CN1146297C - Device and method for signal quality determination - Google Patents

Abstract

The present invention relates to a device for determining the quality of output signals generated by a signal processing circuit according to reference signals, and the signal processing circuit comprises a radio connection mode. The device is provided with a first serial connection circuit and a second serial connection circuit, wherein the first serial connection circuit is used for receiving the output signals, and the second serial connection circuit is used for receiving the reference signals; a combination circuit connected between the two serial connection circuits generates objective quality signals, and a scaling circuit is connected between the two serial connection circuits so as to scale at least a serial connection circuit signal. The poor correlativity between objective quality signals and subjective quality signals which are appraised by human viewers is remarkably improved because a discount group block is built-in in the combination circuit; the discount group block is connected to the scaling circuit so as to receive comparison signals and scales the comparison signals; meanwhile, the discount group block generates quality signals.

Description

Signal quality is determined apparatus and method

The A background of invention

The present invention relates to a kind of device that is used for determining the quality of an output signal being produced according to a reference signal by a signal processing circuit, this device has:

-the first series circuit, it has the first input end that is used to receive output signal, and this first series circuit has

-the first signal processing chunk, it is connected to the first input end of first series circuit, be used to produce one as first signal parameter of the function of time and frequency and

-the first compressed chunks, it is connected to the first signal processing chunk, and be used to compress first signal parameter and produce the first compressed signal parameter,

-the second series circuit, it has second input that is used to receive reference signal, and this second series circuit has

-the second compressed chunks, it is connected to second input, is used to produce the second compressed signal parameter,

-one combinational circuit, it is connected to first output of first series circuit and second output of second series circuit, is used to produce a quality signal, and this combinational circuit has

-one difference chunk, it is connected to two compressed chunks, be used for according to the compressed signal parameter determine a differential signal and

-one integration chunk, it is connected to the difference chunk, is used for by time and frequency the differential signal integration producing quality signal,

-one time average chunk, be used for to by the differential signal behind the frequency integrator by the time integration producing quality signal,

-one convergent-divergent circuit, it is connected between the input of two compressed chunks, and this convergent-divergent circuit has

-one integration chunk again, be used for by frequency to the first series circuit signal and the second series circuit signal integration and

-one chunk relatively, it is connected to integration chunk again, is used for the series circuit signal behind two integrations of comparison and by comparative result at least one series circuit signal is carried out convergent-divergent.

This device is by WO 96/28953, WO 96/28952 and WO 96/28950 are open, this international patent application is defined as an invention that improves existing device, should be disclosed in No. the 12nd, radio engineering meeting periodical the 40th volume by existing device, in December, 1992, particularly publish " a kind of audio quality of sensibility of the sound representation based on tonequality is measured ", especially its Fig. 7 of being shown by John G.Beerends and Jan A.Stemerdink in the 963-978 page or leaf.WO 96/28953 described this existing device is determined by the quality of a signal processing circuit according to the output signal that reference signal produced, and this signal processing circuit for example is a codec for reference signal.Described reference signal is an input signal that for example is delivered to signal processing circuit, though also comprise and can adopt one to calculate good Utopian output signal in advance as reference signal.First signal parameter is produced according to output signal by the first signal processing chunk that is associated with first series circuit as the function of time and frequency, and first signal parameter is by the compression of first compressed chunks that is associated with first series circuit then.In this connected mode, the intermediary operation of described first signal parameter is handled and must not be cancelled.The secondary signal parameter is compressed according to reference signal by second compressed chunks that is associated with second series circuit.In this connected mode, the further operational processes of described secondary signal parameter also must not be cancelled.The differential signal of two compressed signals produces quality signal by time and frequency to the differential signal integration by the integration chunk that is associated with combinational circuit then by the difference chunk decision that is associated with combinational circuit.Owing to increased the convergent-divergent circuit, this existing device is improved.Because this convergent-divergent circuit is arranged, has good correlation by described improved device objective quality signal of evaluating and the subjective quality signal of evaluating by human viewer.

Yet especially, the shortcoming of this device is, when signal processing circuit comprises the dedicated radio link mode, is bad by the correlation of described device objective quality signal of evaluating and the subjective quality signal of being evaluated by human viewer.

WO 96/28952 discloses a kind of similar device, wherein, before producing quality signal, adopts (or reference signal) representation of a kind of assessed signal, carrying out a kind of compensation beyond time-domain or time/frequency domain, thereby improves its correlation.In the time of should determining the situation of performed compensation, this compensation is invalid but in all density (for example, for whole frequencies) of the assessed signal that a definite time point occurs wherein being positioned at.This situation for example is present in, and signal transmits through the dedicated radio link mode.

WO 96/28950 also discloses a kind of similar device, wherein increases an adjustment chunk that is used to reduce the differential signal amplitude on the difference chunk, and correlation is improved.This open scheme that solves above-mentioned compensation problem that also do not provide.

The B summary of the invention

The objective of the invention is, especially, provide the device of the described type of a kind of this paper preamble, make to have correlation preferably by the objective quality signal of described device evaluation and the subjective quality signal of evaluating by human viewer.

For reaching this purpose, device according to the present invention has such feature, and promptly this device also comprises

Handle chunk for-one, be used for process source from comparison signal of chunk relatively and

-one multiplication chunk, comprising

-first input end is connected to an output of the first signal processing chunk,

-the second input, be connected to the integration chunk input and

-one output is connected to an input of time average chunk.

Because method of the present invention, the significantly difference that particularly is present between two series circuit signals is also considered in the integration chunk.Because above-mentioned consideration, even when signal that need to determine its quality is to pass wherein defeated the time through the dedicated radio link mode, also can obtain the objective quality signal evaluated by described device and the subjective quality signal evaluated by human viewer between good correlation.

Especially, the present invention is based on and recognize, bad correlation between objective quality signal of being evaluated by described device and the subjective quality signal evaluated by human viewer also can cause such situation especially, promptly especially when occurring big difference between the series circuit signal, self-evidently mean poor quality.

Adopt treated chunk to be connected to the multiplication chunk of convergent-divergent circuit, can solve the bad problem of correlation.

Should be noted that, when signal processing circuit comprises asynchronous transfer mode and is derived from or when belonging to the signal of reference signal, adopts device of the present invention also can improve correlation when signal processing circuit generation is different from very much.

First embodiment according to device of the present invention has such characteristic, and promptly its convergent-divergent circuit has

-one unit for scaling, comprising

-one input is connected to the output of the first signal processing chunk,

-one output, be connected to first compressed chunks input and

-one control input end is connected to the output that is used for according to comparative result the first series circuit signal being carried out the comparison chunk of convergent-divergent.

Because provide to have the convergent-divergent circuit that is used for the first series circuit signal is carried out the unit for scaling of convergent-divergent, the function of convergent-divergent circuit is brought into play preferably.So correlation is got back and is further improved.

Second embodiment according to device of the present invention has such characteristic, and promptly it is handled chunk comparison signal is promoted to intensity P, wherein 0＜P＜1.

In this connected mode, significantly difference will be carried out convergent-divergent again according to the relation between two series circuit signals.

The 3rd embodiment according to device of the present invention has such characteristic, and promptly its second series circuit also has

-secondary signal is handled chunk, and it is connected to second input, is used to produce the secondary signal parameter as the function of time and frequency, and second compressed chunks is connected to secondary signal and handles chunk with compression secondary signal parameter.

The invention still further relates to a kind of method for quality that is used for determining an output signal being produced according to a reference signal by a signal processing circuit, this method comprises known and according to the step of the characteristic of claim 5.

The further embodiment of the method according to this invention sets forth in the method dependent claims.

The C documents

* WO?96/28953

* WO?96/28950

* WO?96/28952

* radio engineering can periodical No. the 12nd, the 40th volume, in December, 1992, particularly " a kind of audio quality of sensibility of the sound representation based on tonequality is measured " of being shown, 963-978 page or leaf by John G.Beerends and Jan A.Stemerdink.

* " the perception simulation in the mass measurement of music codec ", author John G.Beerends and Jan A.Stemerdink are published in the 96th conference, and on February 26th, 1994 is to March 1, Amsterdam.

All documents are included in the literal of quoting in these documents, all are regarded as the component part of present patent application.

D embodiment

With reference to examples of implementation shown in the drawings the present invention is done more detailed explanation.In the accompanying drawing:

Fig. 1 illustrates according to a device of the present invention, comprising, known signal processing chunk, known compressed chunks, according to convergent-divergent circuit of the present invention with according to combinational circuit of the present invention,

Fig. 2 illustrates the known signal processing chunk that is used for according to device of the present invention,

Fig. 3 illustrates the known compressed chunks that is used for according to device of the present invention,

Fig. 4 illustrate be used for according to device of the present invention according to convergent-divergent circuit of the present invention,

Fig. 5 illustrate be used for according to device of the present invention according to combinational circuit of the present invention.

Basis as shown in Figure 1 the inventive system comprises a band and is useful on the reception output signal The first signal of first input end 7 process chunk 1, this output signal is derived from one and for example is The signal processing circuit of codec. First signal is processed first output of chunk 1 through connecting Line 9 is connected to the first input end of convergent-divergent circuit 3. Also comprise one according to device of the present invention Band is useful on the first signal of second input 8 that receives input signal and processes chunk 2, and this is defeated Entering signal, to transport to one for example be the signal processing circuit of codec. Secondary signal is processed chunk 2 second output is connected to second input of convergent-divergent circuit 3 through connecting line 10. The convergent-divergent electricity First output on road 3 is connected to the first input end of first compressed chunks 4 through connecting line 11, Second output of convergent-divergent circuit 3 is connected to second of second compressed chunks 5 through connecting line 12 Input. First output of first compressed chunks 4 is connected to combinational circuit 6 through connecting line 13 First input end, second output of second compressed chunks 5 is connected to group through connecting line 16 Close second input of circuit 6. The 3rd output of convergent-divergent circuit 3 connects through connecting line 14 To the 3rd input of combinational circuit 6, second output of second compressed chunks 5, or connect Line 16 is connected to the four-input terminal of combinational circuit 6, combinational circuit 6 tools through connecting line 15 Be useful on the output 17 that produces a quality signal. The 4th output of convergent-divergent circuit 3 is through connecting Wiring 18 is connected to the 5th input of combinational circuit 6. First signal is processed chunk 1 and the One compressed chunks 4 is equivalent to first series circuit together, and secondary signal is processed chunk 2 and second Compressed chunks 5 is equivalent to second series circuit together.

Known first (or second) signal shown in Figure 2 is processed chunk 1 (or 2) and is comprised first (or second) multiplier 20 is used in time-domain by means of window function flowing to first (or second) signal process chunk 1 first input end 7 (or second input 8) or be derived from The output signal (or input signal) of signal processing circuit (for example codec) doubles; One Individual first (or second) converter 21 that is connected to first (or second) multiplier 20, being used for will Signal from first (or second) multiplier 20 transforms to frequency domain; First (or second) is exhausted To value chunk 22, for the absolute value of determining from the signal of first (or second) converter 21 To produce first (or second) positive signal parameter as the function of time and frequency; First (or Second) converter 23, be used for will from first (or second) absolute value chunk 22 and with the time Between spectrum and first (or second) positive signal parameter of representing of frequency spectrum be converted to time spectrum with bark First (or second) signal parameter of sound spectrum (Bark spectrum) expression; And first (or Second) the discount device 24, be used for to be derived from first (or second) converter and with time spectrum and First (or second) signal parameter of bark spectral representation is used in the sense of hearing and is imposed discount, then will This signal transmits through connecting line 9 (or 10).

First (or second) compressed chunks, 4 (or 5) shown in Figure 3 receive the signal parameter of first (or second) input that flows to first (or second) adder 30 through connecting line 11 (or 12), its first (or second) output is connected to first (or second) input of the 3rd (or 4th) multiplier 32 on the one hand through connecting line 31, be connected to the non-linear convolution chunk 36 of first (or second) on the other hand, this convolution chunk is connected to first (or second) compressed chunks 37 again, in order to produce first (or second) compressed signal parameter through connecting line 13 (or 16).The 3rd (or 4th) multiplier 32 also has first (or second) input, be used to receive an input signal, with first (or second) output, it is connected to first (or second) input of first (or second) time-delay chunk 34, and first (or second) output of this time-delay chunk is connected to first (or second) input in addition of adder 30.

Convergent-divergent circuit shown in Figure 4 comprises an integration chunk 40 again, its first input end is connected to the first input end of convergent-divergent circuit 3, thereby be connected to connecting line 9, in order to receive the first series circuit signal (with first signal parameter of time spectrum and bark spectral representation), its second input is connected to second input of convergent-divergent circuit 3, thereby be connected to connecting line 10, in order to receive the second series circuit signal (with the secondary signal parameter of time spectrum and bark spectral representation).First output that is used to produce the chunk of integration again 40 of the first series circuit signal behind the integration is connected to the first input end of comparison chunk 41, and second output of integration chunk 40 is connected to second input of comparison chunk 41 again.The first input end of convergent-divergent circuit 3 is connected to the first input end of a unit for scaling 42, and second output is connected to the output of unit for scaling 42, and connecting line 9 is switched to connecting line 11 through convergent-divergent circuit 3 orders through unit for scaling 42.Second input of convergent-divergent circuit 3 is connected to second output, and connecting line 10 connects in turn to connecting line 12 through convergent-divergent circuit 3.The output that is used to produce the comparison chunk 41 of comparison signal is connected to the control input end of unit for scaling 42, and is connected to connecting line 18 through the 4th output of convergent-divergent circuit 3.The output of unit for scaling 42, or connecting line 11, be connected to a ratio and determine the first input end of chunk 43, second input of convergent-divergent circuit 3, or connecting line 10 or connecting line 12, be connected to ratio and determine second input of chunk 43, ratio determines that the output of chunk is connected to the 3rd output of convergent-divergent circuit 3, is connected to connecting line 14 then to produce a scale signal.

Combinational circuit 6 shown in Figure 5 comprises that is compared a chunk 50 again, its first input end is connected to the first input end of combinational circuit 6 to receive the first compressed signal parameter through connecting line 13, and its second input is connected to second input of combinational circuit 6 to receive the second compressed signal parameter through connecting line 16.The first input end of combinational circuit 6 also is connected to the first input end of a difference chunk 54,56.Be used to produce one again scale signal more relatively the output of chunk 50 be connected to the control input end of convergent-divergent chunk 52 through a connecting line 51, the input of convergent-divergent chunk 52 is connected to second input of combinational circuit 6 to receive the second compressed signal parameter through connecting line 16, its output is connected to second input of difference chunk 54,56 in order to determine a differential signal according to the compressed signal parameter of mutual convergent-divergent through connecting line 53.The 3rd input of difference chunk 54,56 is connected to the four-input terminal of combinational circuit 6, in order to receive the second compressed signal parameter that is received through connecting line 16 through connecting line 15.Difference chunk 54,56 comprise a difference engine 54 and the chunk of absolute value again 56 that is used for the absolute value of definite this differential signal that is used to produce differential signal, its output is connected to an input of convergent-divergent chunk 57 again, the control input end of convergent-divergent chunk 57 is connected to the 3rd input of combinational circuit 6 again, in order to receive scale signal through connecting line 14.The output of convergent-divergent chunk 57 is connected to the input of an integration chunk 58,59 again, presses time and frequency integrator in order to the absolute value to the differential signal behind the convergent-divergent.Combinational circuit 6 also has a discount chunk 60,61, handles chunk 60 and a multiplication chunk 61 comprising one.The input of handling chunk 60 is connected to connecting line 18 to receive comparison signal through the 5th input of combinational circuit 6, and the output of handling chunk 60 is connected to the first input end of multiplication chunk 61.Integration chunk 58,59 comprises an integrator 58 and time average chunk 59 that is connected in series, and the latter's output is connected to the output 17 of combinational circuit 6 to produce quality signal.The output of integrator 58 is connected to second input of multiplication chunk 61, and the output of multiplication chunk is connected to the input of time average chunk 59.

Below, as described in the international patent application file of institute's reference, the operation of the known devices that is used for definite quality of output signals is described, this output signal is produced by the signal processing circuit of for example codec, and the discount chunk 60,61 that is not shown specifically as shown in Figure 5 in the known devices.

For example the output signal of the signal processing circuit of codec is transported to input 7, and first signal processing circuit 1 is converted to described output signal by time spectrum and the first represented signal parameter of bark spectrum then.This process is to be passed through to be doubled by the represented output signal of time spectrum by the represented window function of time spectrum by first multiplier 20; Then, what drawn transforms to frequency domain by the represented signal of time spectrum by first converter 21, for example adopts the method for fast fourier transform; Then, the absolute value of the signal of being represented by time spectrum and frequency spectrum that is drawn determined by the first absolute value chunk 22, the method that for example adopts square; Then, the signal parameter of being represented by time spectrum and frequency spectrum that is drawn is converted to by time spectrum and the represented signal parameter of bark spectrum by first transducer 23, for example adopts the method based on the repeated sampling of non-linear frequency scope (being also referred to as the bark frequency range); This signal is adjusted to be suitable for auditory function by the first discount device 24 then, and is perhaps filtered, for example adopts the method by the multiplication of its characteristic of bark spectral representation.

According to corresponding mode, for example the output signal of the signal processing circuit of codec is transported to input 8, and secondary signal treatment circuit 2 is converted to described output signal by time spectrum and the represented secondary signal parameter of bark spectrum then.

The first series circuit signal (by first signal parameter of time spectrum and bark spectral representation) receives and is delivered to the first input end of integration chunk 40 again by the first input end of convergent-divergent circuit 3 through connecting line 9, the second series circuit signal (by the secondary signal parameter of time spectrum and bark spectral representation) receives and is delivered to by second input of convergent-divergent circuit 3 again second input of integration chunk 40 through connecting line 10, this integration chunk is pressed frequency integrator to two series circuit signals, the first series circuit signal behind the integration is delivered to the first input end of comparison chunk 41 through first output of integration chunk 40 again, and the second series circuit signal behind the integration is delivered to second input of comparison chunk 41 through second output of integration chunk 40 again.Relatively the series circuit signal of chunk after with two integrations compares and correspondingly produces comparison signal in view of the above, and this comparison signal is transported to the control input end of convergent-divergent chunk 42.The convergent-divergent chunk (that is to say the function through the described comparison signal of conduct that first input end received of connecting line 9 and convergent-divergent circuit 3, increase or reduce the amplitude of the described first series circuit signal) the first series circuit signal (by first signal parameter of time spectrum and bark spectral representation) carry out convergent-divergent, and produce a convergent-divergent the first series circuit signal, be delivered to first output of convergent-divergent circuit 3 through the output of convergent-divergent chunk 42; In this example, second input of convergent-divergent circuit 3 is connected directly to second output of convergent-divergent circuit 3.In this example, convergent-divergent the first series circuit signal and the second series circuit signal be delivered to first compressed chunks 4 and second compressed chunks 5 respectively through convergent-divergent circuit 3 respectively.

What drawn is converted into the first compressed signal parameter by time spectrum and bark spectral representation by first compressed chunks 4 again by first signal parameter behind the convergent-divergent of time spectrum and bark spectral representation.Utilize first adder 30, the 3rd multiplier 32 and the first time-delay chunk 34, can finish the signal parameter by time spectrum and bark spectral representation be multiply by input signal by the bark spectral representation, the signal that reduces by index law for example, drawn then by the signal parameter of time spectrum and bark spectral representation after time-delay, be added to by time spectrum and bark spectral representation signal parameter, what drawn then carries out convolution algorithm by the signal parameter of time spectrum and bark spectral representation by the first non-linear convolution chunk 36 that has by the spread function of bark spectral representation, drawn then by the signal parameter of time spectrum and bark spectral representation again by 37 compressions of first compressed chunks.

According to corresponding mode, be converted into the second compressed signal parameter through second compressed chunks 5 by time spectrum and bark spectral representation by the represented secondary signal parameter of time spectrum and bark spectrum.

First and second compressed signals are delivered to combinational circuit 6 through separately connecting line 13 and 16 respectively, suppose that at present it is the combinational circuit of a standard, do not have the discount chunk 60,61 during institute in detail as Fig. 5.Two compressed signal parameters are by chunk 50 integrations and mutually relatively relatively again, and then relatively chunk 50 produces a scale signal again, and its expression is the mean ratio between two compressed signal parameters for example.Described scale signal again is transported to convergent-divergent chunk 52, here, this chunk convergent-divergent second compressed signal parameter (that is to say, the second compressed signal parameter as the function of scale signal and increase and decrease).Obviously, convergent-divergent chunk 52 also can be used to the convergent-divergent first compressed signal parameter rather than the second compressed signal parameter by the known mode of those skilled in the art, and can be made into two convergent-divergent chunks by the known mode of those skilled in the art again, be used for simultaneously two compressed signal parameters of convergent-divergent mutually.Differential signal is produced by the compressed signal parameter of difference engine 54 from mutual convergent-divergent, and the absolute value of differential signal is again by absolute value chunk 56 is definite again.The signal that is drawn by integrator 58 by the bark spectral integral and by time average chunk 59 by the time spectral integral, produce quality signals by output 17, it is representing for example objective way of the quality of the signal processing circuit of codec.

Owing to adopt convergent-divergent circuit 3, can between by objective quality signal of evaluating according to device of the present invention and the subjective quality signal evaluated by human viewer, obtain good correlation usually.This is based on especially entirely recognizes, bad correlation between the subjective quality signal that objective quality signal that known devices is evaluated and human viewer are evaluated especially comes from such situation, be that human viewer is difficult to accept for the distortion of the distortion of determining than other more, because adopt two compressed chunks, this bad correlation is improved; Also based on recognizing especially, owing to adopted convergent-divergent circuit 3, be better than with any single compressed chunks with the function of two compressed chunks 4 and 5, this has further improved correlation.

Because second input of convergent-divergent circuit 3, or connecting line 10 or connecting line 12, be connected to ratio and determine second input of chunk 43, and the output of convergent-divergent chunk 42, or connecting line 11, be connected to ratio and determine the first input end of chunk 43, ratio determines that chunk 43 just can draw the first series circuit signal behind the convergent-divergent and the mutual ratios of the second series circuit signal, and determine that by ratio the output of chunk 43 produces the scale signal as the function of this mutual ratios, this scale signal is through the 3rd output of convergent-divergent circuit 3 and be delivered to the 3rd input of combinational circuit 6 again through connecting line 14.Described scale signal is transported to the chunk of convergent-divergent again 57 of combinational circuit 6, and it carries out convergent-divergent (that is to say, the amplitude of described absolute value is increased and decreased) in the mode of the function of described scale signal to the absolute value of the differential signal that is derived from difference chunk 54,56.The result, the correlation of having improved obtains further to improve, this is that consequently integration chunk 58,59 is brought into play function better because still be present in the first series circuit signal behind the convergent-divergent and the difference between the second series circuit signal is subjected to discount in combinational circuit.

If difference engine 54 (or absolute value chunk 56) again has the chunk of adjustment again that does not illustrate among the figure, for example reduce a little with the form of subtracter amplitude with differential signal, correlation can be further improved.The amplitude of differential signal preferably reduces in the mode of a kind of function of series circuit signal, as in this example, the mode of the function of its secondary signal parameter after with the compression that is derived from second compressed chunks 5 reduces, and consequently the function of integration chunk 58,59 can be better.This has made that well correlation obtains further improvement again.

Yet, when signal processing circuit for example comprises the dedicated radio link mode, have bad correlation by described device objective quality signal of evaluating and the subjective quality signal of evaluating by human viewer.This problem just can be resolved according to device of the present invention by adopting, and this device has discount chunk 60,61.

According to the operation that is used for determining by the quality of the output signal that signal processing circuit produced of for example codec of the present invention, as mentioned above, existing do to replenish as follows.

Handle chunk 60 from relatively chunk 41 is through connecting line 18 reception comparison signals, this comparison signal for example is promoted to intensity P, wherein 0＜P＜1.Some probable values of P are, for example, and P=0.2 or P=0.3 or P=0.4 or P=0.5.Utilize multiplication chunk 61, comparison signal after the processing multiply by the signal (pressing the bark spectral integral) behind the integration again, the signal that is drawn produces quality signal by time average chunk 59 by the time spectral integral and by output 17 again, and it is representing the objective way of the quality of signal processing circuit.

Because this device that has discount chunk 60,61 is provided, when having significantly difference between two series circuit signals, can impose discount especially at integration chunk 58,59.Because described discount arranged, can between the subjective quality signal that objective quality signal of being evaluated by described device and human viewer are evaluated, obtain good correlation, even transmit through wireless system when the signal that needs definite its quality.

The present invention is based on especially and recognizes, bad correlation between the subjective quality signal that objective quality signal that known devices is evaluated and human viewer are evaluated especially comes from such situation, especially when having significantly difference between two series circuit signals, self-evidently mean poor quality.

Should be noted that, when signal processing circuit comprises asynchronous transfer mode and when the signal of signal processing circuit generation is different from the signal that is derived from reference signal very much, adopts discount chunk 60,61 also can improve correlation.

Each assembly of the first signal processing chunk 1 shown in Figure 2 is as previously described described according to the known mode of those skilled in the art in documents fully.First multiplier 20 by window function respectively aspect time and the amplitude two to being derived from the digital output signal multiplication of the signal processing circuit of codec for example, this window function for example is a so-called cosine square function of being represented by time spectrum, the signal of being represented by time spectrum that is drawn then transforms to frequency domain by first converter 21, for example adopts the method for fast fourier transform; Then, the absolute value of the signal of being represented by time spectrum and frequency spectrum that is drawn determined by the first absolute value chunk 22, the method that for example adopts square.At last, draw the intensity density function of per time/cps.The another kind of method that obtains described signal is the sub-band filter chunk that is used for digital output signal is carried out filtering, after determining absolute value, the signal parameter as the function of time and frequency of the intensity density function that it is per time/cps that this sub-band filter chunk produces a form.First transducer 23 converts the intensity density function of described per time/cps to the intensity density function of per time/bark unit, the method that adopts is for example based on the repeated sampling of non-linear frequency scope, this scope is called the bark scope again, and this conversion is described in the annex A of the 4th piece of documents to some extent; In order to be suitable for the adjustment of auditory function, the intensity density function of described per time of 24 pairs of the first discount devices/bark unit imposes discount, institute's foundation for example be characteristic by the bark spectral representation.

Each assembly of first compressed chunks 4 shown in Figure 3 is as previously described described according to the known mode of those skilled in the art in the 4th piece of documents fully.Intensity density function by the per time/bark unit of auditory function adjustment multiply by a signal by exponential decrease by multiplier 32, for example exp{-T/ τ (z) }.Herein T equal window function length 50%, thereby represent half of a time interval of determining; After this time interval of determining, first multiplier 20 always multiply by output signal a window function of being represented by time spectrum (for example, 40 milliseconds 50% is 20 milliseconds).In this formula, τ (z) is a characteristic value, and it has illustrated its details by the bark spectral representation in Fig. 6 of first piece of documents.The first time-delay chunk 34 is with product time-delay a time span T, perhaps half of the time interval of Que Dinging of this multiplying.36 pairs of signals that spread function provided by the bark spectral representation of the first non-linear convolution chunk carry out convolution, or with the intensity density function of per time/bark unit along the diffusion of bark scope, in the accessories B of the 4th piece of documents, this is described to some extent.37 pairs of first compression units compress according to the signal that form provided by the intensity density function of per time/bark unit representation, for example accomplish to be promoted to intensity α by the intensity density function of per time/bark unit representation, wherein 0＜α＜1.

The assembly of convergent-divergent circuit 3 shown in Figure 4 can make up by the known mode of those skilled in the art.Integration chunk 40 comprises for example two discrete integrators again, be used for separately to two series circuit signal integrations that provide by the bark spectrum, then, form is for example done division one by one for the signal behind 41 pairs of two integrations of comparison chunk of divider, and result or its opposite number of division arithmetic be delivered to unit for scaling 42 as control signal, this unit for scaling for example is a multiplier or divider in form, it with the second series circuit signal times with or divided by result or its opposite number of above-mentioned division arithmetic be equal-sized so that make two series circuit signals with average angle.Ratio determine after chunk 43 receives first and the convergent-divergent that compressed the second series circuit signal, by the diffusion strength density function of per time/bark unit representation, and impose division one by one, to produce scale signal, this scale signal serve as reasons in form division arithmetic result or its opposite number of per time/bark unit representation are that multiplier still is that divider is decided on the structure of unit for scaling 57 again.

Each assembly of first combinational circuit 6 shown in Figure 5 as previously described except that the part of assembly 57 and assembly 54, is described according to the known mode of those skilled in the art in the 4th piece of documents fully.Relatively chunk 50 comprises for example two discrete integrators again, be used for separately the bark that is separated by for example three parts being provided by two series circuit signal integrations that provide, this chunk also comprises for example divider, be used for the signal behind two integrations of each part of bark spectrum is done division arithmetic one by one, and result or its opposite number of division arithmetic be delivered to convergent-divergent chunk 52 as scale signal, this convergent-divergent chunk for example is a multiplier or divider in form, it with each series circuit signal times with or divided by result or its opposite number of above-mentioned division arithmetic be equal-sized at the each several part of bark spectrum with average angle so that make two series circuit signals.In the annex F of the 4th piece of documents, this there is full description.Difference engine 54 draws the difference between the series circuit signal of two mutual convergent-divergents.If this difference is a negative value, just add a constant, if this difference be on the occasion of, just deduct a constant, for example by detect it be less than or greater than null value, add or deduct this constant then.Yet, can be at first determine the absolute value of these differences to deduct a constant from described absolute value then by absolute value chunk 56 again, in this connected mode, clearly do not allow to draw negative final result.In the end in this case, the absolute value chunk should have a subtraction circuit.And, might from difference, deduct (or a part) series circuit signal rather than this constant in a similar fashion, perhaps join with this constant and deducted together.58 pairs of integrators are derived from the signal of unit for scaling 57 by the bark spectral integral again, and 59 pairs of signals that drawn of time average chunk are again by the time spectral integral, and as its result, the quality signal that draws is more little, and the quality of signal processing circuit is just high more.

Signal processing circuit has implication the most widely, in its connected mode, for example can consider various types of audio frequency and/or video equipment.Therefore, signal processing circuit can be a codec, and in this case, input signal is a reference signal, and the quality of output signal should be able to be determined according to input signal.Signal processing circuit also can be an equalizer, and in this connected mode, the quality of output signal determines according to reference signal, and reference signal be according to one existing, be actually desirable equalizer and calculate, or calculate simply.Signal processing circuit even can be a loud speaker, in this case, level and smooth output signal can be used as reference signal, determines voice output quality of signals (in device according to the present invention, carrying out convergent-divergent automatically) again according to this reference signal.Signal processing circuit can also be the computer model of a loud speaker, it is used for according to the design data loud speaker that is arranged at the loud speaker computer model, in this connected mode, the amount of bass output signal of described loud speaker computer model is as reference signal, and its louder volume output signal is as the output signal of signal processing circuit.

Under the situation of the reference signal of calculating, the secondary signal of second series circuit is handled chunk and can be omitted, and can discount calculating in the middle of the reference signal because in fact handle operation that chunk carries out by secondary signal.

Claims (9)

1. be used for determining the device of the quality of an output signal being produced according to a reference signal by a signal processing circuit, this device has:

-the first series circuit, it has the first input end (7) that is used to receive output signal, and this first series circuit has

-the first signal processing chunk (1), it is connected to the first input end of first series circuit, be used to produce one as first signal parameter of the function of time and frequency and

-the first compressed chunks (4), it is connected to the first signal processing chunk, and be used to compress first signal parameter and produce the first compressed signal parameter,

-the second series circuit, it has second input (8) that is used to receive reference signal, and this second series circuit has

-secondary signal is handled chunk (2), and it is connected to second input (8) of second series circuit, be used to produce as the secondary signal parameter of the function of time and frequency and

-the second compressed chunks (5), it is connected to secondary signal and handles chunk, and be used to compress the secondary signal parameter and produce the second compressed signal parameter,

-one combinational circuit (6), it is connected to first output (13) of first series circuit and second output (16) of second series circuit, is used to produce a quality signal, and this combinational circuit has

-one difference chunk (54,56), it is connected to two compressed chunks (4,5), be used for according to the compressed signal parameter determine a differential signal and

-one integration chunk (58), it is connected to the difference chunk, is used for by frequency the differential signal integration,

-one time average chunk (59) is used for the differential signal behind the integration by the time integration producing quality signal,

-one convergent-divergent circuit (3), its input are connected between the output of signal processing chunk (1,2) and the input (11,12) that its output is connected in two compressed chunks (4,5), and this convergent-divergent circuit has

-one integration chunk (40) again, be used for by frequency to first signal parameter and secondary signal parameter integration and

-one chunk (41) relatively, it is connected to the described chunk of integration again, and be used for the signal parameter behind two integrations of comparison and at least one parameter signals carried out convergent-divergent by comparative result,

It is characterized in that this device also comprises

Handle chunk (60) for-one, be used for process source from comparison signal of chunk relatively and

-one multiplication chunk (61), comprising

-first input end is connected to an output of the first signal processing chunk (60),

-the second input, be connected to integration chunk (58) output and

-one output is connected to an input of time average chunk (59).

2. according to the device of claim 1, it is characterized in that convergent-divergent circuit (3) has

-one unit for scaling (42), comprising

-one input is connected to the output (9) of the first signal processing chunk (1),

-one output (11), be connected to first compressed chunks (4) input and

-one control input end is connected to an output of comparison chunk (41), is used for according to comparative result first signal parameter being carried out convergent-divergent.

3. according to the device of claim 1 or 2, it is characterized in that, described second input (8) is the calculating reference signal that is used to receive as the function of time and frequency, and second compressed chunks (5) is connected directly to second input (8) to compress the reference signal of described calculating.

4. be used for determining the method for quality of an output signal being produced according to a reference signal by a signal processing circuit, this method may further comprise the steps:

-according to first signal parameter of output signal generation as the function of time and frequency,

-according to the secondary signal parameter of reference signal generation as the function of time and frequency;

-first signal parameter and secondary signal parameter are pressed frequency integrator,

-first and second signal parameters behind the integration are compared, and produce comparison signal,

-according to described comparison signal in first and second signal parameters at least one carried out convergent-divergent,

-compression first signal parameter and secondary signal parameter,

-determine a differential signal according to the signal parameter that has compressed, and

-to the differential signal first step by frequency integrator and second step by the time integration producing a quality signal,

It is characterized in that this method also comprises step:

-handle comparison signal and

-differential signal of the first step after by the integration that frequency integrator drew be multiply by comparison signal after the processing to produce a consequential signal, then in second step to consequential signal by the time integration.

5. according to the method for claim 4, it is characterized in that the method comprising the steps of:

-according to comparative result first signal parameter is carried out convergent-divergent.

6. according to the method for claim 5, it is characterized in that the step of handling comparison signal comprises the step that comparison signal is promoted to intensity P, wherein 0＜P＜1.

7. according to claim 4,5, or 6 method, it is characterized in that the step of described generation secondary signal parameter comprises step:

-receiving calculating reference signal as the function of time and frequency, described secondary signal parameter comprises the reference signal of described calculating.

8. according to the method for claim 4, it is characterized in that the step of handling comparison signal comprises the step that comparison signal is promoted to intensity P, wherein 0＜P＜1.

9. method according to Claim 8 is characterized in that the step of described generation secondary signal parameter comprises step:

-receiving calculating reference signal as the function of time and frequency, described secondary signal parameter comprises the reference signal of described calculating.

Images