种干扰 因素, 使得得到的谐波频谱被 "模糊" 了。 这种 "模糊" 可能使原本较强的 谐波被削弱, 也可能在谐波间引入杂波干扰。 在谐波分析时, 这些 "模糊" 可能会导致错误的或性能下降的分析结果。 因此, 如何得到一个尽可能 "清 晰" 的谐波频谱, 是决定谐波分析成败的关键因素。 发明内容
本发明提出了谐波分析方法和装置以及确定谐波间杂波的方法和装置, 旨在解决如何获得清晰的和准确的谐波频谱的问题。
第一方面, 提出了一种谐波分析的方法, 包括: 对输入的时域信号进行 时频转换, 以获得所述时域信号在频域上的频谱;对所述频谱进行滤波处理, 以确定目标频谱; 对所述目标频谱的谐波间杂波做消除处理, 以便确定所述 目标频谱的谐波的结构; 利用所述目标频谱的谐波的结构对所述谐波进行谐 波分析。
结合第一方面, 在第一方面的第一实施方式中, 所述对所述频谱进行滤 波处理, 以确定目标频谱, 包括: 对所述频谱进行 k次滤波处理, 并在所述 频谱经过所述 k次滤波处理的每次滤波处理之后记录波峰个数以获得 k个波 峰个数, 其中 k为大于或等于 2的正整数; 确定所述 k个波峰个数中的最小 波峰个数; 确定得到所述最小波峰个数的滤波处理的频谱为目标频谱。
结合第一方面, 在第一方面的第二实施方式中, 所述对所述频谱进行滤 波处理, 以确定目标频谱, 包括: 对所述频谱进行第一滤波处理, 并记录经 过所述第一滤波处理的频谱的第一波峰个数;对所述频谱进行第二滤波处理, 并记录经过所述第二滤波处理的频谱的第二波峰个数; 对所述频谱进行第三 滤波处理, 并记录经过所述第三滤波处理的频谱的第三波峰个数; 当所述第 二波峰个数小于所述第一波峰个数, 且所述第二波峰个数小于所述第三波峰 个数, 则确定经过所述第二滤波处理得到的频谱为目标频谱。
结合第一方面的第二实施方式, 在第一方面的第三实施方式中, 所述第 一滤波处理为第一阶滤波、 所述第二滤波处理为第二阶滤波, 以及所述第三 滤波处理为第三阶滤波, 所述对所述频谱进行滤波处理, 以确定目标频谱, 包括: 对所述频谱进行第一阶滤波, 并记录经过所述第一阶滤波的频谱的第 一波峰个数; 对所述频谱进行第二阶滤波, 并记录经过所述第二阶滤波的频 谱的第二波峰个数; 对所述频谱进行第三阶滤波, 并记录经过所述第三阶滤 波的频谱的第三波峰个数; 当所述第二波峰个数小于所述第一波峰个数, 且 所述第二波峰个数小于所述第三波峰个数, 则确定经过所述第二阶滤波的频 谱为目标频谱。
结合第一方面, 在第一方面的第四实施方式中, 所述对所述频谱进行滤 波处理, 以确定目标频谱, 包括: 釆用低通滤波器对所述频谱进行低通滤波 处理, 将所述经过所述低通滤波处理后的频语作为目标频谱。
结合第一方面, 在第一方面的第五实施方式中, 所述对所述频谱进行滤 波处理, 以确定目标频谱, 包括: 对所述频谱进行第四阶滤波; 确定经过所 述第四阶滤波得到的频谱为目标频谱, 其中所述第四阶为固定值。
结合第一方面或第一至第五实施方式, 在第一方面的第六实施方式中, 所述对所述目标频谱的谐波间杂波做消除处理, 包括釆用以下方式中的至少 一种: 依据波瓣宽度对所述目标频谱的谐波间杂波做消除处理, 其中所述波 瓣宽度是波峰两侧的两个最邻近的波谷之间的宽度; 依据波峰的倾斜度对所 述目标频谱的谐波间杂波做消除处理, 其中所述波峰的倾斜度是所述波峰两 侧的两个最邻近的波谷之间的连线与水平方向的夹角, 或者所述波峰两侧的 两个最邻近的波谷之间的连线的斜率, 或者所述波峰两侧的两个最邻近的波 谷之间的连线的夹角的三角函数。
结合第一方面的第六实施方式, 在第一方面的第七实施方式中, 所述依 据波瓣宽度对所述目标频谱的谐波间杂波做消除处理, 包括: 确定所述目标 频谱的全部波峰和波谷; 根据确定的所述波峰和所述波谷, 确定所述目标频 谱中的波瓣宽度; 当所述波瓣宽度小于第一波瓣宽度, 则确定所述波瓣宽度 对应的波峰为杂波的波峰; 通过消去所述杂波的波瓣来消除所述目标频谱的 杂波。
结合第一方面的第七实施方式, 在第一方面的第八实施方式中, 所述第 一波瓣宽度为预设的固定值, 或者所述第一波瓣宽度为平均波瓣宽度, 其中 所述平均波瓣宽度是所述目标频谱的宽度与所述第二波峰个数的比值。
结合第一方面的第六实施方式, 在第一方面的第九实施方式中, 所述依 据波峰的倾斜度消除所述目标频谱的杂波, 包括: 确定所述目标频谱的全部 波峰和波谷; 根据确定的所述波峰和所述波谷, 确定所述波峰的倾斜度; 当 所述波峰的倾斜度大于第一阔值, 则确定所述波峰为杂波波峰; 通过消去所 述杂波的波瓣来消除所述目标频谱的杂波。 结合第一方面的第九实施方式, 在第一方面的第十实施方式中, 所述第 一阔值为预设的固定值。
结合第一方面的第七实施方式或第九实施方式, 在第一方面的第十一实 将所述杂波的波瓣位于两个波谷之间的频点幅值以在两个波谷幅值之间的内 插来代替。
第二方面, 提出了一种确定谐波间杂波的方法, 包括: 确定所述频谱的 全部波峰和波谷; 根据确定的所述波峰和所述波谷, 确定所述频谱中的波瓣 宽度, 其中所述波瓣宽度是波峰两侧的两个最邻近的波谷之间的宽度; 当所 述波瓣宽度小于第一波瓣宽度, 则确定所述波瓣宽度对应的波峰为杂波的波 峰。
结合第二方面, 在第二方面的第一实施方式中, 所述第一波瓣宽度为预 设的固定值。
结合第二方面, 在第二方面的第二实施方式中, 所述第一波瓣宽度为平 均波瓣宽度, 其中所述平均波瓣宽度是所述频谱的宽度与所述全部波峰的个 数的比值。
第三方面, 提出了一种确定谐波间杂波的方法, 包括: 确定所述目标频 谱的全部波峰和波谷; 根据确定的所述波峰和所述波谷, 确定所述波峰的倾 斜度; 当所述波峰的倾斜度大于第一阔值, 则确定所述波峰为杂波的波峰。
结合第三方面, 在第三方面的第一实施方式中, 所述波峰的倾斜度是所 述波峰两侧的两个最邻近的波谷之间的连线与水平方向的夹角; 或者所述波 峰的倾斜度是所述波峰两侧的两个最邻近的波谷之间的连线的斜率: 或者所 述波峰的倾斜度是所述波峰两侧的两个最邻近的波谷之间的连线的夹角的三 角函数。
结合第三方面或第三方面的第一实施方式, 在第三方面的第二实施方式 中, 所述第一阔值为预设的固定值。
第四方面, 提出了一种谐波分析的装置, 包括: 时频转换单元, 用于对 输入的时域信号进行时频转换, 以获得所述时域信号在频域上的频谱; 滤波 单元, 用于对所述时频转换单元转换得到的所述频谱进行滤波处理, 以确定 目标频谱; 杂波消除单元, 用于对所述滤波单元确定的所述目标频谱的谐波 间杂波做消除处理, 以便确定所述目标频谱的谐波的结构; 谐波分析单元, 用于利用所述目标频谱的谐波的结构对由所述杂波消除单元消除了杂波的所 述谐波进行谐波分析。
结合第四方面,在第四方面的第一实施方式中,所述滤波单元具体用于: 对所述频谱进行 k次滤波处理, 并在所述频谱经过所述 k次滤波处理的每次 滤波处理之后记录波峰个数以获得 k个波峰个数, 其中 k为大于或等于 2的 正整数; 确定所述 k个波峰个数中的最小波峰个数; 确定得到所述最小波峰 个数的滤波处理的频谱为目标频谱。
结合第四方面,在第四方面的第二实施方式中,所述滤波单元具体用于: 对所述频谱进行第一滤波处理, 并记录经过所述第一滤波处理的频谱的第一 波峰个数; 对所述频谱进行第二滤波处理, 并记录经过所述第二滤波处理的 频谱的第二波峰个数; 对所述频谱进行第三滤波处理, 并记录经过所述第三 滤波处理的频谱的第三波峰个数; 当所述第二波峰个数小于所述第一波峰个 数, 且所述第二波峰个数小于所述第三波峰个数, 则确定经过所述第二滤波 处理得到的频谱为目标频谱。
结合第四方面的第二实施方式, 在第四方面的第三实施方式中, 所述第 一滤波处理为第一阶滤波、 所述第二滤波处理为第二阶滤波, 以及所述第三 滤波处理为第三阶滤波, 所述滤波单元具体用于: 对所述频谱进行第一阶滤 波, 并记录经过所述第一阶滤波的频谱的第一波峰个数; 对所述频谱进行第 二阶滤波, 并记录经过所述第二阶滤波的频谱的第二波峰个数; 对所述频谱 进行第三阶滤波, 并记录经过所述第三阶滤波的频谱的第三波峰个数; 当所 述第二波峰个数小于所述第一波峰个数, 且所述第二波峰个数小于所述第三 波峰个数, 则确定经过所述第二阶滤波的频谱为目标频谱。
结合第四方面,在第四方面的第四实施方式中,所述滤波单元具体用于: 釆用低通滤波器对所述频谱进行低通滤波处理, 将所述经过所述低通滤波处 理后的频语作为目标频谱。
结合第四方面,在第四方面的第五实施方式中,所述滤波单元具体用于: 对所述频谱进行第四阶滤波;确定经过所述第四阶滤波到的频谱为目标频谱; 其中所述第四阶为固定值。
结合第四方面或第一至第五实施方式, 在第四方面的第六实施方式中, 所述杂波消除单元釆用以下方式中的至少一种: 依据波瓣宽度对所述目标频 谱的谐波间杂波做消除处理, 其中所述波瓣宽度是波峰两侧的两个最邻近的 波谷之间的宽度; 依据波峰的倾斜度对所述目标频谱的谐波间杂波做消除处 理, 其中所述波峰的倾斜度是所述波峰两侧的两个最邻近的波谷之间的连线 与水平方向的夹角, 或者所述波峰两侧的两个最邻近的波谷之间的连线的斜 率, 或者所述波峰两侧的两个最邻近的波谷之间的连线的夹角的三角函数。
结合第四方面的第六实施方式, 在第四方面的第七实施方式中, 所述杂 波消除单元具体用于: 确定所述目标频谱的全部波峰和波谷; 根据确定的所 述波峰和所述波谷, 确定所述目标频谱中的波瓣宽度; 当所述波瓣宽度小于 第一波瓣宽度, 则确定所述波瓣宽度对应的波峰为杂波的波峰; 通过消去所 述杂波的波瓣来消除所述目标频谱的杂波。
结合第四方面的第七实施方式, 在第四方面的第八实施方式中, 所述第 一波瓣宽度为预设的固定值, 或者所述第一波瓣宽度为平均波瓣宽度, 其中 所述平均波瓣宽度是所述目标频谱的宽度与所述第二波峰个数的比值。
结合第四方面的第六实施方式, 在第四方面的第九实施方式中, 所述杂 波消除单元具体用于: 确定所述目标频谱的全部波峰和波谷; 根据确定的所 述波峰和所述波谷, 确定所述波峰的倾斜度; 当所述波峰的倾斜度大于第一 阔值, 则确定所述波峰为杂波波峰; 通过消去所述杂波的波瓣来消除所述目 标频谱的杂波。
结合第四方面的第九实施方式, 在第四方面的第十实施方式中, 所述第 一阔值为预设的固定值。
结合第四方面的第七或第九实施方式,在第四方面的第十一实施方式中, 所述杂波消除单元具体用于: 将所述杂波的波瓣位于两个波谷之间的频点幅 值以在两个波谷幅值之间的内插来代替。
第五方面, 提出了一种确定谐波间杂波的装置, 包括: 第一确定单元, 用于确定所述频谱的全部波峰和波谷; 第二确定单元, 用于根据所述第一确 定单元确定的所述波峰和所述波谷, 确定所述频谱中的波瓣宽度, 其中所述 波瓣宽度是波峰两侧的两个最邻近的波谷之间的宽度; 第三确定单元, 用于 当所述波瓣宽度小于第一波瓣宽度, 则确定由所述第二确定单元确定的所述 波瓣宽度对应的波峰为杂波的波峰。
结合第五方面, 在第五方面的第一实施方式中, 所述第一波瓣宽度为预 设的固定值。
结合第五方面, 在第五方面的第一实施方式中, 所述第一波瓣宽度为平 均波瓣宽度, 其中所述平均波瓣宽度是所述频谱的宽度与所述全部波峰的个 数的比值。
第六方面, 提出了一种确定谐波间杂波的装置, 包括: 第四确定单元, 用于确定所述目标频谱的全部波峰和波谷; 第五确定单元, 用于根据所述第 四确定单元确定的所述波峰和所述波谷, 确定所述波峰的倾斜度; 第六确定 单元, 用于当所述波峰的倾斜度大于第一阔值, 则确定由所述第五确定单元 确定的所述波峰为杂波的波峰。
结合第六方面, 在第六方面的第一实施方式中, 所述波峰的倾斜度是所 述波峰两侧的两个最邻近的波谷之间的连线与水平方向的夹角; 或者所述波 峰的倾斜度是所述波峰两侧的两个最邻近的波谷之间的连线的斜率: 或者所 述波峰的倾斜度是所述波峰两侧的两个最邻近的波谷之间的连线的夹角的三 角函数。
结合第六方面或第六方面的第一实施方式, 在第四方面的第二实施方式 中, 所述第一阔值为预设的固定值。
本发明实施例通过滤波确定谐波分析的目标频谱, 并且能够有效地消除 杂波干扰, 从而获得清晰的和准确的谐波频谱, 以利于进行谐波分析。 附图说明
为了更清楚地说明本发明实施例的技术方案, 下面将对本发明实施例中 所需要使用的附图作简单地介绍, 显而易见地, 下面所描述的附图仅仅是本 发明的一些实施例, 对于本领域普通技术人员来讲, 在不付出创造性劳动的 前提下, 还可以根据这些附图获得其他的附图。
图 1是根据本发明实施例的谐波分析的方法的流程图。 图 2是消除杂波的示意图。
图 3示出了通过波峰的倾斜度确定杂波的方法。
图 4是才艮据本发明实施例的确定谐波间杂波的方法的流程图。
图 5是根据本发明另一实施例的确定谐波间杂波的方法的流程图。
图 6是根据本发明实施例的谐波分析的装置的结构示意图。
图 7是才艮据本发明实施例的确定谐波间杂波的装置的结构示意图。
图 8是才艮据本发明另一实施例的确定谐波间杂波的装置的结构示意图。 图 9是根据本发明实施例的编码器的结构示意图。 具体实施方式
下面将结合本发明实施例中的附图, 对本发明实施例中的技术方案进行 清楚、 完整地描述, 显然, 所描述的实施例是本发明的一部分实施例, 而不 是全部实施例。 基于本发明中的实施例, 本领域普通技术人员在没有做出创 造性劳动的前提下所获得的所有其他实施例, 都应属于本发明保护的范围。 以下将结合图 1 , 描述根据本发明实施例的谐波分析的方法, 包括如下步骤。
11 , 对输入的时域信号进行时频转换, 以获得所述时域信号在频域上的 频谱。
例如, 快速傅里叶变换(FFT, Fast Fourier Transform )和离散余弦变换 ( DCT, Discrete Cosine Transform )等是常用的时频转换方法, 通过上述方 法对输入的时域信号进行时频转换, 以获得所述输入的时域信号在频域上的 频谱。
12, 对所述频谱进行滤波处理, 以确定目标频谱。
在一种实施方式中, 对所述频语进行 k次滤波处理, 并在所述频语经过 所述 k次滤波处理的每次滤波处理之后记录波峰个数以获得 k个波峰个数, 其中 k为大于或等于 2的正整数;确定所述 k个波峰个数中的最小波峰个数; 确定得到所述最小波峰个数的滤波处理的频谱为目标频谱。
在另一种实现方式中, 对所述频谱进行第一滤波处理, 并记录经过所述 第一滤波处理的频谱的第一波峰个数; 对所述频谱进行第二滤波处理, 并记 录经过所述第二滤波处理的频谱的第二波峰个数; 对所述频谱进行第三滤波 处理, 并记录经过所述第三滤波处理的频谱的第三波峰个数; 当所述第二波 峰个数小于所述第一波峰个数,且所述第二波峰个数小于所述第三波峰个数, 则确定经过所述第二滤波处理的频谱为目标频谱。
具体而言, 第一滤波处理可以为第一阶滤波、 第二滤波处理可以为第二 阶滤波, 以及第三滤波处理可以为第三阶滤波。 应理解, 第一阶滤波、 第二 阶滤波和第三阶滤波分别对应的滤波阶数是不同的。 这样, 对所述频谱进行 第一阶滤波, 并记录经过所述第一阶滤波的频谱的第一波峰个数; 对所述频 谱进行第二阶滤波, 并记录经过所述第二阶滤波的频谱的第二波峰个数; 对 所述频谱进行第三阶滤波, 并记录经过所述第三阶滤波的频谱的第三波峰个 数; 当所述第二波峰个数小于所述第一波峰个数, 且所述第二波峰个数小于 所述第三波峰个数, 则确定经过所述第二阶滤波的频谱为目标频谱。
通常, 当第一阶滤波、 第二阶滤波和第三阶滤波均为低通滤波时, 更优 的滤波应该导致目标频谱具有更少的频谱波峰个数。 可以理解, 上述滤波过 程是可以循环的。 例如, 在经过第一次滤波组合中的第一阶滤波、 第二阶滤 波和第三阶滤波之后, 若第二波峰个数并不同时小于第一波峰个数和第三波 峰个数, 那么就再次经过第二次滤波组合中的第一阶滤波、 第二阶滤波和第 三阶滤波, 直到再某次滤波组合中寻找到同时小于第一波峰个数和第三波峰 个数的第二波峰个数。 也就是说, 这样的循环滤波寻找最小频谱波峰个数的 过程, 就是一个寻找最优滤波、 确定目标频谱的过程。
这样, 随着滤波的阶数由低向高增加, 每轮滤波后的频谱波峰个数会呈 现出一种先单调递减, 再单调递增的现象。 通常, 最优滤波就出现在波峰个 数折线由递减到递增的拐点处。 为了降低计算复杂度, 本实施方式中对确定 目标频谱的循环过程增加了一个提前终止判断, 即一旦检测到拐点, 则认为 已找到最优滤波, 不再执行剩下的循环。 本实施方式中每循环一次, 比较当 次循环后的目标频谱的波峰个数是否小于等于前次循环后的目标频谱的波峰 个数。 若是, 则继续循环, 若否, 则终止循环, 以前次循环后的目标频语作 为后续步骤的目标频谱, 即将前次循环的滤波作为最优滤波。
在另一种实施方式中, 为了进一步降低计算的复杂度, 滤波阶数不再动 态的确定, 而是对每帧都以一个固定的阶数进行滤波。 经过对大的样本进行 统计可以发现, 对一给定的釆样率和帧长, 最优滤波阶数在某一固定阶数的 概率会显著高于其它阶数。 因此以该最大出现概率阶数作为固定的滤波阶数 可以在节省计算复杂度的同时最大限度的逼近最优滤波结果。 例如, 对所述 频谱进行第四滤波处理;确定经过所述第四滤波处理得到的频谱为目标频谱。 具体地, 第四滤波处理为第四阶滤波, 则对所述频谱进行第四阶滤波, 其中 所述第四阶为固定值; 确定经过所述第四阶滤波的频谱为目标频谱。
在另一种实施方式中, 可以釆用任意形式的低通滤波器对所述频谱进行 低通滤波处理, 以确定目标频谱为经过所述低通滤波处理后的频谱。
13 , 对所述目标频谱的谐波间杂波做消除处理, 以便确定所述目标频谱 的谐波的结构。
例如, 可以釆用以下方式中的至少一种来对所述目标频谱的谐波间杂波 做消除处理: 依据波瓣宽度对所述目标频谱的谐波间杂波做消除处理; 依据 波峰的倾斜度对所述目标频谱的谐波间杂波做消除处理。
依据波瓣宽度对所述目标频谱的谐波间杂波做消除处理的方法包括如下 步骤: 首先, 确定所述目标频谱的全部波峰和波谷, 其中频谱中某一频点的 能量高于它的两个最邻近频点的能量时,则认为该频点处是频谱的一个波峰, 而波谷的定义为两个相邻波峰间的能量最小的频点; 然后, 根据确定的所述 波峰和所述波谷, 确定所述目标频谱中的波瓣宽度, 其中所述波瓣是一个波 峰两侧的两个最邻近波谷之间的频点所组成的集合, 所述波瓣宽度是波峰两 侧的两个最邻近的波谷之间的宽度; 接着, 当所述波瓣宽度小于第一波瓣宽 度, 则确定所述波瓣宽度对应的波峰为杂波的波峰; 最后, 通过消去所述杂 波的波瓣来消除所述目标频谱的杂波, 例如将所述杂波的波瓣位于两个波谷 之间的频点能量以在两个波谷之间的内插来代替。
这里, 第一波瓣宽度为预设的固定值, 或者所述第一波瓣宽度为平均波 瓣宽度, 其中所述平均波瓣宽度是所述目标频谱的宽度与所述第二波峰个数 的比值。 确定所述目标频谱的全部波峰和波谷; 然后, 根据确定的所述波峰和所述波 谷, 确定所述波峰的倾斜度; 接着, 当所述波峰的倾斜度大于第一阔值, 则 确定所述波峰为杂波波峰; 最后, 通过消去所述杂波的波瓣来消除所述目标 频谱的杂波, 例如将所述杂波的波瓣位于两个波谷之间的频点能量以在两个 波谷之间的内插来代替。
这里所述的波峰的倾斜度可以是所述波峰的倾斜度是所述波峰两侧的两 个最邻近的波谷之间的连线与水平方向的夹角; 也可以是所述波峰两侧的两 个最邻近的波谷之间的连线的斜率; 也可以是所述波峰两侧的两个最邻近的 波谷之间的连线的夹角的三角函数。 第一阔值可以为预设的固定值, 也可以 与所述目标频谱的滤波阶数相关。
14, 针利用所述目标频谱的谐波的结构对所述谐波进行谐波分析。
例如, 计算谐波的个数, 计算谐波的基频, 计算谐波的强弱, 计算谐波 的帧内 /帧间相关性, 计算谐噪比等等。
由此可见, 本发明实施例通过滤波确定谐波分析的目标频谱, 并且能够 有效地消除杂波干扰, 从而获得清晰的和准确的谐波频谱, 以利于进行谐波 分析。
下面将结合具体实施例, 详细说明根据本发明实施例的谐波分析的方法。
第一具体实施例, 以音频编码器为例。 一般地, 音频领域常见的时频域 变换包括 FFT, DCT, 改进型的离散余弦变换(MDCT, Modified Discrete Cosine Transform )等。
由于 MDCT是广泛应用于音频编码器的时频域变换工具, 复用 MDCT 系数可以节省谐波分析的计算复杂度, 因此本具体实施例以 MDCT 变换为 例, 说明一个计算 32kHz釆样音频信号在 8k-16kHz频段上的谐波个数的谐 波分析方案, 输入音频帧为 20ms帧长。
第一步, 首先对输入编码器的时域音频帧 s(x)进行 MDCT变换, 得到输 入的时域音频帧的 MDCT系数 M(x), 其中 x=0,l, ...639。 通常, 时域音频帧 在进行 MDCT变换前都要先进行加窗操作,由于是公知技术,这里不再详述, 默认 s(x)是加窗后的时域音频帧。
容易理解, 时域音频帧 s(x)的釆样个数取决于该时域音频帧的帧长和釆 样率。 在本具体实施例中, 根据 20ms帧长和 32Hz釆样率, 可以确定时域音 频帧 s(x)的釆样个数为 640。 同样容易理解, 时频变换后的频域音频帧的样 点个数取决于时域帧釆样个数和时频变换方法。 在本具体实施例中, 根据
MDCT的变换方法, 可以确定频域音频帧 M(x)的样点个数也为 640。
然后, 根据输入的时域音频帧的 MDCT系数 M(x)得到输入的时域音频 帧在 8k-16kHz频段上的对数能量谱 S(i), 其中 i=0,l,...319。 为了方便后续描 述, 简称为频谱 S(i), 频谱 S(i)可以根据以下公式 1进行确定:
S(i) = log2 (M(i + 320) -M(i + 320)) 公式(1) 其中, i=0, 1, ...319。
第二步,对频谱 S (i)做 3阶滤波,得到滤波后的频谱 S[k](j),这里 j=0, 1, ... 319, 其中 S[k )表示 k阶滤波后的频谱。 具体的滤波方法为以下公式(2)至 公式(4) 中之一或其组合:
∑ S(i) , j<(k-l)/2 公式( 2 ) j + (k + l)/2
M j+{klc--\[))I/22
1 X S(i) , (k-l)/2< j<319-(k-l)/2 公式( 3 ) i-3-(k-\)/2
319
zrjz ∑ ), j>319-(k-l)/2 公式( 4 ) 其中 k为滤波器的阶数, 例如 k=3。 实际得到的谐波频谱由于多种因素 的干扰可能是 "模糊" 的, 也就是, 除谐波外, 频谱上还会有很多干扰杂波。 一般而言, 相对于谐波, 干扰杂波可以被看作是一种更高频的成份。 通过滤 波, 可以有效地将部分杂波滤掉, 而保留住谐波成份。
第三步, 求频谱 S[3](j)中波峰的个数 p[3]。 当频谱中某一频点的能量高 于它的两个最邻近频点的能量时, 则认为该频点处是频谱的一个波峰。 此步 执行完后, 记录下求得的波峰个数1[3], 这里说的记录可以由软件实现, 例如 读取后存储在内容中, 也可以由硬件实现, 比如存储在寄存器中。 返回第二 步, 对频谱 S(i)再做 5阶滤波, 对 S[5](j)再记录下其波峰个数 p[5], 以此循 环往复再记录下其余的 Pm、 P[9]等。
第四步, 在所有记录下的 p[k]中找到最小值 ρ[τ], 获取其对应的滤波后的 频谱, 即 S[T](j), 下面称作目标频谱。 目标频谱 S[T](j)的获取可以是从之前 在求 P [kI的循环中记录下的频谱中直接获取,也可以是根据 T值重新 由于每帧谐波频谱的特征不同, 对每帧使用相同的滤波方法不会在每帧都达 到最优的效果。 一般而言, 更优的滤波应该导致目标频谱具有更少的频谱波 峰个数。 因此, 上述循环滤波寻找最小频谱波峰个数的过程, 就是一个寻找 最优滤波、 确定目标频谱的过程。
第五步, 搜索目标频谱 S[T] (j)的所有波峰和波谷, 记录下每个波峰和波 谷的位置和能量。 其中波谷的定义为两个相邻波峰间的能量最小的频点。 特 峰与它所在端点之间的能量最小的频点。 通常, 波峰和波谷总是交替出现的。 波峰的能量和位置分别记作 Ep (m)和 Idxp (m) , 其中 m=0, 1, ... L_l , L为目标频 谱的波峰个数 pm。波谷的能量和位置分别记作 (1 )和 Idxv (n) ,其中 n=0, 1 , ... L。 波谷个数比波峰个数多一个。
第六步, 依据波瓣宽度消除目标频谱的杂波。
通过滤波获取目标频谱的过程已经一定程度上消除了部分杂波, 但一般 而言仍会有杂波未被消除。 此时, 由于杂波的存在, 目标频谱的波峰个数会 大于实际谐波的个数, 从而使目标频谱的平均波瓣宽度会小于谐波波瓣宽度。 这时, 如果目标频谱的某个波峰的波瓣宽度小于平均波瓣宽度, 则该波峰被 认为是杂波。 这里, 平均波瓣宽度的定义为: 目标频谱的带宽与波峰个数的 比值。
搜索出目标频谱 S[T] (j)的所有波峰和波谷后, 再计算每个波瓣宽度 Wp(m), 其中 m=0, 1, ... L-l。 其中, 波瓣的定义为由一个波峰两侧的两个最邻 近波谷之间的频点所组成的集合。 波瓣宽度的定义为该波瓣的两个波谷之间 的宽度, 即波瓣覆盖的频率范围大小。 波瓣宽度可以通过以下公式进行确定:
Wp {m) = Idxv (m + \) - Idxv (m) 公式( 5 ) 被确认的杂波需要被从目标频谱中消去, 具体的消去方法是将杂波的波 瓣消去, 即将杂波波瓣位于两个波谷之间的频点能量以在两个波谷之间的内 插来代替, 参见图 2。 图 1中位于中间的较小波峰即为杂波波峰, 经杂波消去 后, 该波峰即被虚线的频谱所代替。
其数 达为: - S[T] (Idxv (m)) 公式 ( 6 )
Figure imgf000015_0001
其中, 1=0, 1, ...Wp (m) , S[T]'G)表示杂波波瓣被消除后的所在位置的频谱。 为 了方便, 消去杂波后的目标频谱仍称作目标频谱。
第七步, 依据波峰的倾斜度消除目标频谱的杂波。
经第六步后的目标频谱可能仍是包含有杂波的, 还需利用其它方法进一 步消除。 图 3示出了一种典型的杂波频谱。 波峰 A是谐波波峰, 波峰 B和波 峰 C是杂波波峰。 虚线 M是穿过波峰 B的两个波谷的直线, 虚线 N是穿过波 峰 C的两个波谷的直线, 角 α和角 β分别是 M和 N与水平轴的夹角。 可见, 波峰 Β和波峰 C明显倾斜, 具体而言, 角 α和角 β的角度明显较大, 即角 α 和角 β的正切较大, 或直线 Μ和直线 Ν的斜率较大。
通过判断穿过某波峰的两个波谷的直线的斜率, 可以一定程度上确定该 波峰是否为杂波。 具体的, 将求得的斜率与一个门限 thrl做比较, 若大于该 门限, 则认为该波峰为杂波。 门限 thrl可以是一个预先设定的固定值, 也可 以是变量。 当门限是变量时, 可以根据第四步中最优滤波的阶数 T确定, 即 thrl=f (T) , 表示 thrl是滤波器的最优阶数 T的函数。 由于第六步的消除杂 波操作已经改变了目标频谱, 在依据波峰的倾斜度消除目标频谱的杂波之前, 需要重新搜索目标频谱的所有波峰和波谷。 确定出杂波后, 以与第六步中公 式(6 )表示的相同的方法消除, 参照图 2所示的方法将杂波消除。
第八步, 搜索谐波个数。 经过之前的步骤, 目标频谱的杂波已经在很大程度 上消除了, 谐波结构已变得清晰, 可以进行最后的谐波个数确认了。 重新搜 索目标频谱的所有波峰和波谷,计算每个波峰的强度 IP(y),其中 y=0,l,2...K, K为波峰个数。 波峰强度通过以下公式进行确定: , 1) 公式(7 ) 将每个波峰的强度与一个门限 thr2做比较。 若波峰的强度大于该门限, 则认为检测到一个谐波。 与门限 thrl类似, 门限 thr2可以是一个预先设定的 固定值, 也可以是变量。 当门限是变量时, 可以根据第四步中最优滤波的阶 数 T确定, 即 thr2=f(T), 表示 thr2是滤波器的最优阶数 T的函数。
第二具体实施例中, 其与第一具体实施例中的大部分步骤相似, 不同在 于对第一实施例中确定目标频谱的第二至四步的简化处理。 具体而言, 第一具体实施例的第二至四步中的最优滤波器的确定是通过 一个循环过程并取其中的最优滤波器来实现的。 一般而言, 随着滤波的阶数 由低向高增加, 每轮滤波后的频谱波峰个数会呈现出一种先单调递减, 再单 调递增的现象。 通常, 最优滤波就出现在波峰个数折线由递减到递增的拐点 处。 为了降低计算复杂度, 第二具体实施例中对第一具体实施例中的循环过 程增加了一个提前终止判断, 即一旦检测到拐点, 则认为已找到最优滤波, 不再执行剩下的循环。 每循环一次, 比较当次循环后的目标频谱的波峰个数 是否小于等于前次循环后的目标频谱的波峰个数。 若是, 则继续循环, 若否, 则终止循环, 以前次循环后的目标频谱作为后续步骤的目标频谱, 即将前次 循环的滤波作为最优滤波。
第三具体实施例中, 其与第一具体实施例中的大部分步骤相似, 不同在 于对第一实施例中确定目标频谱的第二至四步的简化处理。
具体而言, 滤波阶数不再动态的确定, 而是对每帧都以一个固定的阶数 进行滤波。 经过对大的样本进行统计可以发现, 对一给定的釆样率和帧长, 最优滤波阶数在某一固定阶数的概率会显著高于其它阶数。 因此以该最大出 现概率阶数作为固定的滤波阶数可以在节省计算复杂度的同时最大限度的逼 近最优滤波结果。 以第一具体实施例中 32kHz釆样, 20ms帧长的输入信号 为例, 釆用的固定滤波阶数选为 5。
第四具体实施例中, 其与第一具体实施例中的大部分步骤相似, 不同在 于对第一实施例中确定目标频谱的第二至四步的简化处理。
本发明实施例的输入信号可以不限制为音频信号, 可以是任意一种可以 在频域分析的信号, 如图像信号等。
另外, 对原始频谱滤波获得目标频谱的过程所使用的滤波器可以是任意 形式的低通滤波器。 例如, 以下公式 ( 8 )给出了一阶低通滤波器的一个实施 例。
^]( ) = ^( -1) + ^( 公式(8 ) 其中, i=0,l, ... ,319, /是常系数。
第五具体实施例中, 其与第一具体实施例中的大部分步骤相似, 不同在 于对第一实施例中确定目标频谱的第六步的依据波瓣宽度消除杂波过程中波 瓣宽度的确定方法。
具体而言, 第一具体实施例在第六步依据波瓣宽度消除杂波的步骤中, 杂波的确定是将某个波峰的波瓣宽度与平均波瓣宽度做比较。 然而, 根据特 定的应用, 确定杂波时也可以将波瓣宽度与一个预先设定的固定值做比较。 如要搜索 8000Hz〜 16000Hz频带内的非致密谐波(即在该频带内谐波数少于 10的谐波), 若当谐波数等于 10时,平均波瓣宽度应为 800Hz或 32个频点, 故非致密谐波的波瓣宽度应大于 32。所以可将门限设定为一个固定值 32,若 某波峰的波瓣宽度小于 32, 则认为是杂波。
下面将结合图 4和图 5分别说明根据本发明实施例的确定频谱的谐波间 杂波的方法。
如图 4所示的确定频谱的谐波间杂波的方法包括如下步骤:
41 , 确定频语的全部波峰和波谷;
42, 根据确定的所述波峰和所述波谷, 确定所述频谱中的波瓣宽度, 其 中所述波瓣宽度是波峰两侧的两个最邻近的波谷之间的宽度;
43 , 当所述波瓣宽度小于第一波瓣宽度, 则确定所述波瓣宽度对应的波 峰为杂波的波峰。
其中, 第一波瓣宽度可以为预设的固定值。 或者, 第一波瓣宽度可以为 平均波瓣宽度, 其中所述平均波瓣宽度是所述频谱的宽度与所述全部波峰的 个数的比值。
由此可见, 本发明实施例的确定频谱的谐波间杂波的方法可以有效去除 频谱的谐波间杂波。
如图 5所示的确定频谱的谐波间杂波的方法包括如下步骤:
51 , 确定所述目标频谱的全部波峰和波谷;
52, 根据确定的所述波峰和所述波谷, 确定所述波峰的倾斜度;
53 , 当所述波峰的倾斜度大于第一阔值,则确定所述波峰为杂波的波峰。 其中, 波峰的倾斜度是所述波峰两侧的两个最邻近的波谷之间的连线与 水平方向的夹角; 或者所述波峰的倾斜度是;
其中, 第一阔值可以为预设的固定值。
由此可见, 本发明实施例的确定频谱的谐波间杂波的方法可以有效去除 频谱的谐波间杂波。
此外, 图 4与图 5所示的两种确定频谱的谐波间杂波的方法也可以结合 实施, 这样可以更有效地去除频谱的谐波间杂波。
图 6示出了根据本发明实施例的谐波分析的装置的结构示意图。 其中, 谐波分析的装置 60 包括时频转换单元 61、 滤波单元 62、 杂波消除单元 63 和谐波分析单元 64。 所述时频转换单元 61、 滤波单元 62、 杂波消除单元 63 和谐波分析单元 64 所对应的功能模块集成在数字信号处理(DSP, Digital Signal Processing ) 器件中。
其中, 时频转换单元 61用于对输入的时域信号进行时频转换, 以获得所 述时域信号在频域上的频谱。 滤波单元 62用于对所述时频转换单元 61转换 得到的所述频谱进行滤波处理, 以确定目标频谱。杂波消除单元 63用于对所 述滤波单元 62确定的所述目标频谱的谐波间杂波做消除处理,以便确定所述 目标频谱的谐波的结构。谐波分析单元 64用于利用所述目标频谱的谐波的结 构对由所述杂波消除单元 63消除了杂波的所述谐波进行谐波分析。
进一步, 滤波单元 62具体用于: 对所述频谱进行 k次滤波处理, 并在所 述频谱经过所述 k次滤波处理的每次滤波处理之后记录波峰个数以获得 k个 波峰个数, 其中 k为大于或等于 2的正整数; 确定所述 k个波峰个数中的最 小波峰个数; 确定得到所述最小波峰个数的滤波处理的频谱为目标频谱。
或者, 滤波单元 62用于: 对所述频谱进行第一滤波处理, 并记录经过所 述第一滤波处理的频谱的第一波峰个数; 对所述频谱进行第二滤波处理, 并 记录经过所述第二滤波处理的频谱的第二波峰个数; 对所述频谱进行第三滤 波处理, 并记录经过所述第三滤波处理的频谱的第三波峰个数; 当所述第二 波峰个数小于所述第一波峰个数, 且所述第二波峰个数小于所述第三波峰个 数, 则确定经过所述第二滤波处理得到的频谱为目标频谱。
具体而言, 若所述第一滤波处理为第一阶滤波、 所述第二滤波处理为第 二阶滤波, 以及所述第三滤波处理为第三阶滤波, 滤波单元 62用于: 对所述 频谱进行第一阶滤波, 并记录经过所述第一阶滤波的频谱的第一波峰个数; 对所述频谱进行第二阶滤波, 并记录经过所述第二阶滤波的频谱的第二波峰 个数; 对所述频谱进行第三阶滤波, 并记录经过所述第三阶滤波的频谱的第 三波峰个数; 当所述第二波峰个数小于所述第一波峰个数, 且所述第二波峰 个数小于所述第三波峰个数,则确定经过所述第二阶滤波的频谱为目标频谱。
或者,滤波单元 62用于: 釆用任意形式的低通滤波器对所述频谱进行低 通滤波处理, 将所述经过所述低通滤波处理后的频语作为目标频谱。
或者, 滤波单元 62用于: 对所述频谱进行第四阶滤波, 其中所述第四阶 为固定值; 确定经过所述第四阶滤波的频谱为目标频谱。
进一步地,杂波消除单元 63釆用以下方式中的至少一种:依据波瓣宽度 对所述目标频谱的谐波间杂波做消除处理, 其中所述波瓣宽度是波峰两侧的 两个最邻近的波谷之间的宽度; 依据波峰的倾斜度对所述目标频谱的谐波间 杂波做消除处理, 其中所述波峰的倾斜度是所述波峰两侧的两个最邻近的波 谷之间的连线与水平方向的夹角, 或者所述波峰两侧的两个最邻近的波谷之 间的连线的斜率, 或者所述波峰两侧的两个最邻近的波谷之间的连线的夹角 的三角函数。
具体而言,杂波消除单元 63用于:确定所述目标频谱的全部波峰和波谷; 根据确定的所述波峰和所述波谷, 确定所述目标频谱中的波瓣宽度; 当所述 波瓣宽度小于第一波瓣宽度,则确定所述波瓣宽度对应的波峰为杂波的波峰; 为预设的固定值, 或者所述第一波瓣宽度为平均波瓣宽度, 其中所述平均波 瓣宽度是所述目标频谱的宽度与所述第二波峰个数的比值。
或者, 具体地, 杂波消除单元 63用于: 确定所述目标频谱的全部波峰和 波谷; 根据确定的所述波峰和所述波谷, 确定所述波峰的倾斜度; 当所述波 峰的倾斜度大于第一阔值, 则确定所述波峰为杂波波峰; 通过消去所述杂波 的波瓣来消除所述目标频谱的杂波, 例如, 将所述杂波的波瓣位于两个波谷 之间的频点幅值以在两个波谷幅值之间的内插来代替。 其中, 第一阔值可以 为预设的固定值。
由此可见, 本发明实施例通过滤波确定谐波分析的目标频谱, 并且能够 有效地消除杂波干扰, 从而获得清晰的和准确的谐波频谱, 以利于进行谐波 分析。
图 7示出了根据本发明实施例的确定频谱的谐波间杂波的装置。 如图所 示, 确定频谱的谐波间杂波的装置 70包括第一确定单元 71、 第二确定单元 72和第三确定单元 73。
其中,第一确定单元 71用于确定所述频谱的全部波峰和波谷。第二确定 单元 72用于根据所述第一确定单元 71确定的所述波峰和所述波谷, 确定所 述频谱中的波瓣宽度, 其中所述波瓣宽度是波峰两侧的两个最邻近的波谷之 间的宽度。第三确定单元 73用于当所述波瓣宽度小于第一波瓣宽度,则确定 由所述第二确定单元 72确定的所述波瓣宽度对应的波峰为杂波的波峰。
这里, 第一波瓣宽度可以为预设的固定值。 或者, 第一波瓣宽度可以为 平均波瓣宽度, 其中所述平均波瓣宽度是所述频谱的宽度与所述全部波峰的 个数的比值。
由此可见, 本发明实施例可以通过波瓣宽度与预设值的比较结果确定谐 波间的杂波, 以便再通过消除杂波来获得清楚的谐波结构。
图 8示出了根据本发明实施例的确定频谱的谐波间杂波的装置。 如图所 示, 确定频谱的谐波间杂波的装置 80包括第四确定单元 81、 第五确定单元 82和第六确定单元 83。
其中,第四确定单元 81用于确定所述目标频谱的全部波峰和波谷。第五 确定单元 82用于根据所述第四确定单元 81确定的所述波峰和所述波谷, 确 定所述波峰的倾斜度。第六确定单元 83用于当所述波峰的倾斜度大于第一阔 值, 则确定由所述第五确定单元 82确定的所述波峰为杂波的波峰。
这里, 所述波峰的倾斜度是所述波峰两侧的两个最邻近的波谷之间的连 线与水平方向的夹角; 或者所述波峰的倾斜度是所述波峰两侧的两个最邻近 的波谷之间的连线的斜率: 或者所述波峰的倾斜度是所述波峰两侧的两个最 邻近的波谷之间的连线的夹角的三角函数。 第一阔值可以为预设的固定值。 定谐波间的杂波, 以便再通过消除杂波来获得清楚的谐波结构。
容易理解, 可以将上两者相结合, 即通过确定频谱的谐波间杂波的装置 70确定杂波并消除杂波, 然后再通过确定频谱的谐波间杂波的装置 80进一 步确定杂波并消除杂波, 从而可以更有效地消除杂波。 也可以将确定频谱的 谐波间杂波的装置 70与确定频谱的谐波间杂波的装置 80集成在一起形成一 个确定频谱的谐波间杂波的装置。 这样, 第一确定单元 71 与第四确定单元 81的功能相同, 可以去除一个。 于是, 集成的确定频谱的谐波间杂波的装置 可以包括第一确定单元 71、 第二确定单元 72、 第三确定单元 73、 第五确定 单元 82和第六确定单元 83。
图 9示出了才艮据本发明实施例的编码器 90的结构示意图。其中,该编码 器 90包括处理器 91和存储器 92。
其中, 处理器 91实现根据本发明实施例的谐波分析的方法。 即, 处理器 91用于对输入的时域信号进行时频转换, 以获得所述时域信号在频域上的频 谱; 对所述频谱进行滤波处理, 以确定目标频谱; 对所述目标频谱的谐波间 杂波做消除处理, 以便确定所述目标频谱的谐波的结构; 针利用所述目标频 谱的谐波的结构对所述谐波进行谐波分析。 存储器 92用于存储处理器 91执 行的指令。
应理解, 本发明的每个权利要求所叙述的方案也应看做是一个实施例, 并且是权利要求中的特征是可以结合的, 如本发明中的判断步骤后的执行的 不同分支的步骤可以作为不同的实施例。
本领域普通技术人员可以意识到, 结合本文中所公开的实施例描述的各 示例的单元及算法步骤, 能够以电子硬件、 或者计算机软件和电子硬件的结 合来实现。 这些功能究竟以硬件还是软件方式来执行, 取决于技术方案的特 定应用和设计约束条件。 专业技术人员可以对每个特定的应用来使用不同方 法来实现所描述的功能, 但是这种实现不应认为超出本发明的范围。
所属领域的技术人员可以清楚地了解到, 为描述的方便和简洁, 上述描 述的系统、 装置和单元的具体工作过程, 可以参考前述方法实施例中的对应 过程, 在此不再赘述。
在本申请所提供的几个实施例中, 应该理解到, 所揭露的系统、 装置和 方法, 可以通过其它的方式实现。 例如, 以上所描述的装置实施例仅仅是示 意性的, 例如, 所述单元的划分, 仅仅为一种逻辑功能划分, 实际实现时可 以有另外的划分方式, 例如多个单元或组件可以结合或者可以集成到另一个 系统, 或一些特征可以忽略, 或不执行。 另一点, 所显示或讨论的相互之间 的耦合或直接耦合或通信连接可以是通过一些接口, 装置或单元的间接耦合 或通信连接, 可以是电性, 机械或其它的形式。 为单元显示的部件可以是或者也可以不是物理单元, 即可以位于一个地方, 或者也可以分布到多个网络单元上。 可以根据实际的需要选择其中的部分或 者全部单元来实现本实施例方案的目的。
另外,在本发明各个实施例中的各功能单元可以集成在一个处理单元中, 也可以是各个单元单独物理存在, 也可以两个或两个以上单元集成在一个单 元中。
所述功能如果以软件功能单元的形式实现并作为独立的产品销售或使用 时, 可以存储在一个计算机可读取存储介质中。 基于这样的理解, 本发明的 技术方案本质上或者说对现有技术做出贡献的部分或者该技术方案的部分可 以以软件产品的形式体现出来, 该计算机软件产品存储在一个存储介质中, 包括若干指令用以使得一台计算机设备(可以是个人计算机, 服务器, 或者 网络设备等)执行本发明各个实施例所述方法的全部或部分步骤。 而前述的 存储介质包括: U盘、 移动硬盘、 只读存储器(ROM, Read-Only Memory ), 随机存取存储器 (RAM, Random Access Memory )、 磁碟或者光盘等各种可 以存储程序代码的介质。
以上所述, 仅为本发明的具体实施方式, 但本发明的保护范围并不局限 于此, 任何熟悉本技术领域的技术人员在本发明揭露的技术范围内, 可轻易 想到变化或替换, 都应涵盖在本发明的保护范围之内。 因此, 本发明的保护 范围应所述以权利要求的保护范围为准。

Claims

权 利 要 求 书
1、 一种谐波分析的方法, 其特征在于, 包括:
对输入的时域信号进行时频转换,以获得所述时域信号在频域上的频谱; 对所述频谱进行滤波处理, 以确定目标频谱;
对所述目标频谱的谐波间杂波做消除处理, 以便确定所述目标频谱的谐 波的结构;
利用所述目标频谱的谐波的结构对所述谐波进行谐波分析。
2、根据权利要求 1所述的方法, 其特征在于, 所述对所述频谱进行滤波 处理, 以确定目标频谱, 包括:
对所述频语进行 k次滤波处理, 并在所述频语经过所述 k次滤波处理中 的每次滤波处理之后记录与本次滤波处理对应的波峰个数以获得总共 k个波 峰个数, 其中 k为大于或等于 2的正整数;
确定所述 k个波峰个数中的最小波峰个数;
确定得到所述最小波峰个数的滤波处理的频谱为目标频谱。
3、根据权利要求 1所述的方法, 其特征在于, 所述对所述频谱进行滤波 处理, 以确定目标频谱, 包括:
对所述频谱进行第一滤波处理, 并记录经过所述第一滤波处理的频谱的 第一波峰个数;
对所述频谱进行第二滤波处理, 并记录经过所述第二滤波处理的频谱的 第二波峰个数;
对所述频谱进行第三滤波处理, 并记录经过所述第三滤波处理的频谱的 第三波峰个数;
当所述第二波峰个数小于所述第一波峰个数, 且所述第二波峰个数小于 所述第三波峰个数, 则确定经过所述第二滤波处理得到的频谱为目标频谱。
4、根据权利要求 3所述的方法, 其特征在于, 所述第一滤波处理为第一 阶滤波、 所述第二滤波处理为第二阶滤波, 以及所述第三滤波处理为第三阶 滤波, 所述对所述频谱进行滤波处理, 以确定目标频谱, 包括:
对所述频谱进行第一阶滤波, 并记录经过所述第一阶滤波的频谱的第一 波峰个数; 对所述频谱进行第二阶滤波, 并记录经过所述第二阶滤波的频谱的第二 波峰个数;
对所述频谱进行第三阶滤波, 并记录经过所述第三阶滤波的频谱的第三 波峰个数;
当所述第二波峰个数小于所述第一波峰个数, 且所述第二波峰个数小于 所述第三波峰个数, 则确定经过所述第二阶滤波的频谱为目标频谱。
5、根据权利要求 1所述的方法, 其特征在于, 所述对所述频谱进行滤波 处理, 以确定目标频谱, 包括:
釆用低通滤波器对所述频谱进行低通滤波处理, 将所述经过所述低通滤 波处理后的频谱作为目标频谱。
6、根据权利要求 1所述的方法, 其特征在于, 所述对所述频谱进行滤波 处理, 以确定目标频谱, 包括:
对所述频谱进行第四阶滤波;
确定经过所述第四阶滤波得到的频谱为目标频谱;
其中所述第四阶为固定值。
7、根据权利要求 1至 6中任一项所述的方法, 其特征在于, 所述对所述 目标频谱的谐波间杂波做消除处理, 包括釆用以下方式中的至少一种: 依据波瓣宽度对所述目标频谱的谐波间杂波做消除处理, 其中所述波瓣 宽度是波峰两侧的两个最邻近的波谷之间的宽度; 以及
依据波峰的倾斜度对所述目标频谱的谐波间杂波做消除处理, 其中所述 波峰的倾斜度是所述波峰两侧的两个最邻近的波谷之间的连线与水平方向的 夹角, 或者所述波峰两侧的两个最邻近的波谷之间的连线的斜率, 或者所述 波峰两侧的两个最邻近的波谷之间的连线的夹角的三角函数。
8、根据权利要求 7所述的方法, 其特征在于, 所述依据波瓣宽度对所述 目标频谱的谐波间杂波做消除处理, 包括:
确定所述目标频谱的全部波峰和波谷;
根据确定的所述波峰和所述波谷, 确定所述目标频谱中的波瓣宽度; 当所述波瓣宽度小于第一波瓣宽度, 则确定所述波瓣宽度对应的波峰为 杂波的波峰; 通过消去所述杂波的波瓣来消除所述目标频谱的杂波。
9、根据权利要求 8所述的方法, 其特征在于, 所述第一波瓣宽度为预设 的固定值, 或者所述第一波瓣宽度为平均波瓣宽度, 其中所述平均波瓣宽度 是所述目标频谱的宽度与所述第二波峰个数的比值。
10、 根据权利要求 7所述的方法, 其特征在于, 所述依据波峰的倾斜度 消除所述目标频谱的杂波, 包括:
确定所述目标频谱的全部波峰和波谷;
根据确定的所述波峰和所述波谷, 确定所述波峰的倾斜度;
当所述波峰的倾斜度大于第一阔值, 则确定所述波峰为杂波波峰;
11、根据权利要求 10所述的方法, 其特征在于, 所述第一阔值为预设的 固定值。
12、根据权利要求 8或 10所述的方法, 其特征在于, 所述通过消去所述 杂波的波瓣来消除所述目标频谱的杂波, 包括:
将所述杂波的波瓣位于两个波谷之间的频点幅值以在两个波谷幅值之间 的内插来代替。
13、 一种确定谐波间杂波的方法, 其特征在于, 包括:
确定所述频谱的全部波峰和波谷;
根据确定的所述波峰和所述波谷, 确定所述频谱中的波瓣宽度, 其中所 述波瓣宽度是波峰两侧的两个最邻近的波谷之间的宽度;
当所述波瓣宽度小于第一波瓣宽度, 则确定所述波瓣宽度对应的波峰为 杂波的波峰。
14、根据权利要求 13所述的方法, 其特征在于, 所述第一波瓣宽度为预 设的固定值。
15、根据权利要求 13所述的方法, 其特征在于, 所述第一波瓣宽度为平 均波瓣宽度, 其中所述平均波瓣宽度是所述频谱的宽度与所述全部波峰的个 数的比值。
16、 一种确定谐波间杂波的方法, 其特征在于, 包括:
确定所述目标频谱的全部波峰和波谷; 根据确定的所述波峰和所述波谷, 确定所述波峰的倾斜度; 当所述波峰的倾斜度大于第一阔值, 则确定所述波峰为杂波的波峰。
17、 根据权利要求 16所述的方法, 其特征在于,
所述波峰的倾斜度是所述波峰两侧的两个最邻近的波谷之间的连线与水 平方向的夹角; 或者
所述波峰的倾斜度是所述波峰两侧的两个最邻近的波谷之间的连线的斜 率: 或者
所述波峰的倾斜度是所述波峰两侧的两个最邻近的波谷之间的连线的夹 角的三角函数。
18、 根据权利要求 16或 17所述的方法, 其特征在于, 所述第一阔值为 预设的固定值。
19、 一种谐波分析的装置, 其特征在于, 包括:
时频转换单元, 用于对输入的时域信号进行时频转换, 以获得所述时域 信号在频域上的频谱;
滤波单元,用于对所述时频转换单元转换得到的所述频谱进行滤波处理, 以确定目标频语;
杂波消除单元, 用于对所述滤波单元确定的所述目标频谱的谐波间杂波 做消除处理, 以便确定所述目标频谱的谐波的结构;
谐波分析单元, 用于利用所述目标频谱的谐波的结构对由所述杂波消除 单元消除了杂波的所述谐波进行谐波分析。
20、根据权利要求 19所述的装置,其特征在于,所述滤波单元具体用于: 对所述频谱进行 k次滤波处理, 并在所述频谱经过所述 k次滤波处理的 每次滤波处理之后记录波峰个数以获得 k个波峰个数, 其中 k为大于或等于
2的正整数;
确定所述 k个波峰个数中的最小波峰个数;
确定得到所述最小波峰个数的滤波处理的频谱为目标频谱。
21、根据权利要求 19所述的装置,其特征在于,所述滤波单元具体用于: 对所述频谱进行第一滤波处理, 并记录经过所述第一滤波处理的频谱的 第一波峰个数; 对所述频谱进行第二滤波处理, 并记录经过所述第二滤波处理的频谱的 第二波峰个数;
对所述频谱进行第三滤波处理, 并记录经过所述第三滤波处理的频谱的 第三波峰个数;
当所述第二波峰个数小于所述第一波峰个数, 且所述第二波峰个数小于 所述第三波峰个数, 则确定经过所述第二滤波处理得到的频谱为目标频谱。
22、根据权利要求 21所述的装置, 其特征在于, 所述第一滤波处理为第 一阶滤波、 所述第二滤波处理为第二阶滤波, 以及所述第三滤波处理为第三 阶滤波, 所述滤波单元具体用于:
对所述频谱进行第一阶滤波, 并记录经过所述第一阶滤波的频谱的第一 波峰个数;
对所述频谱进行第二阶滤波, 并记录经过所述第二阶滤波的频谱的第二 波峰个数;
对所述频谱进行第三阶滤波, 并记录经过所述第三阶滤波的频谱的第三 波峰个数;
当所述第二波峰个数小于所述第一波峰个数, 且所述第二波峰个数小于 所述第三波峰个数, 则确定经过所述第二阶滤波的频谱为目标频谱。
23、根据权利要求 19所述的装置,其特征在于,所述滤波单元具体用于: 釆用低通滤波器对所述频谱进行低通滤波处理, 将所述经过所述低通滤 波处理后的频谱作为目标频谱。
24、根据权利要求 19所述的装置,其特征在于,所述滤波单元具体用于: 对所述频谱进行第四阶滤波;
确定经过所述第四阶滤波得到的频谱为目标频谱;
其中所述第四阶为固定值。
25、 根据权利要求 19至 24中任一项所述的装置, 其特征在于, 所述杂 波消除单元釆用以下方式中的至少一种:
依据波瓣宽度对所述目标频谱的谐波间杂波做消除处理, 其中所述波瓣 宽度是波峰两侧的两个最邻近的波谷之间的宽度;
依据波峰的倾斜度对所述目标频谱的谐波间杂波做消除处理, 其中所述 波峰的倾斜度是所述波峰两侧的两个最邻近的波谷之间的连线与水平方向的 夹角, 或者所述波峰两侧的两个最邻近的波谷之间的连线的斜率, 或者所述 波峰两侧的两个最邻近的波谷之间的连线的夹角的三角函数。
26、根据权利要求 25所述的装置, 其特征在于, 所述杂波消除单元具体 用于:
确定所述目标频谱的全部波峰和波谷;
根据确定的所述波峰和所述波谷, 确定所述目标频谱中的波瓣宽度; 当所述波瓣宽度小于第一波瓣宽度, 则确定所述波瓣宽度对应的波峰为 杂波的波峰;
27、根据权利要求 26所述的装置, 其特征在于, 所述第一波瓣宽度为预 设的固定值, 或者所述第一波瓣宽度为平均波瓣宽度, 其中所述平均波瓣宽 度是所述目标频谱的宽度与所述第二波峰个数的比值。
28、根据权利要求 25所述的装置, 其特征在于, 所述杂波消除单元具体 用于:
确定所述目标频谱的全部波峰和波谷;
根据确定的所述波峰和所述波谷, 确定所述波峰的倾斜度;
当所述波峰的倾斜度大于第一阔值, 则确定所述波峰为杂波波峰;
29、根据权利要求 28所述的装置, 其特征在于, 所述第一阔值为预设的 固定值。
30、 根据权利要求 26或 28所述的装置, 其特征在于, 所述杂波消除单 元具体用于:
将所述杂波的波瓣位于两个波谷之间的频点幅值以在两个波谷幅值之间 的内插来代替。
31、 一种确定谐波间杂波的装置, 其特征在于, 包括:
第一确定单元, 用于确定所述频谱的全部波峰和波谷;
第二确定单元,用于根据所述第一确定单元确定的所述波峰和所述波谷, 确定所述频谱中的波瓣宽度, 其中所述波瓣宽度是波峰两侧的两个最邻近的 波谷之间的宽度;
第三确定单元, 用于当所述波瓣宽度小于第一波瓣宽度, 则确定由所述 第二确定单元确定的所述波瓣宽度对应的波峰为杂波的波峰。
32、根据权利要求 31所述的装置, 其特征在于, 所述第一波瓣宽度为预 设的固定值。
33、根据权利要求 32所述的装置, 其特征在于, 所述第一波瓣宽度为平 均波瓣宽度, 其中所述平均波瓣宽度是所述频谱的宽度与所述全部波峰的个 数的比值。
34、 一种确定谐波间杂波的装置, 其特征在于, 包括:
第四确定单元, 用于确定所述目标频谱的全部波峰和波谷;
第五确定单元,用于根据所述第四确定单元确定的所述波峰和所述波谷, 确定所述波峰的倾斜度;
第六确定单元, 用于当所述波峰的倾斜度大于第一阔值, 则确定由所述 第五确定单元确定的所述波峰为杂波的波峰。
35、 根据权利要求 34所述的装置, 其特征在于,
所述波峰的倾斜度是所述波峰两侧的两个最邻近的波谷之间的连线与水 平方向的夹角; 或者
所述波峰的倾斜度是所述波峰两侧的两个最邻近的波谷之间的连线的斜 率: 或者
所述波峰的倾斜度是所述波峰两侧的两个最邻近的波谷之间的连线的夹 角的三角函数。
36、 根据权利要求 34或 35所述的装置, 其特征在于, 所述第一阔值为 预设的固定值。
PCT/CN2014/080562 2013-06-26 2014-06-24 谐波分析方法和装置以及确定谐波间杂波的方法和装置 WO2014206265A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201310260866.1A CN104251934B (zh) 2013-06-26 2013-06-26 谐波分析方法和装置以及确定谐波间杂波的方法和装置
CN201310260866.1 2013-06-26

Publications (1)

Publication Number Publication Date
WO2014206265A1 true WO2014206265A1 (zh) 2014-12-31

Family

ID=52141058

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2014/080562 WO2014206265A1 (zh) 2013-06-26 2014-06-24 谐波分析方法和装置以及确定谐波间杂波的方法和装置

Country Status (2)

Country Link
CN (1) CN104251934B (zh)
WO (1) WO2014206265A1 (zh)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105550516A (zh) * 2015-12-11 2016-05-04 合肥工业大学 基于Adams算法的频率自适应谐波间谐波分析方法
CN105574331A (zh) * 2015-12-11 2016-05-11 合肥工业大学 基于Adams算法的指定频率的谐波间谐波分析方法
CN106483375A (zh) * 2016-09-22 2017-03-08 电子科技大学 一种多频率分次谐波检测方法

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105510032B (zh) * 2015-12-11 2017-12-26 西安交通大学 基于谐噪比指导的解卷积方法
CN113341223B (zh) * 2021-06-02 2022-12-06 贵州电网有限责任公司 一种适用于电网谐波分析及谐波源定位的方法
CN113282785B (zh) * 2021-07-19 2021-11-05 北京快鱼电子股份公司 一种基于未标注关键词数据的关键词检测方法及系统
CN113823277A (zh) * 2021-11-23 2021-12-21 北京百瑞互联技术有限公司 基于深度学习的关键词识别方法、系统、介质及设备
CN114121004B (zh) * 2021-11-23 2024-05-03 北京百瑞互联技术股份有限公司 基于深度学习的语音识别方法、系统、介质及设备

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US3978287A (en) * 1974-12-11 1976-08-31 Nasa Real time analysis of voiced sounds
JP2001516902A (ja) * 1997-09-18 2001-10-02 マトラ ノーテル コミュニカショーン デジタル音声信号中のノイズを抑制する方法
CN1342968A (zh) * 2000-09-13 2002-04-03 中国科学院自动化研究所 用于语音识别的高精度高分辨率基频提取方法
CN101113995A (zh) * 2007-08-29 2008-01-30 湖南大学 基于Nuttall窗双峰插值FFT的基波与谐波检测方法
CN101641734A (zh) * 2007-03-23 2010-02-03 三星电子株式会社 对音频信号编码的方法和设备以及对音频信号解码的方法和设备
CN101915874A (zh) * 2010-07-20 2010-12-15 北海市深蓝科技发展有限责任公司 一种基于傅立叶变换的谐波检测方法
US20120278020A1 (en) * 2011-04-29 2012-11-01 Analog Devices, Inc. Apparatus and method for real time harmonic spectral analyzer

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1848240B (zh) * 2005-04-12 2011-12-21 佳能株式会社 基于离散对数傅立叶变换的基音检测方法、设备和介质

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US3978287A (en) * 1974-12-11 1976-08-31 Nasa Real time analysis of voiced sounds
JP2001516902A (ja) * 1997-09-18 2001-10-02 マトラ ノーテル コミュニカショーン デジタル音声信号中のノイズを抑制する方法
CN1342968A (zh) * 2000-09-13 2002-04-03 中国科学院自动化研究所 用于语音识别的高精度高分辨率基频提取方法
CN101641734A (zh) * 2007-03-23 2010-02-03 三星电子株式会社 对音频信号编码的方法和设备以及对音频信号解码的方法和设备
CN101113995A (zh) * 2007-08-29 2008-01-30 湖南大学 基于Nuttall窗双峰插值FFT的基波与谐波检测方法
CN101915874A (zh) * 2010-07-20 2010-12-15 北海市深蓝科技发展有限责任公司 一种基于傅立叶变换的谐波检测方法
US20120278020A1 (en) * 2011-04-29 2012-11-01 Analog Devices, Inc. Apparatus and method for real time harmonic spectral analyzer

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105550516A (zh) * 2015-12-11 2016-05-04 合肥工业大学 基于Adams算法的频率自适应谐波间谐波分析方法
CN105574331A (zh) * 2015-12-11 2016-05-11 合肥工业大学 基于Adams算法的指定频率的谐波间谐波分析方法
CN105550516B (zh) * 2015-12-11 2019-01-04 合肥工业大学 基于Adams算法的频率自适应谐波间谐波分析方法
CN105574331B (zh) * 2015-12-11 2019-03-26 合肥工业大学 基于Adams算法的指定频率的谐波间谐波分析方法
CN106483375A (zh) * 2016-09-22 2017-03-08 电子科技大学 一种多频率分次谐波检测方法
CN106483375B (zh) * 2016-09-22 2019-02-01 电子科技大学 一种多频率分次谐波检测方法

Also Published As

Publication number Publication date
CN104251934A (zh) 2014-12-31
CN104251934B (zh) 2018-08-14

Similar Documents

Publication Publication Date Title
WO2014206265A1 (zh) 谐波分析方法和装置以及确定谐波间杂波的方法和装置
US11031029B2 (en) Pitch detection algorithm based on multiband PWVT of teager energy operator
Jwo et al. Windowing techniques, the welch method for improvement of power spectrum estimation
Sharma et al. Improved eigenvalue decomposition-based approach for reducing cross-terms in Wigner–Ville distribution
CN108921014A (zh) 一种基于改进噪声包络信号识别的螺旋桨轴频搜索方法
CN112244835B (zh) 信号处理方法及相关装置
CN112190268A (zh) 生理信号处理设备
CN113341378B (zh) 基于频谱差分熵检测的自适应信道化接收方法
CN114325740A (zh) 一种应用于激光多普勒测速仪的速度测量方法与系统
US20060178881A1 (en) Method and apparatus for detecting voice region
CN114724573A (zh) 一种啸叫抑制方法、装置、计算机可读存储介质以及系统
CN115762551A (zh) 鼾声检测方法、装置、计算机设备及存储介质
CN113243900B (zh) 一种视频心率检测的rPPG信号预处理方法及系统
JP7152112B2 (ja) 信号処理装置、信号処理方法および信号処理プログラム
CN114858430A (zh) 一种基于lcd-小波新阈值的机械密封声发射信号降噪方法
CN113948088A (zh) 基于波形模拟的语音识别方法及装置
CN116057628A (zh) 用于语音和音乐录音的嗡嗡噪声检测和去除
JP7461192B2 (ja) 基本周波数推定装置、アクティブノイズコントロール装置、基本周波数の推定方法及び基本周波数の推定プログラム
CN117727314B (zh) 用于生态音频信息的滤波增强方法
CN115376548B (zh) 一种音频信号有声段端点检测方法及系统
CN112932410B (zh) 基于切比雪夫多项式拟合的生理包络信号幅值计算方法
CN113643689B (zh) 一种数据滤波方法和相关设备
CN117037834B (zh) 一种会议语音数据智能采集方法及系统
CN114333880B (zh) 一种信号处理方法、装置、设备及存储介质
CN109655913B (zh) 地震信号动态滤波方法及系统

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 14817714

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 14817714

Country of ref document: EP

Kind code of ref document: A1