CN105551501B - Harmonic signal fundamental frequency estimation algorithm and device - Google Patents

Harmonic signal fundamental frequency estimation algorithm and device Download PDF

Info

Publication number
CN105551501B
CN105551501B CN201610044926.XA CN201610044926A CN105551501B CN 105551501 B CN105551501 B CN 105551501B CN 201610044926 A CN201610044926 A CN 201610044926A CN 105551501 B CN105551501 B CN 105551501B
Authority
CN
China
Prior art keywords
fundamental frequency
spectrum
normal
signal
harmonic
Prior art date
Legal status (The legal status 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 status listed.)
Active
Application number
CN201610044926.XA
Other languages
Chinese (zh)
Other versions
CN105551501A (en
Inventor
张维维
李敏
许爽
董玉华
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Dalian Sailing Technology Co Ltd
Original Assignee
Dalian Nationalities University
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 Dalian Nationalities University filed Critical Dalian Nationalities University
Priority to CN201811495768.5A priority Critical patent/CN109524023A/en
Priority to CN201610044926.XA priority patent/CN105551501B/en
Priority to CN201811494641.1A priority patent/CN109410980A/en
Priority to CN201811494657.2A priority patent/CN109493880A/en
Publication of CN105551501A publication Critical patent/CN105551501A/en
Application granted granted Critical
Publication of CN105551501B publication Critical patent/CN105551501B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • GPHYSICS
    • G10MUSICAL INSTRUMENTS; ACOUSTICS
    • G10LSPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L25/00Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
    • G10L25/03Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters
    • G10L25/18Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters the extracted parameters being spectral information of each sub-band
    • GPHYSICS
    • G10MUSICAL INSTRUMENTS; ACOUSTICS
    • G10LSPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L19/00Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis
    • G10L19/02Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis using spectral analysis, e.g. transform vocoders or subband vocoders
    • GPHYSICS
    • G10MUSICAL INSTRUMENTS; ACOUSTICS
    • G10LSPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L25/00Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
    • G10L25/27Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the analysis technique
    • GPHYSICS
    • G10MUSICAL INSTRUMENTS; ACOUSTICS
    • G10LSPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L25/00Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
    • G10L25/45Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of analysis window

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Computational Linguistics (AREA)
  • Signal Processing (AREA)
  • Health & Medical Sciences (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Human Computer Interaction (AREA)
  • Acoustics & Sound (AREA)
  • Multimedia (AREA)
  • Spectroscopy & Molecular Physics (AREA)
  • Auxiliary Devices For Music (AREA)
  • Compression, Expansion, Code Conversion, And Decoders (AREA)

Abstract

A kind of harmonic signal fundamental frequency estimation algorithm and device, belong to field of signal processing, in order to solve the problem of it is quick, accurately extract fundamental frequency and reduce influence of the spectral leakage to high frequency spectral peak, technical essential includes: that S1. makees normal Q transformation to audio signal, obtains normal Q conversion spectrum;S2. calculate to fold harmonic wave summation spectrum using normal Q conversion spectrum;S3. by the spectral peak preliminary screening fundamental frequency of folding harmonic wave summation spectrum;S4. the posterior probability density function for establishing fundamental frequency candidate extracts simultaneously output harmonic wave signal fundamental frequency according to maximum posteriori criterion.Effect is: being converted with normal Q and realizes multiresolution Power estimation, folds harmonic wave summation spectrum based on the definition of normal Q conversion spectrum, and be used for preliminary fundamental frequency screening.Fundamental frequency estimation is realized by maximum a posteriori probability method, and wherein prior probability is found out by folding harmonic wave summation spectrum, and likelihood function is determined by the matching degree of harmonic wave spectral peak and observation frequency spectrum.

Description

Harmonic signal fundamental frequency estimation algorithm and device
Technical field
The invention belongs to field of signal processing, are related to a kind of harmonic signal fundamental frequency estimation algorithm.
Background technique
Currently used fundamental frequency estimation algorithm includes correlation method, Cepstrum Method, Wavelet Transform, harmonic wave summation etc..From Correlation method has the advantages that principle is simple, is easily achieved, but its calculation amount is as every frame signal length is at a square multiplication.Cepstrum Method In voice switching, since signal-to-noise ratio reduction often results in Cepstrum peak substantial deviation fundamental frequency.Wavelet Transform is closed by detecting glottis Pitch period is extrapolated in signal characteristic mutation when conjunction, but multiple threshold values in wavelet transformation are difficult to accurately select.Harmonic wave summation side Method takes into consideration only the summation of each harmonic amplitude weighting, without considering that each harmonic frequency may be slightly offset from harmonic wave integer Times, spectral leakage phenomenon causes the pseudo- peak number amount increase of signal in high-frequency range and part high fdrequency component can be by low frequency component Secondary lobe is buried.
It is high that the Chinese invention patent application of Publication No. CN1342968A discloses a kind of high-precision for speech recognition Resolution ratio fundamental frequency extracting method, and specifically describe in fundamental frequency extracting method have signal segmentation, adding window, determine fundamental frequency candidate, Filtering has too low HperValue or RperThe step of value, finally asks fundamental frequency path with DP algorithm.The purpose is to can in the time domain, Related coefficient evaluation and test is carried out to candidate fundamental frequency.However, when this kind of way extracts fundamental frequency candidate, although the operation of DP can be reduced Amount, still, which belongs to correlation method, still has computationally intensive, the slow disadvantage of extraction rate.
Summary of the invention
In order to solve the problem of it is quick, accurately extract fundamental frequency and reduce influence of the spectral leakage to high frequency spectral peak, the present invention A kind of harmonic signal fundamental frequency estimation algorithm is proposed, to improve the speed and accuracy of fundamental frequency extraction.
To achieve the goals above, the technical solution adopted by the present invention is that: a kind of harmonic signal fundamental frequency estimation algorithm, packet Include: S1. makees normal Q to AF harmonic signal and converts, and obtains normal Q conversion spectrum;S2. it calculates to fold harmonic wave using normal Q conversion spectrum to sum Spectrum;S3. by the spectral peak preliminary screening fundamental frequency of folding harmonic wave summation spectrum;S4. the posterior probability density function of fundamental frequency candidate, root are established Simultaneously output harmonic wave signal fundamental frequency is extracted according to maximum posteriori criterion.
The utility model has the advantages that AF harmonic signal fundamental frequency estimation algorithm of the present invention, is converted with normal Q and realizes that multiresolution spectrum is estimated Meter folds harmonic wave summation spectrum based on the definition of normal Q conversion spectrum, and is used for preliminary fundamental frequency screening.Pass through maximum a posteriori probability method Realize fundamental frequency estimation, wherein prior probability is found out by folding harmonic wave summation spectrum, and likelihood function is by harmonic wave spectral peak and observes frequency spectrum Matching degree determines.
AF harmonic signal fundamental frequency estimation is a key technology of field of signal processing, in speech recognition, music signal Processing, Underwater acoustic signal processing etc. play a significant role, and fundamental frequency estimation algorithm proposed by the present invention may be used on all kinds of having In the fundamental frequency estimation of the signal of harmonic structure, the method for proposition can be realized frequency spectrum multiresolution analysis, propose that folding harmonic wave asks Preliminary screening with spectrum as fundamental frequency estimates fundamental frequency using maximum a posteriori probability method.The energy that this method makes full use of frequency spectrum to provide The characteristics of information such as amount, harmonic wave, spectrum matching, realize fundamental frequency estimation, have arithmetic speed fast, and parameter is few, strong robustness.
Detailed description of the invention
Fig. 1 is the frame music signal diagram in embodiment 10;
Fig. 2 is the normal Q conversion spectrum of the music signal in embodiment 10;
Fig. 3 is the folding harmonic wave summation spectrum in embodiment 10;
Fig. 4 is the fundamental frequency posterior probability density spectra in embodiment 10;
Fig. 5 is the time-domain signal diagram in embodiment 11;
Fig. 6 is the normal Q conversion spectrum in embodiment 11;
Fig. 7 is the folding harmonic wave summation spectrum in embodiment 11;
Fig. 8 is the fundamental frequency posterior probability density spectra in embodiment 11;
Fig. 9 is the software flow pattern of the method for the invention.
Specific embodiment
Embodiment 1:A kind of harmonic signal fundamental frequency estimation algorithm, comprising steps of
S1. make normal Q to AF harmonic signal to convert, obtain normal Q conversion spectrum;
S2. calculate to fold harmonic wave summation spectrum using normal Q conversion spectrum;
S3. by the spectral peak preliminary screening fundamental frequency of folding harmonic wave summation spectrum;
S4. the posterior probability density function for establishing fundamental frequency candidate extracts simultaneously output harmonic wave according to maximum posteriori criterion Signal fundamental frequency.
AF harmonic signal fundamental frequency estimation algorithm of the present invention is converted with normal Q and realizes multiresolution Power estimation, based on humorous Wave summation energy definition folds harmonic wave summation spectrum, and is used for preliminary fundamental frequency screening.Base is realized by maximum a posteriori probability method Frequency estimates that wherein prior probability is found out by folding harmonic wave summation spectrum, and likelihood function is determined by the matching degree of harmonic wave spectral peak and frequency spectrum.
AF harmonic signal fundamental frequency estimation is a key technology of field of signal processing, in speech recognition, music signal Processing, Underwater acoustic signal processing etc. play a significant role, and fundamental frequency estimation algorithm proposed by the present invention may be used on all kinds of having In the fundamental frequency estimation of the signal of harmonic structure, the method for proposition can be realized frequency spectrum multiresolution analysis, propose that folding harmonic wave asks Preliminary screening with spectrum as fundamental frequency estimates fundamental frequency using maximum a posteriori probability method.The energy that this method makes full use of frequency spectrum to provide The characteristics of information such as amount, harmonic wave, spectrum matching, realize fundamental frequency estimation, have arithmetic speed fast, and parameter is few, strong robustness.
Embodiment 2:It, more specifically, can be first to audio before normal Q transformation with technical solution same as Example 1 Harmonic signal is divided, windowing process.
In the step S1, sub-frame processing first is made to the non-stationary AF harmonic signal of input, then gives framing in short-term Signal adds Hanning window suppressed sidelobes amplitude, and length of window is consistent with the time window width that normal Q is converted.
Embodiment 3:With technical solution identical with embodiment 1 or 2, more specifically, in the step S3, screening rule Then to choose the corresponding frequency of at least preceding 3 maximum in folding harmonic wave summation spectrum as candidate fundamental frequency, the present embodiment is selected Preceding 3 maximum is as candidate fundamental frequency, and in experiment, candidate fundamental frequency quantity increases, and accuracy can be improved, and chooses 3, quasi- Exactness is very high.
Embodiment 4:With being more specifically basis in step S1 with embodiment 1 or 2 or 3 identical technical solutions Human hearing characteristic calculates the normal Q conversion spectrum of logarithmic frequency domain, the normal Q conversion spectrum are as follows:
(1) in formula, Q is quality factor, is constant, and N [k] is that normal Q is converted in the corresponding time window width of k-th of frequency point It spends, in (1) formula, x (l, m) indicates the sampled point of the serial number m in l frame signal;If x1(n), n=0,1 ... .M-1 is indicated Length is M, sample rate FsAudio signal, the audio signal truncation be every segment length be N [k] frame, if frame move be L, Then l frame signal can indicate are as follows:
X (l, m)=x1(m+lL), m=0,1 ... N [k] -1 (2)
(1) in formula, wN[k](m) indicate that length is the Hanning window of N [k], it may be assumed that
Embodiment 5:With with embodiment 1 or 2 or 3 or 4 identical technical solutions, more specifically, fold harmonic wave summation Spectrum is defined as:
Wherein: h is overtone order, and H is highest subharmonic, 0 < α < 1, XQ(k, l) is normal Q conversion spectrum, round () fortune It calculates result and is equal to nearest integer, it is assumed that every octave takes Oct point, and the lowest frequency components of normal Q transform analysis are fmin, most High frequency components are fmax, thenWherein ceil () operation result is equal to just infinite side To nearest integer, the corresponding frequency of k-th of Frequency point of normal Q transformation are as follows:
Embodiment 6:With with embodiment 1 or 2 or 3 or 4 or 5 identical technical solutions, more specifically, fundamental frequency posteriority is general Rate density is defined as:
Wherein: Fi,lCandidate, the p (F in above formula for i-th of fundamental frequency of l framei,l) it is i-th of fundamental frequency F of l framei,lPriori Probability defines p (Fi,l)=Xc(Fi,l,l);P (X in above formulac(fk,l)|Fi,l) measure given fundamental frequency Fi,lIt obtains observing normal Q frequency spectrum Probability.
Embodiment 7:With with embodiment 1 or 2 or 3 or 4 or 5 identical technical solutions, more specifically, definition:
Wherein:
(8) β=0.03 in formula, if often fundamental frequency candidate F in Q spectrumi,lH subharmonic be spectral peak then γl(h,Fi,l) it is 1, It otherwise is 0, thereforeCharacterize the quantity of preceding H order harmonic components included in the normal Q spectrum of l frame.Similarly, if XQ(k, It l) is spectral peak then λl(k) it is 1, is otherwise 0, thereforeCharacterize F in the normal Q spectrum of l framei,lPreceding H subharmonic frequency All spectral peak numbers within the scope of rate.
The normalization weighted sum that harmonic wave summation spectrum is fundamental frequency and each harmonic component is folded, if fundamental frequency is F0, then fold humorous Wave summation spectrum is in F0Functional value at position is that normal Q is converted in F0,2F0,3F0,...,HF0Etc. amplitude normalization weighted sum; Harmonic wave summation spectrum is folded in 2F0Functional value at position is that normal Q is converted in 2F0,4F0,6F0,...,2HF0Etc. amplitude normalizing Change weighted sum.The signals main energetic such as daily voice, music concentrates on low-frequency range, therefore even if often Q transformation fundamental frequency amplitude is less than two Subharmonic amplitude, larger output can also be obtained at fundamental frequency position by folding harmonic wave summation spectrum.By in folding harmonic wave summation spectrum Preceding several peak values obtain fundamental frequency candidate, can be realized the preliminary screening of fundamental frequency.
First item in maximum a posteriori probability density function is defined as p (Fi,l)=Xc(Fi,l, l), i.e. folding harmonic wave summation Spectrum is in Fi,lThe amplitude at place, p (Fi,l) take fold harmonic wave summation spectrum and very Q conversion spectrum is because of sometimes certain harmonic components Amplitude can be more than the amplitude of fundamental frequency, and even if folding harmonic wave summation spectrum in fundamental frequency position when harmonic amplitude is higher than fundamental frequency amplitude The value at the place of setting is still greater than the value at higher hamonic wave, and then improves the accuracy rate of fundamental frequency estimation, reduces False Rate.
Embodiment 8:With with embodiment 1 or 2 or 3 or 4 or 5 or 6 or 7 identical technical solutions, more specifically, fundamental frequency Estimation formulas are as follows:
Embodiment 9:A kind of harmonic signal fundamental frequency estimation algorithm, includes the following steps:
1. a pair AF harmonic signal makees normal Q transformation, normal Q conversion spectrum is obtained, the normal Q conversion spectrum are as follows:
(1) in formula, Q is quality factor, is constant, and N [k] is that normal Q is converted in the corresponding time window width of k-th of frequency point It spends, in (1) formula, x (l, m) indicates the sampled point of the serial number m in l frame signal;If x1(n), n=0,1 ... .M-1 is indicated Length is M, sample rate FsAudio signal, the audio signal truncation be every segment length be N [k] frame, if frame move be L, Then l frame signal can indicate are as follows:
X (l, m)=x1(m+lL), m=0,1 ... N [k] -1 (2)
(1) in formula, wN[k](m) indicate that length is the Hanning window of N [k], it may be assumed that
Harmonic wave summation spectrum is folded 2. being asked by normal Q conversion spectrum, it may be assumed that
Wherein: h is overtone order, and H is highest subharmonic, 0 < α < 1, XQ(k, l) is normal Q conversion spectrum, round () fortune It calculates result and is equal to nearest integer, it is assumed that every octave takes Oct point, and the lowest frequency components of normal Q transform analysis are fmin, most High frequency components are fmax, thenWherein ceil () operation result is equal to just infinite side To nearest integer, the corresponding frequency of k-th of Frequency point of normal Q transformation are as follows:
3. obtaining the corresponding frequency of 3 peak-peaks as fundamental frequency candidate from folding in harmonic wave summation spectrum, each base is then sought The candidate posterior probability density of frequency, it may be assumed that
Wherein: Fi,lCandidate, the p (F for i-th of fundamental frequency of l framei,l) it is i-th of fundamental frequency F of l framei,lPrior probability, it is fixed Justice is p (Fi,l)=Xc(Fi,l, l), i.e. folding harmonic wave summation spectrum is in Fi,lThe amplitude at place.p(Xc(fk,l)|Fi,l) measure given base Frequency Fi,lObtain observing the likelihood function of normal Q frequency spectrum, the likelihood function is defined as:
Wherein:
(8) β=0.03 in formula, if often fundamental frequency candidate F in Q spectrumi,lH subharmonic be spectral peak then γl(h,Fi,l) it is 1, It otherwise is 0, thereforeCharacterize the quantity of preceding H order harmonic components included in the normal Q spectrum of l frame.Similarly, if XQ (k, l) is spectral peak then λl(k) it is 1, is otherwise 0, thereforeCharacterize F in the normal Q spectrum of l framei,lPreceding H subharmonic All spectral peak numbers in frequency range.
Embodiment 10:The present embodiment carries out experimental verification to the method to fundamental frequency estimation in the various embodiments described above:
The present embodiment experiment is to have used above-described embodiment fundamental frequency estimation method, has carried out fundamental frequency to a frame music signal and has estimated Meter, and emulate and obtain Fig. 1-Fig. 4, the situation of the present embodiment verifying is: the amplitude of fundamental component is greater than second harmonic width in signal Degree.
Attached drawing 1 is a frame music signal.
Attached drawing 2 is the normal Q conversion spectrum of above-mentioned music signal: having harmonic component abundant by the way that normal Q conversion spectrum is visible.Each time Harmonic component, which should be apparent that, to be come, and normal Q spectrum only has very narrow spectral line at each frequency component, and spectral leakage is unobvious.
Attached drawing 3 is to fold harmonic wave summation spectrum: folding harmonic wave summation spectrum can increase the difference of fundamental frequency and each harmonic component.From In select three frequencies with maximum folded harmonic amplitude as candidate fundamental frequency, be selected the candidate fundamental frequency difference come in this figure Fundamental frequency, second harmonic and the third-harmonic component of corresponding original time-domain signal, then ask the posterior probability of these three candidate fundamental frequencies close Degree.
Attached drawing 4 is fundamental frequency posterior probability density spectra: the candidate fundamental frequency with maximum a posteriori probability density, i.e. 181.45Hz Frequency is chosen as final fundamental frequency, coincide with the fact.
Spectral leakage can effectively be inhibited from the visible normal Q transformation of attached drawing 1- attached drawing 4, each frequency component in prominent signal. Step (2)-(4) of the present invention are all based on normal Q spectrum, and fold and only take a small amount of (3) in harmonic wave summation spectrum and have larger folding harmonic wave The fundamental frequency candidate of summation energy participates in maximum a posteriori probability density and calculates, and has less operand, computation complexity is low, can be fast Speed is realized.The experimental results showed that the maximum a posteriori probability function of building can be accurate since harmonic signal has good harmonic wave The each harmonic of candidate fundamental frequency and the matching relationship of observation frequency spectrum are mapped out, the fundamental frequency of harmonic signal can be accurately filtered out.
Embodiment 11:The applicable experimental situations of the present embodiment are: the amplitude of fundamental component is less than two in a frame music signal Subharmonic amplitude.
The frame music signal time-domain signal is as shown in Figure 5;
Normal Q conversion spectrum is as shown in Figure 6;
Although fold harmonic wave summation spectrum as shown in fig. 7, practical fundamental component amplitude less than second harmonic component amplitude, But it folds the corresponding amplitude of fundamental frequency in harmonic wave summation spectrum significantly to be amplified, and is more than the amplitude of second harmonic;
Fundamental frequency posterior probability density spectra is as shown in Figure 8;By above-mentioned experiment attached drawing as it can be seen that being less than harmonic amplitude in fundamental frequency amplitude In the case where, this method still is able to accurately estimate fundamental frequency.
Embodiment 12:A kind of harmonic signal fundamental frequency estimation device, comprising: normal Q conversion module is made AF harmonic signal normal Q transformation, obtains normal Q conversion spectrum;Harmonic wave summation spectrum computing module is folded, calculates to fold harmonic wave summation spectrum using normal Q conversion spectrum;Just Step screening fundamental module, by the spectral peak preliminary screening fundamental frequency of folding harmonic wave summation spectrum;
Harmonic signal fundamental frequency output module, establishes the posterior probability density function of fundamental frequency candidate, according to maximum a posteriori probability Criterion is extracted and output harmonic wave signal fundamental frequency.Device described in the present embodiment, it is real to execute method described in embodiment 1-9 The technical solution applied in a 1-9 is suitable for the present embodiment.
The foregoing is only a preferred embodiment of the present invention, but scope of protection of the present invention is not limited thereto, Anyone skilled in the art in the technical scope disclosed by the present invention, according to the technique and scheme of the present invention and its Inventive concept is subject to equivalent substitution or change, should be covered by the protection scope of the present invention.

Claims (9)

1. a kind of harmonic signal fundamental frequency estimation method, which is characterized in that comprising steps of
S1. make normal Q to AF harmonic signal to convert, obtain normal Q conversion spectrum;
S2. calculate to fold harmonic wave summation spectrum using normal Q conversion spectrum;
S3. by the spectral peak preliminary screening fundamental frequency of folding harmonic wave summation spectrum;
S4. the posterior probability density function for establishing fundamental frequency candidate extracts simultaneously output harmonic wave signal according to maximum posteriori criterion Fundamental frequency.
2. harmonic signal fundamental frequency estimation method as described in claim 1, which is characterized in that in the step S1, to input Non-stationary AF harmonic signal makees sub-frame processing, then adds Hanning window suppressed sidelobes amplitude, length of window to framing signal in short-term It is consistent with the time window width of normal Q transformation.
3. harmonic signal fundamental frequency estimation method as described in claim 1, which is characterized in that in the step S3, screening rule To choose the corresponding frequency of at least preceding 3 maximum in folding harmonic wave summation spectrum as candidate fundamental frequency.
4. harmonic signal fundamental frequency estimation method as described in claim 1, which is characterized in that in step S1, normal Q conversion spectrum are as follows:
(1) in formula, Q is quality factor, is constant, and N [k] is that normal Q is converted in the corresponding time window width of k-th of frequency point, (1) In formula, x (l, m) indicates the sampled point of the serial number m in l frame signal;If x1(n), n=0,1 ... .M-1 indicates length For M, sample rate FsAudio signal, the audio signal truncation be every segment length be N [k] frame, if frame move be L, l Frame signal can indicate are as follows:
X (l, m)=x1(m+lL), m=0,1 ... N [k] -1
(1) in formula, wN[k](m) indicate that length is the Hanning window of N [k], it may be assumed that
5. harmonic signal fundamental frequency estimation method as described in claim 1, which is characterized in that fold harmonic wave summation spectrum is defined as:
Wherein: h is overtone order, and H is highest subharmonic, 0 < α < 1, XQ(k, l) is normal Q conversion spectrum, round () operation knot Fruit is equal to nearest integer, it is assumed that every octave takes Oct point, and the lowest frequency components of normal Q transform analysis are fmin, most high frequency Rate component is fmax, thenWherein ceil () operation result is equal to positive infinity most Close integer, the corresponding frequency of k-th of Frequency point of normal Q transformation are as follows:
6. harmonic signal fundamental frequency estimation method as claimed in claim 5, which is characterized in that the definition of fundamental frequency posterior probability density Are as follows:
Wherein: Fi,lCandidate, the p (F in above formula for i-th of fundamental frequency of l framei,l) it is i-th of fundamental frequency F of l framei,lPrior probability, It is defined as p (Fi,l)=Xc(Fi,l,l);P (X in above formulac(fk,l)|Fi,l) measure given fundamental frequency Fi,lIt obtains observing normal Q frequency spectrum Probability.
7. harmonic signal fundamental frequency estimation method as claimed in claim 6, which is characterized in that definition:
Wherein:
β=0.03,Indicate that the normal Q of l frame composes preceding H subharmonic spectral peak quantity,Characterize l frame Score peak number amount within the scope of preceding H subfrequency.
8. harmonic signal fundamental frequency estimation method as claimed in claim 7, which is characterized in that fundamental frequency estimation formula are as follows:
9. a kind of harmonic signal fundamental frequency estimation device characterized by comprising
Normal Q conversion module is made normal Q to AF harmonic signal and is converted, obtains normal Q conversion spectrum;
Harmonic wave summation spectrum computing module is folded, calculates to fold harmonic wave summation spectrum using normal Q conversion spectrum;
Preliminary screening fundamental module, by the spectral peak preliminary screening fundamental frequency of folding harmonic wave summation spectrum;
Harmonic signal fundamental frequency output module, establishes the posterior probability density function of fundamental frequency candidate, according to maximum posteriori criterion Extract simultaneously output harmonic wave signal fundamental frequency.
CN201610044926.XA 2016-01-22 2016-01-22 Harmonic signal fundamental frequency estimation algorithm and device Active CN105551501B (en)

Priority Applications (4)

Application Number Priority Date Filing Date Title
CN201811495768.5A CN109524023A (en) 2016-01-22 2016-01-22 A kind of method of pair of fundamental frequency estimation experimental verification
CN201610044926.XA CN105551501B (en) 2016-01-22 2016-01-22 Harmonic signal fundamental frequency estimation algorithm and device
CN201811494641.1A CN109410980A (en) 2016-01-22 2016-01-22 A kind of application of fundamental frequency estimation algorithm in the fundamental frequency estimation of all kinds of signals with harmonic structure
CN201811494657.2A CN109493880A (en) 2016-01-22 2016-01-22 A kind of method of harmonic signal fundamental frequency preliminary screening

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201610044926.XA CN105551501B (en) 2016-01-22 2016-01-22 Harmonic signal fundamental frequency estimation algorithm and device

Related Child Applications (3)

Application Number Title Priority Date Filing Date
CN201811494641.1A Division CN109410980A (en) 2016-01-22 2016-01-22 A kind of application of fundamental frequency estimation algorithm in the fundamental frequency estimation of all kinds of signals with harmonic structure
CN201811495768.5A Division CN109524023A (en) 2016-01-22 2016-01-22 A kind of method of pair of fundamental frequency estimation experimental verification
CN201811494657.2A Division CN109493880A (en) 2016-01-22 2016-01-22 A kind of method of harmonic signal fundamental frequency preliminary screening

Publications (2)

Publication Number Publication Date
CN105551501A CN105551501A (en) 2016-05-04
CN105551501B true CN105551501B (en) 2019-03-15

Family

ID=55830649

Family Applications (4)

Application Number Title Priority Date Filing Date
CN201811495768.5A Pending CN109524023A (en) 2016-01-22 2016-01-22 A kind of method of pair of fundamental frequency estimation experimental verification
CN201811494641.1A Pending CN109410980A (en) 2016-01-22 2016-01-22 A kind of application of fundamental frequency estimation algorithm in the fundamental frequency estimation of all kinds of signals with harmonic structure
CN201811494657.2A Pending CN109493880A (en) 2016-01-22 2016-01-22 A kind of method of harmonic signal fundamental frequency preliminary screening
CN201610044926.XA Active CN105551501B (en) 2016-01-22 2016-01-22 Harmonic signal fundamental frequency estimation algorithm and device

Family Applications Before (3)

Application Number Title Priority Date Filing Date
CN201811495768.5A Pending CN109524023A (en) 2016-01-22 2016-01-22 A kind of method of pair of fundamental frequency estimation experimental verification
CN201811494641.1A Pending CN109410980A (en) 2016-01-22 2016-01-22 A kind of application of fundamental frequency estimation algorithm in the fundamental frequency estimation of all kinds of signals with harmonic structure
CN201811494657.2A Pending CN109493880A (en) 2016-01-22 2016-01-22 A kind of method of harmonic signal fundamental frequency preliminary screening

Country Status (1)

Country Link
CN (4) CN109524023A (en)

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108108333B (en) * 2017-05-02 2021-10-19 大连民族大学 Method for pseudo-bispectrum separation of signals with same harmonic frequency components
CN107833581B (en) * 2017-10-20 2021-04-13 广州酷狗计算机科技有限公司 Method, device and readable storage medium for extracting fundamental tone frequency of sound
CN108363109B (en) * 2018-01-12 2019-06-21 吉林大学 The harmonic wave interference fundamental frequency method for fast searching of ground nuclear magnetic resonance data
CN113933590B (en) * 2020-07-14 2024-07-30 森兰信息科技(上海)有限公司 Method, system, medium and device for calculating frequency spectrum of wave
CN111855192B (en) * 2020-07-31 2021-04-23 北京航空航天大学 Singular value decomposition method for denoising encoder signal
CN112307931A (en) * 2020-10-26 2021-02-02 西安电子科技大学 Electromagnetic information leakage rapid detection method based on template matching technology

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102054480A (en) * 2009-10-29 2011-05-11 北京理工大学 Method for separating monaural overlapping speeches based on fractional Fourier transform (FrFT)
CN104036785A (en) * 2013-03-07 2014-09-10 索尼公司 Speech signal processing method, speech signal processing device and speech signal analyzing system
CN104538024A (en) * 2014-12-01 2015-04-22 百度在线网络技术(北京)有限公司 Speech synthesis method, apparatus and equipment

Family Cites Families (16)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
IN177084B (en) * 1991-05-27 1996-11-02 Satake Eng Co Ltd
JP3091343B2 (en) * 1993-03-01 2000-09-25 株式会社河合楽器製作所 Electronic musical instrument
CN1151490C (en) * 2000-09-13 2004-05-26 中国科学院自动化研究所 High-accuracy high-resolution base frequency extracting method for speech recognization
FR2853125A1 (en) * 2003-03-27 2004-10-01 France Telecom METHOD FOR ANALYZING BASIC FREQUENCY INFORMATION AND METHOD AND SYSTEM FOR VOICE CONVERSION USING SUCH ANALYSIS METHOD.
DE102006008260B3 (en) * 2006-02-22 2007-07-05 Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. Device for analysis of audio data, has semitone analysis device to analyze audio data with reference to audibility information allocation over quantity from semitone
JP4630980B2 (en) * 2006-09-04 2011-02-09 独立行政法人産業技術総合研究所 Pitch estimation apparatus, pitch estimation method and program
JP4322283B2 (en) * 2007-02-26 2009-08-26 独立行政法人産業技術総合研究所 Performance determination device and program
US7925502B2 (en) * 2007-03-01 2011-04-12 Microsoft Corporation Pitch model for noise estimation
WO2008123920A1 (en) * 2007-04-10 2008-10-16 Exxonmobil Upstream Research Company Separation and noise removal for multiple vibratory source seismic data
CN101159136A (en) * 2007-11-13 2008-04-09 中国传媒大学 Low bit rate music signal coding method
JP2011065041A (en) * 2009-09-18 2011-03-31 Brother Industries Ltd Basic frequency estimating device, musical notation device and program
US8716586B2 (en) * 2010-04-05 2014-05-06 Etienne Edmond Jacques Thuillier Process and device for synthesis of an audio signal according to the playing of an instrumentalist that is carried out on a vibrating body
EP2638541A1 (en) * 2010-11-10 2013-09-18 Koninklijke Philips Electronics N.V. Method and device for estimating a pattern in a signal
EP2685448B1 (en) * 2012-07-12 2018-09-05 Harman Becker Automotive Systems GmbH Engine sound synthesis
JP2014219607A (en) * 2013-05-09 2014-11-20 ソニー株式会社 Music signal processing apparatus and method, and program
GB201310861D0 (en) * 2013-06-18 2013-07-31 Nokia Corp Audio signal analysis

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102054480A (en) * 2009-10-29 2011-05-11 北京理工大学 Method for separating monaural overlapping speeches based on fractional Fourier transform (FrFT)
CN104036785A (en) * 2013-03-07 2014-09-10 索尼公司 Speech signal processing method, speech signal processing device and speech signal analyzing system
CN104538024A (en) * 2014-12-01 2015-04-22 百度在线网络技术(北京)有限公司 Speech synthesis method, apparatus and equipment

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
《基于常量Q变换的音符起始点检测》;桂文明 刘睿凡 邵曦 白光一;《计算机工程》;20131031;第39卷(第10期);全文
CQ变换的快速算法及在音调频率估计中的误差分析;丁志中;《信息与电子工程》;20051231;第3卷(第4期);全文
基于改进CQT语谱图的单旋律识别法;孔秋强 等;《信息系统工程》;20120531;全文

Also Published As

Publication number Publication date
CN109524023A (en) 2019-03-26
CN109493880A (en) 2019-03-19
CN109410980A (en) 2019-03-01
CN105551501A (en) 2016-05-04

Similar Documents

Publication Publication Date Title
CN105551501B (en) Harmonic signal fundamental frequency estimation algorithm and device
CN103854662B (en) Adaptive voice detection method based on multiple domain Combined estimator
CN111128213B (en) Noise suppression method and system for processing in different frequency bands
CN102054480B (en) Method for separating monaural overlapping speeches based on fractional Fourier transform (FrFT)
CN103594094B (en) Adaptive spectra subtraction real-time voice strengthens
Jang et al. A probabilistic approach to single channel blind signal separation
CN103559888A (en) Speech enhancement method based on non-negative low-rank and sparse matrix decomposition principle
CN102982801A (en) Phonetic feature extracting method for robust voice recognition
TW201108201A (en) Apparatus, method and computer program for obtaining a parameter describing a variation of a signal characteristic of a signal
CN103021405A (en) Voice signal dynamic feature extraction method based on MUSIC and modulation spectrum filter
CN106373559A (en) Robustness feature extraction method based on logarithmic spectrum noise-to-signal weighting
CN111899750B (en) Speech enhancement algorithm combining cochlear speech features and hopping deep neural network
CN115146687B (en) Fault feature extraction method based on second-order variable-scale parameter self-matching stochastic resonance
CN112183225B (en) Underwater target signal feature extraction method based on probability latent semantic analysis
CN110808057A (en) Voice enhancement method for generating confrontation network based on constraint naive
CN112086100A (en) Quantization error entropy based urban noise identification method of multilayer random neural network
Dang et al. Noise reduction using modified phase spectra and Wiener Filter
CN103971697A (en) Speech enhancement method based on non-local mean filtering
Hansson-Sandsten et al. Optimal cepstrum estimation using multiple windows
CN104036781A (en) Voice signal bandwidth expansion device and method
CN112652321B (en) Deep learning phase-based more friendly voice noise reduction system and method
Wang et al. Research on speech enhancement based on deep neural network
CN112820318A (en) Impact sound model establishment and impact sound detection method and system based on GMM-UBM
CN106971733A (en) The method and system and intelligent terminal of Application on Voiceprint Recognition based on voice de-noising
Ma et al. A fast instantaneous frequency estimation for underwater acoustic target feature extraction

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant
TR01 Transfer of patent right

Effective date of registration: 20190611

Address after: Room 507, 5th floor, No. 3 Gaoxin Street, Dalian Hi-tech Industrial Park, Liaoning Province

Patentee after: Dalian Sailing Technology Co., Ltd.

Address before: 116600 No. 18 Liaohe West Road, Jinzhou New District, Dalian City, Liaoning Province

Patentee before: Dalian ethnic university

TR01 Transfer of patent right