CN102231274A - Fundamental tone period estimated value correction method, fundamental tone estimation method and related apparatus - Google Patents

Abstract

The invention provides a fundamental tone period estimated value correction method which comprises the following steps: when a sequence maximum value MAmax(i+1) of a present subframe in a present frequency area MCAMD is larger than a weighted value of an intermediate variable MAmax, substituting the intermediate variable MAmax and an intermediate variable Topt with the MAmax(i+1) and delay of MAmax(i+1) respectively; if a ratio of the intermediate variable Topt and fundamental tone period estimated value Tpre_mid_o of first odd number of subframes of the present subframe is less than correction factor r1 or is greater than correction factor r2 and sequence maximum value MAmax0 of the present subframe in Tpre_mid_o contiguous scope is greater than product of the intermediate variable MAmax and experience factor pho2, correcting the intermediate variable Topt with delay T0 corresponding to the MAmax0; carrying out median filtering on the fundamental tone period estimated value of first even number of subframe of the present subframe and the intermediate variable Topt.

Description

Pitch period estimated value modification method, pitch estimation method and relevant apparatus

Technical field

The present invention relates to the signal Processing field, relate in particular to pitch period estimated value modification method, pitch estimation method and relevant apparatus.

Background technology

In field of voice signal, the original implication of fundamental tone is meant the caused periodic feature of vocal cord vibration when sending out voiced sound, and pitch period is the inverse of vibration frequency of vocal band.In the Audio Signal Processing field, fundamental tone also has similar implication.From the time domain angle, the obvious characteristic of periodic signal is the similarity of waveform.The cardinal principle that relies on the homophylic Pitch Detection Algorithm of waveform is to determine pitch period by the similarity between the signal after comparison original signal and its displacement.If translocation distance equals pitch period, so, two signals have maximum similarity (perhaps simple crosscorrelation maximum).No matter be field of voice signal or Audio Signal Processing field, the detection of fundamental tone or the estimation of pitch period all are very important technology, because, by pitch Detection and estimation, extract the fundamental frequency of signal, just can know the variation speed of signal, thereby understand the feature of signal, for further signal Processing provides necessary reference.

Pitch Detection and algorithm for estimating are more, usually be divided into time domain approach and frequency domain method two big classes, wherein, time domain approach mainly comprises autocorrelation function method (Autocorrelation Function, ACF), average magnitude difference function method (Average Magnitude Difference Function, AMDF) and simple inverse filter tracking method (Simple Inverse Filtering Tracking, SIFT) etc., frequency domain method comprises that mainly harmonic wave amasss spectrometry (Harmonic Product Spectrum Method) and cepstrum method (Ceptrum Method, CM) consider from the computation complexity aspect, correlation method is compared with additive method, on the low and performance of calculated amount and additive method similar.

In numerous pitch Detection and algorithm for estimating, a kind of method that prior art provides is based on the pitch estimation method of Normalized Cross Correlation Function.In the method, at first the signal to input carries out pre-service, comprises high-pass filtering, low-pass filtering, goes average, numerical value is level and smooth etc.; Then, to through its normalized crosscorrelation sequence of pretreated calculated signals, obtain simple crosscorrelation sequence R (i), the delay of the maximal value correspondence of the R that calculates (i) can be used as the pitch period estimated value; At last, use median filter that the pitch period estimated value is carried out smoothly, level and smooth purpose is mainly in order to remove " wild point " in the fundamental tone estimated value, i.e. frequency multiplication/half frequency mistake.

In the above-mentioned prior art, the Normalized Cross Correlation Function method is to compare by current demand signal and past signal, and peak value can appear in the waveform similarity then Normalized Cross Correlation Function of signal.If the pitch period of signal is less, and have preferably periodically, then the simple crosscorrelation sequence can comprise a plurality of cycles,, a plurality of maximum value can occur in the simple crosscorrelation sequence that is.Because influence of harmonic, in these maximum value, maximal value often is not the maximum value of first periodic point correspondence, therefore, half frequency mistake appears in Normalized Cross Correlation Function method easily, thereby causes median filter that the pitch period estimated value is carried out level and smooth poor effect.

Summary of the invention

The embodiment of the invention provides pitch period estimated value modification method, pitch estimation method and relevant apparatus, is used to solve prior art is carried out poor effect when level and smooth to the pitch period estimated value problem.

The embodiment of the invention provides a kind of pitch period estimated value modification method, comprising: more current subframe is revised circulation average amplitude difference sequence maximal value MA in current frequency field _Max(i+1) with the first intermediate variable MA _MaxWeighted value, if described MA _Max(i+1) greater than described MA _MaxWeighted value, then respectively with described MA _Max(i+1) and described MA _Max(i+1) Dui Ying delay T (i+1) substitutes the first intermediate variable MA _MaxWith the second intermediate variable T _Opt, repeat described comparison procedure, until described current frequency field not within the fundamental frequency scope;

Calculate the described second intermediate variable T _OptEstimate intermediate value T with the pitch period of odd number subframe before the described current subframe _{Pre_mid_o}Ratio, if described ratio is less than first correction factor or greater than second correction factor and described T _{Pre_mid_o}The correction circulation average amplitude difference sequence maximal value MA of described current subframe in the nearby sphere _Max0Greater than the described first intermediate variable MA _MaxWith the product of the second experience factor, then use described MA _Max0Corresponding delay T ₀Proofread and correct the second intermediate variable T _Opt

The pitch period estimated value and the second intermediate variable T with even number subframe before the described current subframe _OptCarry out medium filtering, with the value of carrying out obtaining behind the described medium filtering pitch period estimated value as middle subframe;

The described first intermediate variable MA _MaxWeighted value be described MA _MaxWith the product of the first experience factor, describedly use described MA _Max0Corresponding delay T ₀Proofread and correct the second intermediate variable T _OptBe with the second intermediate variable T _OptCurrency be set to described MA _Max0Corresponding delay T ₀, described middle subframe is that some subframe neutron frame numbers that preceding even number subframe of described current subframe and described current subframe constitute are the subframe of intermediate value, described even number is the maximum even number less than described odd number.

The embodiment of the invention provides a kind of pitch period estimated value correcting device, comprising: comparison module is used for more current subframe and revises circulation average amplitude difference sequence maximal value MA in current frequency field _Max(i+1) with the first intermediate variable MA _MaxWeighted value, if described MA _Max(i+1) greater than described MA _MaxWeighted value, then respectively with described MA _Max(i+1) and described MA _Max(i+1) Dui Ying delay T (1) substitutes the first intermediate variable MA _MaxWith the second intermediate variable T _Opt, repeat described comparison procedure, until described current frequency field not within the fundamental frequency scope;

Correction module is used to calculate the described second intermediate variable T _OptEstimate intermediate value T with the pitch period of odd number subframe before the described current subframe _{Pre_mid_o}Ratio, if described ratio is less than first correction factor or greater than second correction factor and described T _{Pre_mid_o}The correction circulation average amplitude difference sequence maximal value MA of described current subframe in the nearby sphere _Max0Greater than the described first intermediate variable MA _MaxWith the product of the second experience factor, then use described MA _Max0Corresponding delay T ₀Proofread and correct the second intermediate variable T _Opt

The medium filtering module is used for the pitch period estimated value and the second intermediate variable T with even number subframe before the described current subframe _OptCarry out medium filtering, with the pitch period estimated value of carrying out obtaining behind the described medium filtering pitch period estimated value as middle subframe;

The described first intermediate variable MA _MaxWeighted value be described MA _MaxWith the product of the first experience factor, describedly use described MA _Max0Corresponding delay T ₀Proofread and correct the second intermediate variable T _OptBe with the second intermediate variable T _OptCurrency be set to described MA _Max0Corresponding delay T ₀, some subframe neutron frame numbers that preceding even number subframe that described middle subframe is described current subframe and described current subframe constitute are the subframe of intermediate value, described even number is the maximum even number less than described odd number.

The embodiment of the invention provides a kind of pitch estimation method, comprising: the signal that receives is carried out pre-service, and described signal comprises voice signal or sound signal;

To its normalized crosscorrelation sequence of the pretreated calculated signals of described process, ask for the correction circulation average amplitude difference sequence of described normalized crosscorrelation sequence according to the described normalized crosscorrelation sequence that obtains;

According to the delay of the maximal value correspondence of described correction circulation average amplitude difference sequence in the fundamental frequency scope, the pitch delay estimated value is revised, be the pitch period estimated value of described signal with the delay estimated value of described correction gained.

The embodiment of the invention provides a kind of fundamental tone estimation unit, comprising: pretreatment module, be used for the signal that receives is carried out pre-service, and described signal comprises voice signal or sound signal;

Sequence is asked for module, is used for its normalized crosscorrelation sequence of the pretreated calculated signals of described process, asks for the correction circulation average amplitude difference sequence of described normalized crosscorrelation sequence according to the described normalized crosscorrelation sequence that obtains;

Correcting module is used for the delay according to the maximal value correspondence of described correction circulation average amplitude difference sequence in the fundamental frequency scope, the pitch delay estimated value revised, and be the pitch period estimated value of described signal with the delay estimated value of described correction gained.

The embodiment that provides from the invention described above as can be known, owing to revising circulation average amplitude difference sequence is to derive and get from having good periodic normalized crosscorrelation sequence, handle and revise circulation average amplitude difference sequence maximal value and length of delay thereof, by recycle ratio than revising circulation average amplitude difference sequence maximal value in the side frequency zone, final correction circulation average amplitude difference sequence maximal value and the delay thereof of determining current subframe, can reduce the probability that occurs half frequency/frequency multiplication mistake when the normalized crosscorrelation sequence algorithm is estimated pitch period, and the pitch period estimated value of even number subframe before current subframe correction circulation peaked delay of average amplitude difference sequence and the current subframe is carried out medium filtering, half frequency/frequency multiplication the mistake that occurs in the time of then can further reducing the estimation pitch period has been equivalent to further improve the accuracy that pitch period is estimated.

Description of drawings

In order to be illustrated more clearly in the technical scheme of the embodiment of the invention, to do to introduce simply to the accompanying drawing of required use in prior art or the embodiment description below, apparently, accompanying drawing in describing below only is some embodiments of the present invention, for those of ordinary skills, under the prerequisite of not paying creative work, can also obtain other accompanying drawing as these accompanying drawings.

Fig. 1 is the pitch period estimated value modification method schematic flow sheet that the embodiment of the invention provides;

Fig. 2 is the pitch period estimated value modification method schematic flow sheet that another embodiment of the present invention provides;

Fig. 3 is the pitch estimation method schematic flow sheet that the embodiment of the invention provides;

Fig. 4 is a kind of pitch period estimated value correcting device logical organization synoptic diagram that the embodiment of the invention provides;

Fig. 5 is a kind of fundamental tone estimation unit logical organization synoptic diagram that the embodiment of the invention provides;

Fig. 6 is a kind of fundamental tone estimation unit logical organization synoptic diagram that another embodiment of the present invention provides;

Fig. 7 is a kind of fundamental tone estimation unit logical organization synoptic diagram that another embodiment of the present invention provides;

Fig. 8 is the structural representation that the fundamental tone estimation unit is used for the regular technology modules of time of speech audio Unified coding;

Fig. 9 is the voice signal fundamental tone estimating system logical organization synoptic diagram that the embodiment of the invention provides.

Embodiment

Below in conjunction with the accompanying drawing in the embodiment of the invention, the technical scheme in the embodiment of the invention is clearly and completely described, obviously, described embodiment only is the present invention's part embodiment, rather than whole embodiment.Based on the embodiment among the present invention, those of ordinary skills belong to the scope of protection of the invention not making the every other embodiment that is obtained under the creative work prerequisite.

Seeing also accompanying drawing 1, is the pitch period estimated value modification method schematic flow sheet that the embodiment of the invention provides, and mainly comprises step:

S101, more current subframe is revised circulation average amplitude difference sequence maximal value MA in current frequency field _Max(i+1) with the first intermediate variable MA _MaxWeighted value, if described MA _Max(i+1) greater than described MA _MaxWeighted value, then respectively with described MA _Max(i+1) and described MA _Max(i+1) Dui Ying delay T (i+1) substitutes the first intermediate variable MA _MaxWith the second intermediate variable T _Opt, repeat described comparison procedure, until described current frequency field not within the fundamental frequency scope.

Considering can be with a frame signal, and for example, voice signal or sound signal are divided into 16 subframes, therefore, in embodiments of the present invention, can be to be unit with a subframe, carries out the correction of pitch period estimated value.Further, because the simple crosscorrelation sequence has good period, therefore, in embodiments of the present invention, can conversion of signals be become the normalized crosscorrelation sequence by the normalized crosscorrelation sequence of signal calculated, and then ask for and revise circulation average amplitude difference sequence maximal value.So, in fundamental tone is estimated, can consider the zero-time of original signal, but can utilize first peaked position of normalized crosscorrelation sequence to estimate pitch period.

For voice signal or sound signal, the fundamental frequency scope and can be concentrated between 80Hz～2500Hz by the fundamental frequency scope that people's ear is felt generally between 60Hz～4000Hz.For pitch period is estimated, can be in advance with several frequency fields of fundamental frequency scope division.

In an embodiment provided by the invention, with the fundamental frequency scope division of 80Hz～2500Hz become the first frequency zone [80,240) Hz, second frequency zone [240,720) Hz and the 3rd frequency field [720,2500) Hz.Then, more current subframe is revised circulation average amplitude poor (MCAMD, Modified Circular Average Magnitude Difference) sequence maximal value MA in current frequency field _Max(i+1) with the first intermediate variable MA _MaxWeighted value, for example, more current subframe in the second frequency zone [240,720) interior MCAMD sequence maximal value MA _Max(2) with the first intermediate variable MA _MaxWeighted value and more current subframe the 3rd frequency field [720,2500) interior MCAMD sequence maximal value MA _Max(3) with the first intermediate variable MA _MaxWeighted value.

Because what fundamental tone was estimated the easiest appearance is frequency multiplication/half frequently wrong (being " the wild point " in the fundamental tone estimated value), this mistake just appears for fear of at the beginning the time, should make two frequency sizes of adjacent MCAMD sequence maximal value correspondence be twice or half times of relation as far as possible, promptly frequency multiplication/half concerns frequently.For example, when the frequency of a MCAMD sequence maximal value correspondence is 100Hz, then should make the frequency of adjacent MCAMD sequence maximal value correspondence is 200Hz or 50Hz.In embodiments of the present invention, when the fundamental frequency scope with 80Hz～4000Hz is divided into some frequency fields in the following manner: two end values of each frequency field are the multiple relation, frequency field is continuous, and promptly the left end value of the right-hand member value of previous frequency field (promptly in two end values of this frequency field bigger) and a back frequency field (promptly in two end values of this frequency field less) is overlapping.For example, between fundamental frequency scope 80Hz～4000Hz, dividing the first frequency zone is [80Hz, 160Hz), the second frequency zone is [160Hz, 320Hz), the 3rd frequency field is [320Hz, 640Hz), the 4th frequency field be [640Hz, 1280Hz) and the 5th frequency field be [1280Hz, 4000Hz).So, adjacent MCAMD sequence maximal value is bigger in the possibility that two frequency sizes of adjacent frequency field correspondence are twice or half times of relation, thereby makes follow-up correction to the fundamental tone estimated value more accurate.

If current subframe is MCAMD sequence maximal value MA in current frequency field _Max(i+1) greater than the first intermediate variable MA _MaxWeighted value, then with current subframe MCAMD sequence maximal value MA in current frequency field _Max(i+1) substitute the first intermediate variable MA _Max, with current subframe MCAMD sequence maximal value MA in current frequency field _Max(i+1) Dui Ying delay T (i+1) substitutes the second intermediate variable T _Opt, and repeat above-mentioned comparison procedure, until current frequency field not within the fundamental frequency scope.For example, initial at algorithm, set the first intermediate variable MA _MaxFor current subframe in the first frequency zone [80Hz, 160Hz) in MCAMD sequence maximal value MA _Max(1), sets the second intermediate variable T _OptBe MA _Max(1) Dui Ying delay T (1).The first experience factor is ρ ₁, suppose current subframe in the second frequency zone [160Hz, 320Hz) in MCAMD sequence maximal value MA _Max(2) greater than MA _MaxWith ρ ₁Product, i.e. MA _Max(2)＞MA _Max* ρ ₁, then make MA _Max=MA _Max(2), T _Opt=T (2).Continue more current subframe at the 3rd frequency field [320Hz, 640Hz) interior MCAMD sequence maximal value MA _Max(3) and MA _Max* ρ ₁, if MA _Max(3)＜MA _Max* ρ ₁, then keep MA _MaxAnd T _OptConstant.Continuing relatively, the 4th frequency field is [640Hz, 1280Hz] interior MCAMD sequence maximal value MA _Max(4) and MA _Max* ρ ₁, if MA _Max(4)＞MA _Max* ρ ₁, then make MA _Max=MA _Max(4), T _Opt=T (4) ..., after this continue relatively, until finishing current subframe at the 5th frequency field [1280Hz, 4000Hz) interior MCAMD sequence maximal value MA _Max(5) and current MA _MaxWith ρ ₁The comparison of product.After relatively finishing, the second intermediate variable T _OptBe endowed a value, for example, current subframe is at the 4th frequency field [640Hz, 1280Hz) interior MCAMD sequence maximal value MA _Max(4) Dui Ying delay T (4).

In embodiments of the present invention, the first experience factor ρ ₁Can get 0.95.

S102 calculates the described second intermediate variable T _OptEstimate intermediate value T with the pitch period of odd number subframe before the described current subframe _{Pre_mid_o}Ratio, if described ratio is less than first correction factor or greater than second correction factor and described T _{Pre_mid_o}The correction circulation average amplitude difference sequence maximal value MA of described current subframe in the nearby sphere _Max0Greater than the described first intermediate variable MA _MaxWith the product of the second experience factor, then use described MA _Max0Corresponding delay T ₀Proofread and correct the second intermediate variable T _Opt

If T _OptBe endowed a value, for example, T _Opt=MA _Max(4), then calculate T _Opt/ T _{Pre_mid_o}Here, T _{Pre_mid_o}The pitch period that is odd number subframe before the current subframe is estimated intermediate value.In embodiment provided by the invention, T _{Pre_mid_o}Get the pitch period of preceding 5 subframes of current subframe and estimate that intermediate value can obtain estimated result preferably.For example, suppose that current subframe is subframe F ₈(F ₈The numbering of subscript " 8 " expression subframe, down with), note subframe F ₈Preceding 5 subframes are followed successively by subframe F ₇, subframe F ₆, subframe F ₅, subframe F ₄With subframe F ₃, subframe F then ₈The pitch period of preceding 5 subframes estimates that intermediate value is subframe F ₇, subframe F ₆, subframe F ₅, subframe F ₄With subframe F ₃The pitch period estimated value through obtaining behind the medium filtering, also be the pitch period estimated value that is numbered the subframe of intermediate value in preceding 5 subframes of current subframe, i.e. subframe F ₅The pitch period estimated value.

If described ratio is less than first correction factor or greater than second correction factor and described T _{Pre_mid_o}The correction circulation average amplitude difference sequence maximal value MA of described current subframe in the nearby sphere _Max0Greater than the described first intermediate variable MA _MaxWith the product of the second experience factor, then use described MA _Max0Corresponding delay T ₀Proofread and correct the second intermediate variable T _Opt

Remember that first correction factor, second correction factor and the second experience factor are respectively r ₁, r ₂And ρ ₂, if the second intermediate variable T _OptEstimate intermediate value T with the pitch period of odd number subframe before the current subframe _{Pre_mid_o}Ratio less than first correction factor or greater than second correction factor, that is, if T _Opt/ T _{Pre_mid_o}＜r ₁Or T _Opt/ T _{Pre_mid_o}＞r ₂, then at T _{Pre_mid_o}In the nearby sphere, for example, at T _{Pre_mid_o}Seek the MCAMD sequence maximal value MA of current subframe in contiguous 1ms (millisecond) scope _Max0If T _{Pre_mid_o}The correction circulation average amplitude difference sequence maximal value MA of current subframe in the nearby sphere _Max0Greater than the first intermediate variable MA _MaxWith the product of the second experience factor, that is, and MA _Max0＞MA _Max* ρ ₂, the second intermediate variable T then _OptCurrency be set to T _{Pre_mid_o}The correction circulation average amplitude difference sequence maximal value MA of described current subframe in the nearby sphere _Max0Corresponding delay T ₀

S103 is with the pitch period estimated value and the second intermediate variable T of even number subframe before the described current subframe _OptCarry out medium filtering together, with the value of carrying out obtaining behind the described medium filtering pitch period estimated value as middle subframe.

Middle subframe is that some subframe neutron frame numbers that preceding even number subframe of current subframe and current subframe constitute are the subframe of intermediate value, and this even number is the maximum even number less than the odd number among the step S102.The pitch period estimated value and the second intermediate variable T with even number subframe before the current subframe _OptCarry out medium filtering together, with the pitch period estimated value of carrying out obtaining behind the described medium filtering pitch period estimated value as middle subframe.

For example, suppose that current subframe is subframe F ₈, subframe F ₈Preceding 4 subframes are F ₇, subframe F ₆, subframe F ₅With subframe F ₄The pitch period estimated value of each subframe is respectively T ₇, T ₆, T ₅And T ₄, then with T _Opt, T ₇, T ₆, T ₅And T ₄Carry out medium filtering, with the pitch period estimated value of carrying out obtaining behind the described medium filtering pitch period estimated value, promptly as subframe F as middle subframe ₆The pitch period estimated value.

In embodiments of the present invention, the first correction factor r ₁, the second correction factor r ₂With the second experience factor ρ ₂Can get 0.75,1.4 and 0.85 respectively.

The embodiment that provides from the invention described above as can be known, owing to revising circulation average amplitude difference sequence is to derive and get from having good periodic normalized crosscorrelation sequence, handle to revise circulation average amplitude difference sequence maximal value and length of delay thereof, by recycle ratio than revising circulation average amplitude difference sequence maximal value and default MA in each fundamental frequency zone _MaxWeighted value, final correction circulation average amplitude difference sequence maximal value and the delay thereof of determining current subframe, can reduce the probability that occurs half frequency/frequency multiplication mistake when the normalized crosscorrelation sequence algorithm is estimated pitch period, and the pitch delay estimated value of the correction that obtains with current subframe correction circulation average amplitude difference sequence, carry out medium filtering with the pitch period estimated value of even number subframe before the current subframe, half frequency/frequency multiplication the mistake that occurs in the time of then can further reducing the estimation pitch period has been equivalent to further improve the accuracy that pitch period is estimated.

Below in conjunction with the flow process of accompanying drawing 2, the technical scheme when not satisfying some assumed condition in the foregoing description is described, for example, suppose among the step S101 that current subframe is MCAMD sequence maximal value MA in current frequency field _Max(i+1) and MA _MaxThe comparative result of weighted value be MA _Max(i+1) less than MA _Max* ρ ₁

Among accompanying drawing 2 embodiment, being that 5 frequency fields are the example explanation with the fundamental frequency scope division.Those skilled in the art can understand, when the fundamental frequency scope division is the frequency field of other numbers, 3 frequency fields for example, the flow process of processing is similar, the treatment scheme when therefore repeating no more the fundamental frequency scope division to the frequency field of other numbers.

S201 makes MA _Max=MA _Max(1), T _Opt=T (1), i=2.The MCAMD sequence maximal value and corresponding delay of this maximal value that are about to the first frequency zone compose to the first intermediate variable MA as preset initial value _MaxWith the second intermediate variable T _Opt, move forward a frequency field (i=2), promptly move to the second frequency zone.

S202, MA _Max(i)＞MA _Max* ρ ₁Promptly judge current subframe in new frequency field MCAMD sequence maximal value whether greater than the first experience factor and the MA _MaxProduct, if greater than, then flow process enters step S203, otherwise flow process enters step S204.

S203 makes MA _Max=MA _Max(i), T _Opt=T (i).Promptly when satisfying step S202 condition, the first intermediate variable MA _MaxWith the second intermediate variable T _OptUse the maximal value of current subframe MCAMD sequence in new frequency field respectively and postpone replacement.

S204 makes i=i+1.Promptly move forward a frequency field.

S205, i＞5? judge that promptly current frequency field is whether within the fundamental frequency scope, whether perhaps current frequency field is in 5 frequency fields dividing in the fundamental frequency scope, if go beyond the scope (being i＞5), then flow process enters step S206, otherwise the repetition said process is promptly got back to step S202.

S206 makes r=T _Opt/ T _{Pre_mid_o}Promptly calculate the second intermediate variable T _OptEstimate intermediate value T with the pitch period of preceding 5 subframes of current subframe _{Pre_mid_o}Ratio.

S207, r＜r ₁Or r＞r ₂Promptly judge the second intermediate variable T _OptEstimate intermediate value T with the pitch period of preceding 5 subframes of current subframe _{Pre_mid_o}Ratio whether less than the first correction factor r ₁Or not greater than the second correction factor r ₂If then enter step S208, otherwise change step S211 over to.

In the present embodiment, r ₁And r ₂Can get 0.75 and 1.4 respectively.

S208 is at T _{Pre_mid_o}Ask for the MCAMD sequence maximal value MA of current subframe in the nearby sphere _Max0And delay T ₀For example, at T _{Pre_mid_o}Ask for the MCAMD sequence maximal value MA of current subframe in 1ms (millisecond) scope _Max0And delay T ₀

S209, MA _Max0＞MA _Max* ρ ₂Promptly judge T _{Pre_mid_o}The MCAMD sequence maximal value MA of current subframe in the nearby sphere _Max0Whether greater than the first intermediate variable MA _MaxWith the second experience factor ρ ₂Product.If, then enter step S210, otherwise, change step S211 over to.

S210 makes T _Opt=T ₀Promptly use T _{Pre_mid_o}The MCAMD sequence maximal value MA of current subframe in the nearby sphere _Max0Corresponding delay T ₀Proofread and correct the second intermediate variable T _Opt

S211 is with the pitch period estimated value and the second intermediate variable T of even number subframe before the current subframe _OptCarry out medium filtering together.Carry out the pitch period estimated value that obtains behind the medium filtering pitch period estimated value as middle subframe.The middle subframe here is that some subframe neutron frame numbers that preceding even number subframe of current subframe and current subframe constitute are the subframe of intermediate value, for example, supposes that current subframe is subframe F ₈, subframe F ₈Preceding 4 subframes are F ₇, subframe F ₆, subframe F ₅With subframe F ₄, then middle subframe is subframe F ₆

Seeing also accompanying drawing 3, is the pitch estimation method schematic flow sheet that the embodiment of the invention provides, and mainly comprises step:

S301 carries out pre-service to the signal that receives, and described signal comprises voice signal or sound signal.

Because real signal; for example; voice signal or sound signal are usually mixing ground unrest, harmonic wave and formant frequency or the like; these ground unrests and frequency component make the waveform of signal become very complicated; this usually causes the erroneous judgement of pitch Detection or estimation; for carrying out removing these unfavorable factors as far as possible before fundamental tone is estimated, in embodiment provided by the invention, at first the signal that receives is carried out pre-service.Pre-service comprises to be judged the quiet frame of signal and non-quiet frame.For example, owing to the feature of signals such as voice or audio frequency changed along with the time, only a period of time at interval in, these signals just keep relative stability (steadily), this specific character of signal is called " short-time characteristic ".Analysis and processing to voice or sound signal generally are based upon on " short-time characteristic " basis, promptly adopt segmentation or branch frame to handle to voice or audio signal stream.Dividing frame generally to adopt movably the finite length window that voice or sound signal are weighted realizes.Divide frame both can adopt continuation mode, also can adopt the method for overlapping segmentation to realize.Therefore, in embodiment provided by the invention, can by adopt short-time average energy (short-time average energy of window weighting be equivalent to " square " output of signal by a linear filter) come the quiet/non-quiet frame of signal is judged, promptly when the short-time energy of signal during less than certain threshold value, be judged as quiet frame, otherwise be non-quiet frame.

Because cycle of quiet frame is not obvious or do not have a cycle, it is bigger that the pitch period of quiet frame is often estimated that pitch period with reality departs from, thus the accuracy that influence was estimated the whole signal pitch cycle.In view of the foregoing, in embodiments of the present invention,, do not carry out the estimation and the last handling process of pitch period for the quiet frame of signal.In addition, in the pitch estimation method that the embodiment of the invention provides, in order to be applicable to regular technology of time, the pitch period of quiet frame is made as the fundamental tone intermediate value rather than zero of last subframe, its reason is, if the pitch period of quiet frame is made as zero, then the torsion resistance of extracting by the fundamental tone estimated value (regular technology desired parameters of time) can be very big at non-quiet frame and quiet frame linking point place, the distortion that leads to errors, and then the regular technical feature of time of the pitch estimation method that the embodiment of the invention provides is used in influence.

Before the quiet frame of signal and non-quiet frame were judged, pre-service also comprised high-pass filtering, removes average and numerical filter.For example, can adopt lower frequency limit is that the Hi-pass filter of 50Hz carries out filtering to signal, to remove the interference of power supply signal.Consider when signal has Non-zero Mean that the normalized crosscorrelation sequence all has higher numerical value in all delays, the pre-service that the embodiment of the invention provides also comprises the process of average, and removing average signal afterwards is s ' (+)=s (n)-u, herein,

For signal being carried out smoothly, the pre-service that the embodiment of the invention provides comprises that also the wave filter that adopts certain exponent number carries out numerical filter to signal.Facts have proved that when signal was carried out numerical filter, it is undesirable that exponent number is lower than the filter effect on 5 rank, the filter effect that is higher than 5 rank promotes obvious inadequately, also brings bigger delay.In order not cause delay and to obtain filter effect preferably, in embodiment provided by the invention, can adopt 5 exponent number value filtering devices that signal is carried out numerical filter, the numerical filter formula of 5 exponent number value filtering devices is

S302 to its normalized crosscorrelation sequence of the pretreated calculated signals of described process, asks for the correction circulation average amplitude difference sequence of described normalized crosscorrelation sequence according to the described normalized crosscorrelation sequence that obtains.

Because the simple crosscorrelation sequence has good period, therefore, in embodiments of the present invention, can will become the normalized crosscorrelation sequence through pretreated conversion of signals by calculating Normalized Cross Correlation Function.For example, by Normalized Cross Correlation Function (NCCF, Normalized Cross-Correlation Function) conversion of signals is become normalized crosscorrelation sequence R (j):

$R\left(j\right)=\frac{\underset{n=0}{\overset{N-1}{\mathrm{\Σ}}}\tilde{s}\left(n\right)\tilde{s}(n-j)}{\sqrt{\underset{n=0}{\overset{N-1}{\mathrm{\Σ}}}{\tilde{s}}^2\left(n\right)\underset{n=0}{\overset{N-1}{\mathrm{\Σ}}}{\tilde{s}}^2(n-j)}},j=0,K,M$

Wherein, elongatedness when M is simple crosscorrelation, N are the simple crosscorrelation sequence length.

Afterwards, to the normalized crosscorrelation (NCC that obtains, Normalized Cross-Correlation) sequence, calculate its circulation average magnitude difference function (CAMDF, Circular Average Magnitude Difference Function), further convert signal to circulation average amplitude difference sequence A (j):

$A\left(j\right)=\frac{1}{M}\underset{n=0}{\overset{M-1}{\mathrm{\Σ}}}|R\left((n+j)\mathrm{mod}M\right)-R\left(n\right)|,j=0,K,\frac{M}{2}$

Further, to the poor (CAMD of circulation average amplitude, Circular Average Magnitude Difference) sequence is revised, ask for correction circulation average amplitude poor (MCAMD, Modified Circular Average Magnitude Difference) the sequence M (j) of NCC sequence:

$\mathrm{MA}\left(j\right)=A_{\max }-A\left(j\right),j=0,K,\frac{M}{2},$ Herein, A _MaxMaximal value for A (j) sequence.

The delay of the maximal value correspondence of MA (j) is the possible pitch period estimated value of a subframe (analyzing if signal is divided into some subframes).

As previously mentioned, because cycle of quiet frame is not obvious or do not have a cycle, it is bigger that the pitch period of quiet frame is often estimated that pitch period with reality departs from, thus the accuracy that influence was estimated the whole signal pitch cycle.Therefore, in embodiments of the present invention, to its normalized crosscorrelation sequence of the pretreated calculated signals of described process, the correction circulation average amplitude difference sequence of asking for described normalized crosscorrelation sequence according to the normalized crosscorrelation sequence that obtains specifically comprises: to calculating its normalized crosscorrelation sequence through pretreated non-quiet frame signal, ask for the correction circulation average amplitude difference sequence of the normalized crosscorrelation sequence of this non-quiet frame signal according to the normalized crosscorrelation sequence of the non-quiet frame signal that obtains.

S303 according to the delay of the maximal value correspondence of described correction circulation average amplitude difference sequence in the fundamental frequency scope, revises the pitch delay estimated value, is the pitch period estimated value of described signal with the delay estimated value of described correction gained.

Note MCAMD sequence maximal value is MA _Max(i), then in embodiments of the present invention,, the pitch delay estimated value revised comprise step S3031, S3032 and S3033 according to the delay of revising the maximal value correspondence of circulation average amplitude difference sequence in the fundamental frequency scope:

S3031, more current subframe is revised circulation average amplitude difference sequence maximal value MA in current frequency field _Max(i+1) with the first intermediate variable MA _MaxWeighted value, if described MA _Max(i+1) greater than described MA _MaxWeighted value, then respectively with described MA _Max(i+1) and described MA _Max(i+1) Dui Ying delay T (i+1) substitutes the first intermediate variable MA _MaxWith the second intermediate variable T _Opt, repeat described comparison procedure, until described current frequency field not within the fundamental frequency scope.

The subframe here can be a frame signal that has been divided into 16 subframes, for example, voice signal or sound signal, any one subframe in its 16 subframes, frequency field is some frequency separations of dividing in the fundamental frequency scope.For example, five frequency fields that are divided in the fundamental frequency scope between 80Hz～4000Hz: first frequency zone [80,160) Hz, second frequency zone [160,320) Hz, the 3rd frequency field [320,640) Hz, the 4th frequency field [640,1280) Hz and the 5th frequency field [1280,4000) Hz.

Step S3031 more detailed description can be consulted the related description of the step S101 of accompanying drawing 1 example, does not do herein and gives unnecessary details.

S3032 calculates the second intermediate variable T _OptEstimate intermediate value T with the pitch period of odd number subframe before the current subframe _{Pre_mid_o}Ratio.If the second intermediate variable T _OptEstimate intermediate value T with the pitch period of odd number subframe before the current subframe _{Pre_mid_o}Ratio less than first correction factor or greater than second correction factor and described T _{Pre_mid_o}The correction circulation average amplitude difference sequence maximal value MA of described current subframe in the nearby sphere _Max0Greater than the described first intermediate variable MA _MaxWith the product of the second experience factor, then use described MA _Max0Corresponding delay T ₀Proofread and correct the second intermediate variable T _Opt

If T _OptBe endowed a value, for example, T _Opt=MA _Max(4), then calculate T _Opt/ T _{Pre_mid_o}Here, T _{Pre_mid_o}The pitch period that is odd number subframe before the current subframe is estimated intermediate value.In embodiment provided by the invention, T _{Pre_mid_o}Get the pitch period of preceding 5 subframes of current subframe and estimate that intermediate value can obtain estimated result preferably.For example, suppose that current subframe is subframe F ₈(F ₈The numbering of subscript " 8 " expression subframe, down with), note subframe F ₈Preceding 5 subframes are followed successively by subframe F ₇, subframe F ₆, subframe F ₅, subframe F ₄With subframe F ₃, subframe F then ₈The pitch period of preceding 5 subframes estimates that intermediate value is subframe F ₇, subframe F ₆, subframe F ₅, subframe F ₄With subframe F ₃The pitch period estimated value through obtaining behind the medium filtering, also be the pitch period estimated value that is numbered the subframe of intermediate value in preceding 5 subframes of current subframe, i.e. subframe F ₅The pitch period estimated value.

At present embodiment, the first correction factor r ₁, the second correction factor r ₂With the second experience factor ρ ₂Can get 0.75,1.4 and 0.85 respectively.

S3033 is with the pitch period estimated value and the second intermediate variable T of even number subframe before the described current subframe _OptCarry out medium filtering together, with the value of carrying out obtaining behind the described medium filtering pitch period estimated value as middle subframe.

Step S3033 more detailed description can be consulted the related description of the step S103 of accompanying drawing 1 example, does not do herein and gives unnecessary details.

From the example of accompanying drawing 3 as can be known, the present invention is a correction circulation average amplitude difference sequence of asking for the normalized crosscorrelation sequence according to the normalized crosscorrelation sequence that obtains, and then according to the delay of the maximal value correspondence of described correction circulation average amplitude difference sequence in the fundamental frequency scope, the pitch delay estimated value is revised, and is the pitch period estimated value of described signal with the delay estimated value of described correction gained.In general, the fundamental tone algorithm for estimating that the embodiment of the invention provides based on normalized crosscorrelation weighting correction circulation average magnitude difference function, compare with the time domain fundamental tone algorithm for estimating of prior art and to have low time delay, compare with frequency domain algorithm on the low and performance of computation complexity and be more or less the same.On the other hand, owing to revising circulation average amplitude difference sequence is to derive and get from having good periodic normalized crosscorrelation sequence, therefore can reduce the probability that occurs half frequency/frequency multiplication mistake when the normalized crosscorrelation sequence algorithm is estimated pitch period, and according to the delay of the maximal value correspondence of described correction circulation average amplitude difference sequence in the fundamental frequency scope, the pitch delay estimated value (is for example revised, the pitch delay estimated value of the correction that obtains with current subframe correction circulation average amplitude difference sequence, carry out medium filtering with the pitch period estimated value of even number subframe before the current subframe) then can further reduce the half frequency/frequency multiplication mistake that occurs when estimating pitch period, be equivalent to further improve the accuracy that pitch period is estimated.

Seeing also accompanying drawing 4, is a kind of pitch period estimated value correcting device logical organization synoptic diagram that the embodiment of the invention provides.For convenience of explanation, only show the part relevant with the embodiment of the invention.Functional module/unit that the pitch period estimated value correcting device of accompanying drawing 4 examples comprises can be software module/unit, hardware module/unit or the software and hardware module/unit that combines, comprise comparison module 401, correction module 402 and medium filtering module 403, wherein:

Comparison module 401 is used for more current subframe and revises circulation average amplitude difference sequence maximal value MA in current frequency field _Max(i+1) with the first intermediate variable MA _MaxWeighted value, if described MA _Max(i+1) greater than described MA _MaxWeighted value, then respectively with described MA _Max(i+1) and described MA _Max(i+1) Dui Ying delay T (i+1) substitutes the first intermediate variable MA _MaxWith the second intermediate variable T _Opt, repeat described comparison procedure, until described current frequency field not within the fundamental frequency scope, the described first intermediate variable MA _MaxWeighted value be described MA _MaxWith the product of the first experience factor, described frequency field is some frequency separations of dividing in described fundamental frequency scope;

Correction module 402 is used to calculate the described second intermediate variable T _OptEstimate intermediate value T with the pitch period of odd number subframe before the described current subframe _{Pre_mid_o}Ratio, if described ratio is less than first correction factor or greater than second correction factor and described T _{Pre_mid_o}The correction circulation average amplitude difference sequence maximal value MA of described current subframe in the nearby sphere _Max0Greater than the described first intermediate variable MA _MaxWith the product of the second experience factor, then use described MA _Max0Corresponding delay T ₀Proofread and correct the second intermediate variable T _Opt

Describedly use described MA _Max0The delay T that answers ₀Proofread and correct the second intermediate variable T _OptBe with the second intermediate variable T _OptCurrency be set to described MA _Max0Corresponding delay T ₀

Medium filtering module 403 is used for the pitch period estimated value and the second intermediate variable T with even number subframe before the described current subframe _OptCarry out medium filtering, with the value of carrying out obtaining behind the described medium filtering pitch period estimated value as middle subframe;

Subframe is that the preceding even number subframe of described current subframe and some subframe neutron frame numbers of described current subframe formation are the subframe of intermediate value in the middle of described, and described even number is the maximum even number less than described odd number.

Need to prove, in the embodiment of above pitch period estimated value correcting device, the division of each functional module only illustrates, can be as required in the practical application, for example the facility of the realization of the configuration requirement of corresponding hardware or software is considered, and the above-mentioned functions distribution is finished by different functional modules, the inner structure that is about to described pitch period estimated value correcting device is divided into different functional modules, to finish all or part of function described above.And in the practical application, the corresponding functional modules in the present embodiment can be to be realized by relevant hardware, also can carry out corresponding software and finish by relevant hardware, and for example, in the aforesaid correction module, can be to have the aforementioned calculating second intermediate variable T of execution _OptEstimate intermediate value T with the pitch period of odd number subframe before the current subframe _{Pre_mid_o}The hardware of ratio, ratio calculation device for example, thus also can be to carry out general processor or other hardware devices that the corresponding computer program is finished aforementioned functional; Aforesaid for another example medium filtering module can be to have aforementioned the pitch period estimated value and the second intermediate variable T with even number subframe before the current subframe of execution _OptCarrying out the hardware of medium filtering function, as median filter, thereby also can be to carry out general processor or other hardware devices that the corresponding computer program is finished aforementioned functional.

In the pitch period estimated value correcting device of accompanying drawing 4 examples, the fundamental frequency scope can be [80Hz, 4000Hz), described frequency field be [80Hz, the first frequency zone of dividing in 4000Hz) [80Hz, 160Hz), second frequency zone [160Hz, 320Hz), the 3rd frequency field [320Hz, 640Hz), the 4th frequency field [640Hz, 1280Hz) and the 5th frequency field [1280Hz, 4000Hz).At this moment, comparison module 401 specifically is used for: the default first intermediate variable MA _MaxValue is that current subframe is at [80Hz, 160Hz) the interior circulation average amplitude difference sequence maximal value MA that revises _Max(1), the default second intermediate variable T _OptValue is described MA _Max(1) Dui Ying delay T (1); More current subframe is at [160Hz, 320Hz) the interior circulation average amplitude difference sequence maximal value MA that revises _Max(2) with the first intermediate variable MA _MaxWeighted value and to described MA _MaxWith described T _OptRevise, more current subframe is in that [320Hz revises circulation average amplitude difference sequence maximal value MA in 640Hz) _Max(3) with the first intermediate variable MA _MaxWeighted value and to described MA _MaxWith described T _OptRevise, more current subframe is in that [640Hz revises circulation average amplitude difference sequence maximal value MA in 1280Hz) _Max(4) with the first intermediate variable MA _MaxWeighted value and to described MA _MaxWith described T _OptRevise, and more current subframe is at [1280Hz, 4000Hz) the interior circulation average amplitude difference sequence maximal value MA that revises _Max(5) with the first intermediate variable MA _MaxWeighted value and to described MA _MaxWith described T _OptRevise.

In the pitch period estimated value correcting device of accompanying drawing 4 examples, the pitch period of odd number subframe is estimated intermediate value T before the current subframe _{Pre_mid_o}Can estimate intermediate value T for the pitch period of preceding 5 subframes of described current subframe _{Pre_mid_o}, T _{Pre_mid_o}The correction circulation average amplitude difference sequence maximal value MA of described current subframe in the nearby sphere _Max0Greater than the described first intermediate variable MA _MaxWith the product of the second experience factor be described T _{Pre_mid_o}The correction circulation average amplitude difference sequence maximal value MA of described current subframe in contiguous 1 millisecond of scope _Max0Greater than the described first intermediate variable MA _MaxProduct with the second experience factor.At this moment, correction module 402 specifically is used to calculate the described second intermediate variable T _OptWith the ratio of the pitch period estimated value of the subframe that is numbered intermediate value in preceding 5 subframes of described current subframe, if described ratio is less than first correction factor or greater than second correction factor and described T _{Pre_mid_o}The correction circulation average amplitude difference sequence maximal value MA of described current subframe in contiguous 1 millisecond of scope _Max0Greater than the described first intermediate variable MA _MaxWith the product of the second experience factor, then with the second intermediate variable T _OptCurrency described MA is set _Max0Corresponding delay T ₀

In the pitch period estimated value correcting device of accompanying drawing 4 examples, medium filtering module 403 specifically is used for the pitch period estimated value T with preceding 4 subframes of described current subframe _{Pre_e}With the second intermediate variable T _OptCarry out medium filtering.

Seeing also accompanying drawing 5, is a kind of fundamental tone estimation unit logical organization synoptic diagram that the embodiment of the invention provides.For convenience of explanation, only show the part relevant with the embodiment of the invention.The fundamental tone estimation unit of accompanying drawing 5 examples can be used in the regular technology modules of time of speech audio Unified coding, and the fundamental tone that also can be used for voice signal estimates, it comprises that pretreatment module 501, sequence ask for module 502 and correcting module 503, wherein:

Pretreatment module 501 is used for the signal that receives is carried out pre-service, and described signal comprises voice signal or sound signal;

Sequence is asked for module 502, is used for its normalized crosscorrelation sequence of the pretreated calculated signals of described process, asks for the correction circulation average amplitude difference sequence of described normalized crosscorrelation sequence according to the described normalized crosscorrelation sequence that obtains;

Correcting module 503, be used for delay according to the maximal value correspondence of described correction circulation average amplitude difference sequence in the fundamental frequency scope, the pitch delay estimated value is revised, and is the pitch period estimated value of described signal with the delay estimated value of described correction gained.

The pretreatment module 501 of accompanying drawing 5 examples can also comprise judging unit 601, as the fundamental tone estimation unit of accompanying drawing 6 examples.After judging unit 601 is used for the signal that receives carried out pre-service, quiet frame and non-quiet frame to described signal are judged, at this moment, sequence is asked for module 502 and specifically is used for the pretreated non-quiet frame signal of described process is calculated its normalized crosscorrelation sequence, asks for the correction circulation average amplitude difference sequence of the normalized crosscorrelation sequence of described non-quiet frame signal according to the normalized crosscorrelation sequence of the described non-quiet frame signal that obtains.

The correcting module 503 of accompanying drawing 5 or accompanying drawing 6 examples can also comprise comparing unit 701, correcting unit 702 and median filter unit 703, as the fundamental tone estimation unit of accompanying drawing 7 examples, wherein:

Comparing unit 701 is used for more current subframe and revises circulation average amplitude difference sequence maximal value MA in current frequency field _Max(i+1) with the first intermediate variable MA _MaxWeighted value, if described MA _Max(i+1) greater than described MA _MaxWeighted value, then respectively with described MA _Max(i+1) and described MA _Max(i+1) Dui Ying delay T (i+1) substitutes the first intermediate variable MA _MaxWith the second intermediate variable T _Opt, repeat described comparison procedure, until described current frequency field not within the fundamental frequency scope, the described first intermediate variable MA _MaxWeighted value be described MA _MaxWith the product of the first experience factor, described frequency field is some frequency separations of dividing in described fundamental frequency scope;

Correcting unit 702 is used to calculate the described second intermediate variable T _OptEstimate intermediate value T with the pitch period of odd number subframe before the described current subframe _{Pre mid_o}Ratio, if described ratio is less than first correction factor or greater than second correction factor and described T _{Pre_mid_o}The correction circulation average amplitude difference sequence maximal value MA of described current subframe in the nearby sphere _Max0Greater than the described first intermediate variable MA _MaxWith the product of the second experience factor, then use described MA _Max0Corresponding delay T ₀Proofread and correct the second intermediate variable T _Opt

Describedly use described MA _Max0Corresponding delay T ₀Proofread and correct the second intermediate variable T _OptBe with the second intermediate variable T _OptCurrency be set to described MA _Max0Corresponding delay T ₀,

Median filter unit 703 is used for the pitch period estimated value and the second intermediate variable T with even number subframe before the described current subframe _OptCarry out medium filtering, with the value of carrying out obtaining behind the described medium filtering pitch period estimated value as middle subframe;

Subframe is that some subframe neutron frame numbers that preceding even number subframe of described current subframe and described current subframe constitute are the subframe of intermediate value in the middle of described, and described even number is the maximum even number less than described odd number.

Need to prove, contents such as the information interaction between each module/unit of said apparatus, implementation, since with the inventive method embodiment based on same design, its technique effect that brings is identical with the inventive method embodiment, particular content can repeat no more referring to the narration among the inventive method embodiment herein.

Accompanying drawing 8 has provided accompanying drawing 5 is used for time regular (TW, the Time Warp) technology modules of speech audio Unified coding to the fundamental tone estimation unit of accompanying drawing 7 arbitrary examples structural representation.The regular technology modules of time of the speech audio Unified coding of accompanying drawing 8 examples comprises: (Modified Discrete Cosine Transform, MDCT) module 809 and accompanying drawing 5 are to the fundamental tone estimation unit 802 of accompanying drawing 7 arbitrary examples in the discrete cosine transform of psychologic acoustics control module 801, TW information scrambler 803, TW information decoding device 804, TW mapping model structure module 805, bit output module 806, TW resampling module 807, windowing mapping block 808, correction.

In regular technology modules of the time of accompanying drawing 8 examples, if according to the signal discriminant classification, judge and use the Frequency Domain Coding mode to encode, when scrambler adopts the TW pattern, then signal to be encoded is sent into psychologic acoustics control module 801, fundamental tone estimation unit 802 and TW resampling module 807 respectively.Psychologic acoustics control module 801 is extracted the psychoacoustic parameter of signal, and the window type that is used for auxiliary windowing mapping block 808 is selected; The tonal variations of 802 pairs of signals of fundamental tone estimation unit is followed the tracks of, and extracts Pitch Information, mainly comprises pitch period estimated value (in the present embodiment, pitch period can be used for calculating TW mapping curve parameter information); The time domain that TW resampling module 807 is finished signal to be encoded remaps, to improve the spectrum concentration degree of signal to be encoded at frequency domain.The TW mapping curve parameter information of 803 pairs of inputs of TW information scrambler carries out coded quantization and is stored in the code stream, and the TW mapping curve parameter information behind 804 pairs of coded quantizations of TW information decoding device is decoded.In the present embodiment, use decoded TW mapping curve parameter information to make up the TW mapping model, the mapping parameters of encoding and decoding end is consistent, thereby can recover and the on all four mapping curve of coding side according to this parameter information in decoding end, avoid the error of bringing because of quantization encoding.Decoded TW mapping curve parameter information inputs to the TW mapping model and makes up module 805.The TW mapping model makes up module 805 and makes up mapping curve according to decoded TW mapping curve parameter information, and calculates control informations such as required resampling of time-frequency conversion operation and window function.After signal process TW resampling module 807 resamplings to be encoded and the TW windowing mapping block 808 windowings mapping, obtain TW territory signal, send into the discrete cosine transform module 809 of correction and carry out conversion, thereby obtain the coefficient of TW-MDCT.

Because during the TW coding, every frame needs 16 Pitch Information to calculate mapping curve, when every frame data length adopts at 1024 during realization, promptly every fundamental tone data of 64 output.The specific embodiments of fundamental tone tracking module is: adopt 128 rectangular window, the mode that each translation is 64 is carried out pitch period and is estimated that every frame data calculate 16 pitch value.

Accompanying drawing 9 has provided fundamental tone that accompanying drawing 5 to the fundamental tone estimation unit of accompanying drawing 7 arbitrary examples is used for voice signal when estimating, voice signal fundamental tone estimating system logical organization synoptic diagram.For convenience of explanation, only show the part relevant with the embodiment of the invention.The voice signal fundamental tone estimating system of accompanying drawing 9 examples comprises that pretreatment module 901, voiceless sound voiced sound judge module 902, center clipping module 903, pitch period zero setting module 904, sequence ask for module 905 and correcting module 906, wherein, pretreatment module 901, sequence are asked for module 905 and correcting module 906 can be respectively that accompanying drawing 5 to pretreatment module, sequence in accompanying drawing 7 arbitrary example fundamental tone estimation units asked for module and correcting module:

Pretreatment module 901 is used for voice signal to input and carries out the high pass low-pass filtering, goes average and numerical value level and smooth etc.;

Voiceless sound voiced sound judge module 902 is used for judging that according to short-time energy and short-time zero-crossing rate the voice signal of handling through pretreatment module 901 is voiced sound or voiceless sound;

Center clipping module 903 is used for removing voiced sound signal energy smaller portions;

Pitch period zero setting module 904 is used for the pitch period zero setting to the voiceless sound signal;

Sequence is asked for module 905, is used for its normalized crosscorrelation sequence of calculated signals through described center clipping module 903 processing is asked for the correction circulation average amplitude difference sequence of described normalized crosscorrelation sequence according to the described normalized crosscorrelation sequence that obtains;

Correcting module 906, be used for asking for the delay of the maximal value correspondence of correction circulation average amplitude difference sequence in the fundamental frequency scope that module 905 asks for according to described sequence, the pitch delay estimated value is revised, and is the pitch period estimated value of described signal with the delay estimated value of described correction gained.

One of ordinary skill in the art will appreciate that all or part of step in the whole bag of tricks of the foregoing description is to instruct relevant hardware to finish by program, this program can be stored in the computer-readable recording medium, storage medium can comprise: ROM (read-only memory) (ROM, Read Only Memory), random access memory (RAM, Random Access Memory), disk or CD etc.

More than pitch period estimated value modification method, pitch estimation method and relevant apparatus that the embodiment of the invention is provided be described in detail, used specific case herein principle of the present invention and embodiment are set forth, the explanation of above embodiment just is used for helping to understand method of the present invention and core concept thereof; Simultaneously, for persons skilled in the art, according to thought of the present invention, the part that all can change in specific embodiments and applications, in sum, this description should not be construed as limitation of the present invention.

Claims (16)

1. pitch period estimated value modification method is characterized in that described method comprises:

More current subframe is revised circulation average amplitude difference sequence maximal value MA in current frequency field _Max(i+1) with the first intermediate variable MA _MaxWeighted value, if described MA _Max(i+1) greater than described MA _MaxWeighted value, then respectively with described MA _Max(i+1) and described MA _Max(i+1) Dui Ying delay T (i+1) substitutes the first intermediate variable MA _MaxWith the second intermediate variable T _Opt, repeat described comparison procedure, until described current frequency field not within the fundamental frequency scope;

Calculate the described second intermediate variable T _OptEstimate intermediate value T with the pitch period of odd number subframe before the described current subframe _{Pre_mid_o}Ratio, if described ratio is less than first correction factor or greater than second correction factor and described T _{Pre_mid_o}The correction circulation average amplitude difference sequence maximal value MA of described current subframe in the nearby sphere _Max0Greater than the described first intermediate variable MA _MaxWith the product of the second experience factor, then use described MA _Max0Corresponding delay T ₀Proofread and correct the second intermediate variable T _Opt

The pitch period estimated value and the second intermediate variable T with even number subframe before the described current subframe _OptCarry out medium filtering, with the value of carrying out obtaining behind the described medium filtering pitch period estimated value as middle subframe;

The described first intermediate variable MA _MaxWeighted value be described MA _MaxWith the product of the first experience factor, describedly use described MA _Max0Corresponding delay T ₀Proofread and correct the second intermediate variable T _OptBe with the second intermediate variable T _OptCurrency be set to described MA _Max0Corresponding delay T ₀, described middle subframe is that some subframe neutron frame numbers that preceding even number subframe of described current subframe and described current subframe constitute are the subframe of intermediate value, described even number is the maximum even number less than described odd number.

2. the method for claim 1, it is characterized in that, described fundamental frequency scope be [80Hz, 4000Hz), described frequency field is at [80Hz, first frequency zone [the 80Hz that divides 4000Hz), 160Hz), the second frequency zone [160Hz, 320Hz), the 3rd frequency field [320Hz, 640Hz), the 4th frequency field [640Hz, 1280Hz) and the 5th frequency field [1280Hz, 4000Hz);

Described more current subframe is revised circulation average amplitude difference sequence maximal value MA in current frequency field _Max(i+1) with the first intermediate variable MA _MaxWeighted value comprise:

The default first intermediate variable MA _MaxValue is that current subframe is at [80Hz, 160Hz) the interior circulation average amplitude difference sequence maximal value MA that revises _Max(1), the default second intermediate variable T _OptValue is described MA _Max(1) Dui Ying delay T (1);

If current subframe is at [160Hz, 320Hz) the interior circulation average amplitude difference sequence maximal value MA that revises _Max(2) greater than described MA _MaxWeighted value, then respectively with described MA _Max(2) and described MA _Max(2) Dui Ying delay T (2) substitutes described MA _MaxWith described T _Opt

If current subframe is at [320Hz, 640Hz) the interior circulation average amplitude difference sequence maximal value MA that revises _Max(3) greater than described MA _MaxWeighted value, then respectively with described MA _Max(3) and described MA _Max(3) Dui Ying delay T (3) substitutes described MA _MaxWith described T _Opt

If current subframe is at [640Hz, 1280Hz) the interior circulation average amplitude difference sequence maximal value MA that revises _Max(4) greater than described MA _MaxWeighted value, then respectively with described MA _Max(4) and described MA _Max(4) Dui Ying delay T (4) substitutes described MA _MaxWith described T _Opt

If current subframe is at [1280Hz, 4000Hz) the interior circulation average amplitude difference sequence maximal value MA that revises _Max(5) greater than described MA _MaxWeighted value, then respectively with described MA _Max(5) and described MA _Max(5) Dui Ying delay T (5) substitutes described MA _MaxWith described T _Opt

3. the method for claim 1 is characterized in that, the pitch period of odd number subframe is estimated intermediate value T before the described current subframe _{Pre_mid_o}For the pitch period of preceding 5 subframes of described current subframe is estimated intermediate value T _{Pre_mid_o}

The pitch period of preceding 5 subframes of described current subframe is estimated intermediate value T _{Pre_mid_o}For being numbered the pitch period estimated value of the subframe of intermediate value in preceding 5 subframes of described current subframe;

The described second intermediate variable T of described calculating _OptEstimate intermediate value T with the pitch period of odd number subframe before the described current subframe _{Pre_mid_o}Ratio be:

Calculate the described second intermediate variable T _OptRatio with the pitch period estimated value of the subframe that is numbered intermediate value in preceding 5 subframes of described current subframe.

4. the method for claim 1 is characterized in that, described T _{Pre_mid_o}The correction circulation average amplitude difference sequence maximal value MA of described current subframe in the nearby sphere _Max0Greater than the described first intermediate variable MA _MaxWith the product of the second experience factor be:

Described T _{Pre_mid_o}The correction circulation average amplitude difference sequence maximal value MA of described current subframe in contiguous 1 millisecond of scope _Max0Greater than the described first intermediate variable MA _MaxProduct with the second experience factor.

Describedly use described MA _Max0Corresponding delay T ₀Proofread and correct the second intermediate variable T _OptBe with the second intermediate variable T _OptCurrency be set to T _{Pre_mid_o}The correction circulation average amplitude difference sequence maximal value MA of described current subframe in the nearby sphere _Max0Corresponding delay T ₀

5. the method for claim 1 is characterized in that, described the pitch period estimated value and the second intermediate variable T with even number subframe before the described current subframe _OptCarrying out medium filtering comprises:

The pitch period estimated value and the second intermediate variable T with preceding 4 subframes of described current subframe _OptCarry out medium filtering.

6. as any described method of claim 1 to 5, it is characterized in that the described first experience factor is 0.95, described first correction factor is 0.75, and described second correction factor is 1.4, and the described second experience factor is 0.85.

7. pitch period estimated value correcting device is characterized in that described device comprises:

Comparison module is used for more current subframe and revises circulation average amplitude difference sequence maximal value MA in current frequency field _Max(i+1) with the first intermediate variable MA _MaxWeighted value, if described MA _Max(i+1) greater than described MA _MaxWeighted value, then respectively with described MA _Max(i+1) and described MA _Max(i+1) Dui Ying delay T (1) substitutes the first intermediate variable MA _MaxWith the second intermediate variable T _Opt, repeat described comparison procedure, until described current frequency field not within the fundamental frequency scope;

Correction module is used to calculate the described second intermediate variable T _OptEstimate intermediate value T with the pitch period of odd number subframe before the described current subframe _{Pre_mid_o}Ratio, if described ratio is less than first correction factor or greater than second correction factor and described T _{Pre_mid_o}The correction circulation average amplitude difference sequence maximal value MA of described current subframe in the nearby sphere _Max0Greater than the described first intermediate variable MA _MaxWith the product of the second experience factor, then use described MA _Max0Corresponding delay T ₀Proofread and correct the second intermediate variable T _Opt

The medium filtering module is used for the pitch period estimated value and the second intermediate variable T with even number subframe before the described current subframe _OptCarry out medium filtering, with the pitch period estimated value of carrying out obtaining behind the described medium filtering pitch period estimated value as middle subframe;

The described first intermediate variable MA _MaxWeighted value be described MA _MaxWith the product of the first experience factor, describedly use described MA _Max0Corresponding delay T ₀Proofread and correct the second intermediate variable T _OptBe with the second intermediate variable T _OptCurrency be set to described MA _Max0Corresponding delay T ₀, some subframe neutron frame numbers that preceding even number subframe that described middle subframe is described current subframe and described current subframe constitute are the subframe of intermediate value, described even number is the maximum even number less than described odd number.

8. device as claimed in claim 7, it is characterized in that, described fundamental frequency scope be [80Hz, 4000Hz), described frequency field is at [80Hz, first frequency zone [the 80Hz that divides 4000Hz), 160Hz), the second frequency zone [160Hz, 320Hz), the 3rd frequency field [320Hz, 640Hz), the 4th frequency field [640Hz, 1280Hz) and the 5th frequency field [1280Hz, 4000Hz);

Described comparison module specifically is used for:

The default first intermediate variable MA _MaxValue is that current subframe is at [80Hz, 160Hz) the interior circulation average amplitude difference sequence maximal value MA that revises _Max(1), the default second intermediate variable T _OptValue is described MA _Max(1) Dui Ying delay T (1);

If current subframe is at [160Hz, 320Hz) the interior circulation average amplitude difference sequence maximal value MA that revises _Max(2) greater than described MA _MaxWeighted value, then respectively with described MA _Max(2) and described MA _Max(2) Dui Ying delay T (2) substitutes described MA _MaxWith described T _Opt

If current subframe is at [320Hz, 640Hz) the interior circulation average amplitude difference sequence maximal value MA that revises _Max(3) greater than described MA _MaxWeighted value, then respectively with described MA _Max(3) and described MA _Max(3) Dui Ying delay T (3) substitutes described MA _MaxWith described T _Opt

If current subframe is at [640Hz, 1280Hz) the interior circulation average amplitude difference sequence maximal value MA that revises _Max(4) greater than described MA _MaxWeighted value, then respectively with described MA _Max(4) and described MA _Max(4) Dui Ying delay T (4) substitutes described MA _MaxWith described T _Opt

If current subframe is at [1280Hz, 4000Hz) the interior circulation average amplitude difference sequence maximal value MA that revises _Max(5) greater than described MA _MaxWeighted value, then respectively with described MA _Max(5) and described MA _Max(5) Dui Ying delay T (5) substitutes described MA _MaxWith described T _Opt

9. device as claimed in claim 7 is characterized in that, the pitch period of odd number subframe is estimated intermediate value T before the described current subframe _{Pre_mid_o}For the pitch period of preceding 5 subframes of described current subframe is estimated intermediate value T _{Pre_mid_o}

The pitch period of preceding 5 subframes of described current subframe is estimated intermediate value T _{Pre_mid_o}For being numbered the pitch period estimated value of the subframe of intermediate value in preceding 5 subframes of described current subframe;

Described T _{Pre_mid_o}The correction circulation average amplitude difference sequence maximal value MA of described current subframe in the nearby sphere _Max0Greater than the described first intermediate variable MA _MaxWith the product of the second experience factor be described T _{Pre_mid_o}The correction circulation average amplitude difference sequence maximal value MA of described current subframe in contiguous 1 millisecond of scope _Max0Greater than the described first intermediate variable MA _MaxProduct with the second experience factor;

Describedly use described MA _Max0Corresponding delay T ₀Proofread and correct the second intermediate variable T _OptBe with the second intermediate variable T _OptCurrency be set to described MA _Max0Corresponding delay T ₀

Described correction module specifically is used for:

Calculate the described second intermediate variable T _OptRatio with the pitch period estimated value of the subframe that is numbered intermediate value in preceding 5 subframes of described current subframe;

If described ratio is less than first correction factor or greater than second correction factor and described T _{Pre_mid_o}The correction circulation average amplitude difference sequence maximal value MA of described current subframe in contiguous 1 millisecond of scope _Max0Greater than the described first intermediate variable MA _MaxWith the product of the second experience factor, then with the second intermediate variable T _OptCurrency be set to described MA _Max0Corresponding delay T ₀

10. device as claimed in claim 7 is characterized in that, described medium filtering module specifically is used for:

Pitch period estimated value T with preceding 4 subframes of described current subframe _{Pre_e}With the second intermediate variable T _OptCarry out medium filtering.

11. a pitch estimation method is characterized in that, described method comprises:

The signal that receives is carried out pre-service, and described signal comprises voice signal or sound signal;

To its normalized crosscorrelation sequence of the pretreated calculated signals of described process, ask for the correction circulation average amplitude difference sequence of described normalized crosscorrelation sequence according to the described normalized crosscorrelation sequence that obtains;

According to the delay of the maximal value correspondence of described correction circulation average amplitude difference sequence in the fundamental frequency scope, the pitch delay estimated value is revised, be the pitch period estimated value of described signal with the delay estimated value of described correction gained.

12. method as claimed in claim 11 is characterized in that, describedly the signal that receives is carried out pre-service comprises the quiet frame and the non-quiet frame of described signal are judged;

Described to its normalized crosscorrelation sequence of the pretreated calculated signals of described process, the correction circulation average amplitude difference sequence of asking for described normalized crosscorrelation sequence according to described normalized crosscorrelation sequence is:

The pretreated non-quiet frame signal of described process is calculated its normalized crosscorrelation sequence, ask for the correction circulation average amplitude difference sequence of the normalized crosscorrelation sequence of described non-quiet frame signal according to the normalized crosscorrelation sequence of the described non-quiet frame signal that obtains.

13. method as claimed in claim 11 is characterized in that, described delay according to the maximal value correspondence of described correction circulation average amplitude difference sequence in the fundamental frequency scope is revised the pitch delay estimated value and to be comprised:

More current subframe is revised circulation average amplitude difference sequence maximal value MA in current frequency field _Max(i+1) with the first intermediate variable MA _MaxWeighted value, if described MA _Max(i+1) greater than described MA _MaxWeighted value, then respectively with described MA _Max(i+1) and described MA _Max(i+1) Dui Ying delay substitutes the first intermediate variable MA _MaxWith the second intermediate variable T _Opt, repeat described comparison procedure, until described current frequency field not within the fundamental frequency scope;

Calculate the described second intermediate variable T _OptEstimate intermediate value T with the pitch period of odd number subframe before the described current subframe _{Pre_mid_o}Ratio, if described ratio is less than first correction factor or greater than second correction factor and described T _{Pre_mid_o}The correction circulation average amplitude difference sequence maximal value MA of described current subframe in the nearby sphere _Max0Greater than the described first intermediate variable MA _MaxWith the product of the second experience factor, then use described MA _Max0Corresponding delay T ₀Proofread and correct the second intermediate variable T _Opt

The pitch period estimated value and the second intermediate variable T with even number subframe before the described current subframe _OptCarry out medium filtering, with the value of carrying out obtaining behind the described medium filtering pitch period estimated value as middle subframe;

The described first intermediate variable MA _MaxWeighted value be described MA _MaxWith the product of the first experience factor, describedly use described MA _Max0Corresponding delay T ₀Proofread and correct the second intermediate variable T _OptBe with the second intermediate variable T _OptCurrency be set to described MA _Max0Corresponding delay T ₀, described middle subframe is that some subframe neutron frame numbers that preceding even number subframe of described current subframe and described current subframe constitute are the subframe of intermediate value, described even number is the maximum even number less than described odd number.

14. a fundamental tone estimation unit is characterized in that, described device comprises:

Pretreatment module is used for the signal that receives is carried out pre-service, and described signal comprises voice signal or sound signal;

Sequence is asked for module, is used for its normalized crosscorrelation sequence of the pretreated calculated signals of described process, asks for the correction circulation average amplitude difference sequence of described normalized crosscorrelation sequence according to the described normalized crosscorrelation sequence that obtains;

Correcting module is used for the delay according to the maximal value correspondence of described correction circulation average amplitude difference sequence in the fundamental frequency scope, the pitch delay estimated value revised, and be the pitch period estimated value of described signal with the delay estimated value of described correction gained.

15. device as claimed in claim 14 is characterized in that, described pretreatment module also comprises judging unit, is used for that the signal that receives is carried out pre-service and comprises the quiet frame and the non-quiet frame of described signal are judged;

Described sequence is asked for module and specifically is used for:

The pretreated non-quiet frame signal of described process is calculated its normalized crosscorrelation sequence, ask for the correction circulation average amplitude difference sequence of the normalized crosscorrelation sequence of described non-quiet frame signal according to the normalized crosscorrelation sequence of the described non-quiet frame signal that obtains.

16. device as claimed in claim 14 is characterized in that, described correcting module comprises:

Comparing unit is used for more current subframe and revises circulation average amplitude difference sequence maximal value MA in current frequency field _Max(i+1) with the first intermediate variable MA _MaxWeighted value, if described MA _Max(i+1) greater than described MA _MaxWeighted value, then respectively with described MA _Max(i+1) and described MA _Max(i+1) Dui Ying delay T (1) substitutes the first intermediate variable MA _MaxWith the second intermediate variable T _Opt, repeat described comparison procedure, until described current frequency field not within the fundamental frequency scope;

Correcting unit is used to calculate the described second intermediate variable T _OptEstimate intermediate value T with the pitch period of odd number subframe before the described current subframe _{Pre_mid_o}Ratio, if described ratio is less than first correction factor or greater than second correction factor and described T _{Pre_mid_o}The correction circulation average amplitude difference sequence maximal value MA of described current subframe in the nearby sphere _Max0Greater than the described first intermediate variable MA _MaxWith the product of the second experience factor, then use described MA _Max0Corresponding delay T ₀Proofread and correct the second intermediate variable T _Opt

Median filter unit is used for the pitch period estimated value and the second intermediate variable T with even number subframe before the described current subframe _OptCarry out medium filtering, with the pitch period estimated value of carrying out obtaining behind the described medium filtering pitch period estimated value as middle subframe;

The described first intermediate variable MA _MaxWeighted value be described MA _MaxWith the product of the first experience factor, describedly use described MA _Max0Corresponding delay T ₀Proofread and correct the second intermediate variable T _OptBe with the second intermediate variable T _OptCurrency be set to described MA _Max0Corresponding delay T ₀, described middle subframe is that some subframe neutron frame numbers that preceding even number subframe of described current subframe and described current subframe constitute are the subframe of intermediate value, described even number is the maximum even number less than described odd number.

Images