JP2013246107A - Signal processing device and signal processing method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a signal processing device which outputs an estimated value of a level of a desired signal from a noise signal component and an input signal which is a desired signal.SOLUTION: A signal processing device includes an A/D conversion part 10 for subjecting an input signal to A/D conversion to generate an A/D converted input signal A, a first PDF processing part 12 for subjecting the A/D converted input signal A to PDF processing indicating an amplitude distribution to generate a first PDF signal C, a filter 13 for extracting, from the A/D converted input signal, a predetermined band signal B which is a predetermined frequency including a noise signal component and higher than a desired signal, a second PDF processing part 14 for subjecting the predetermined band signal to PDF processing indicating an amplitude distribution to generate a second PDF signal D, a correlation value calculation part 15 for subjecting the first PDF signal and the second PDF signal to obtain a correlation value, and an estimated value calculation part 16 for estimating the level of the desired signal on the basis of the maximum correlation value to output is as an estimated value.

Description

  The present invention relates to a signal processing apparatus and a signal processing method for sensor signals used between various sensors related to cloud computer networks and M2M communication and the network.

  A sensor usually senses a physical quantity that reacts to a measurement target and converts it into an electrical signal such as a voltage or current that can be easily processed. This electric signal becomes a means for signal processing when quantitatively measuring a predetermined physical quantity. The sensor signal converted into the electric signal has time frequency characteristics according to the measurement object.

  In the process of converting the sensed physical quantity into an electrical signal, the signal is output as an electrical signal in which unnecessary noise other than a physical phenomenon is superimposed on the signal. For example, physical quantities such as thermometers and scales are converted to a low frequency range near DC, but ambient noise such as ambient temperature and wind, and supply voltage and current noise for conversion to electrical signals are It is normal to have the above relatively high frequency fluctuations. It is known that processing for extracting a signal in a low frequency region near DC such as a physical quantity such as a thermometer and a scale requires a long processing time due to the uncertainty of time frequency.

JP 2008-268068 A

  As a research example using noise generation learning information in order to reduce the processing time, there is also a report in which a sensor signal is A / D converted and a low-pass filter using digital signal processing is applied to noise removal (for example, patent document) 1). The low-pass filter can remove even low-frequency noise components by setting the cutoff frequency width low.

  For example, in the conventional weighing device of Patent Document 1 shown in FIG. 11, a vibration period calculation unit 81 that calculates a period of steady low-period vibration possessed by the conveyance unit and generates a trigger corresponding to the period, Phase and amplitude to generate a vibration correction waveform that approximates the waveform of an unloaded weighing signal output by the weighing means when there is no measurement object in the means from the basic waveform function having the period calculated by the vibration period calculation unit A waveform condition storage unit 82 that calculates and stores a waveform generation condition including the waveform signal, and a measurement signal from a condition for generating a vibration waveform stored in the waveform condition storage unit with reference to a trigger generated from the vibration period calculation unit A correction waveform generator 83 that generates a correction signal for correction, and a difference between a measurement signal (input signal X) output from a measurement unit (not shown) and a correction signal output from the correction waveform generation unit. And a correction unit 84 for calculating a weight value of the object. In Patent Document 1, a low-pass filter (not shown) is used to remove a high-frequency component from the measurement signal, and the peak value of the AC component of the measurement signal filtered by the low-pass filter is detected to calculate the vibration period. Yes.

  As described above, research has been conducted to remove a noise component from a sensor signal by a low-pass filter and extract the signal component with high speed and high accuracy. However, the calculation method is based on a linear operation.

  However, in signal processing using linear calculation, the analysis and evaluation procedures are uniquely determined, but the filter response time based on the processing delay time and the uncertainty of the time frequency is accompanied by fundamental restrictions. Specifically, when the cut-off frequency of the low-pass filter is set low, there is a problem that it is necessary to lengthen the sensing period, and the response speed is slowed accordingly.

  In addition to the low-pass filter, a band-pass filter, a Hilbert transform method, and the like are used. With a bandpass filter, for example, when a low-pass bandpass filter having a passband of 5 Hz to 40 Hz is designed, it is extremely difficult to attenuate a DC signal. In the Hilbert transform method, the direct current can be attenuated, but the low-frequency signal is attenuated. Further, when the DC component is suppressed, the number of taps of the digital filter is remarkably increased and the processing delay is remarkably increased.

  These filtering methods extract a low-frequency component, which is a target signal component, using locality and bias of noise of amplitude frequency characteristics and phase frequency characteristics of an electrical signal as a sensor signal. When these processes are considered as a system, the calculation model is a process that processes the frequency phase component of the signal by convolving the sensor signal (frequency phase component) with the impulse response of the processing system in the time domain. It can be formulated. This convolution operation has the property that the lower the frequency of the low frequency component to be extracted, the more the desired low frequency can not be extracted unless the number of response points of the impulse response is increased. That is, the processing time is long, and the processing time is represented by (number of response points of impulse response) * (sampling time T) / 2. Although this processing delay uses an analog method, τ time is proportional to exp (−t / τ) in terms of the time constant τ of the processing system which is the main mode of analog processing. Therefore, in the low-frequency processing, τ increases, and the delay time until a normal signal is transmitted is increased.

  If such a processing delay is allowed for the sensor signal, the nature of the signal that changes from time to time is overlooked, and there is a possibility that the acquisition opportunity of the signal that fluctuates with time may be lost, so it is desirable to solve this processing delay. It was.

  Therefore, the present invention has been made in view of the above problems, and relates to a technique for providing a signal processing apparatus for outputting an estimated value of a desired signal level from an input signal composed of a noise signal component and an analog desired signal. An object of the present invention is to provide a technique for processing in high accuracy and in a short time.

In order to achieve the above object, the signal processing device (100) according to claim 1 comprises:
A signal processing apparatus for outputting an estimated value of a level of a desired signal from an input signal (X) composed of a noise signal component and an analog desired signal (Y),
An A / D converter (10) for A / D converting the input signal to generate an A / D converted input signal (A);
A first PDF processing unit (12) for generating a first PDF signal (C) by performing PDF (Probability Density Function) processing indicating amplitude distribution on the A / D converted input signal (A); ,
A filter (13) for extracting a predetermined band signal (B) that includes the noise signal component and is a predetermined frequency band higher than the desired signal from the A / D converted input signal;
A second PDF processing unit (14) for performing a PDF process indicating an amplitude distribution on the predetermined band signal to generate a second PDF signal (D);
A correlation value calculation unit (15) for calculating a correlation value by performing a cross-correlation operation between the first PDF signal and the second PDF signal;
An estimated value calculation unit (16) that estimates the level of the desired signal based on the maximum correlation value and outputs the estimated value as an estimated value.

に基づいて前記相関値であるv(β)を最大とする前記平均値パラメータβを求め、当該βを推定値とすることを特徴とする。 The signal processing device according to claim 2 is the signal processing device (100) according to claim 1,
The estimated value calculation unit includes a lower limit value (PL_1) and an upper limit value (PM_1) of an amplitude value present in the first PDF signal, and a lower limit value (PL_2) of an amplitude value present in the second PDF signal. The upper limit value (PM_2) is obtained,
A lower limit value (PL_12) which is a lower limit among the lower limit value of the amplitude value present in the first PDF signal or the lower limit value of the amplitude value present in the second PDF signal;
A maximum upper limit value (PM_12) which is a maximum upper limit value among the upper limit value of the amplitude value present in the first PDF signal or the upper limit value of the amplitude value present in the second PDF signal;
An average value parameter (β) including candidates of estimated values to be obtained;
A random variable (k) representing the amplitude component;
A function (f1) of the first PDF signal C;
When the function (f2) of the second PDF signal D is used,

The average value parameter β that maximizes the correlation value v (β) is obtained based on the above and β is used as an estimated value.

In order to achieve the above object, the signal processing method according to claim 3 comprises:
A signal processing method for outputting an estimated value of a level of a desired signal from an input signal (X) comprising a noise signal component and an analog desired signal (Y),
A / D conversion step (S101) for A / D converting the input signal to generate an A / D converted input signal (A);
A first PDF processing step (S102) for generating a first PDF signal (C) by performing PDF (Probability Density Function) processing indicating amplitude distribution on the A / D converted input signal;
A band extraction step (S103) for extracting a predetermined band signal (B) that includes the noise signal component and is a predetermined frequency band higher than the desired signal from the A / D converted input signal;
A second PDF processing step (S104) for generating a second PDF signal (D) by performing a PDF process indicating an amplitude distribution on the predetermined band signal;
A correlation value calculating step (S105) for calculating a correlation value by performing a cross-correlation operation between the first PDF signal and the second PDF signal;
And an estimated value calculating step (S107) for estimating the level of the desired signal based on the maximum correlation value and outputting the estimated value as an estimated value.

に基づいて前記相関値であるv(β)を最大とする前記平均値パラメータβを求め、当該βを推定値とすることを特徴とする。 The signal processing method according to claim 4 is the signal processing method according to claim 3,
A lower limit value (PL_1) and an upper limit value (PM_1) of an amplitude value present in the first PDF signal, and a lower limit value (PL_2) and an upper limit value (PM_2) of an amplitude value present in the second PDF signal. A further upper and lower limit value calculating step (S106) to be obtained;
In the estimated value calculating step, the lower limit value of the amplitude value existing in the first PDF signal (or the lower limit value (PL — 12) which is the lowest limit among the lower limit values of the amplitude value existing in the second PDF signal). )When,
A maximum upper limit value (PM_12) which is a maximum upper limit value among the upper limit value of the amplitude value present in the first PDF signal or the upper limit value of the amplitude value present in the second PDF signal;
An average value parameter (β) including candidates of estimated values to be obtained;
A random variable (k) representing the amplitude component;
A function (f1) of the first PDF signal C;
When the function (f2) of the second PDF signal D is used,

The average value parameter β that maximizes the correlation value v (β) is obtained based on the above and β is used as an estimated value.

  According to the signal processing apparatus and the signal processing method of the present invention, an input signal is A / D converted to generate an A / D converted input signal, and the first PDF signal C is generated from the A / D converted input signal. On the other hand, a second PDF signal is generated from a predetermined band signal that includes a noise signal component from the A / D converted input signal and is a predetermined frequency band higher than the desired signal. The correlation value is obtained by correlation processing, the level of the desired signal is estimated based on the maximum value of the correlation value, and the result is output as an estimated value. Since real-time processing is performed to extract the results, the desired signal that could not be obtained without taking a long time with the conventional signal processing apparatus and signal processing method that calculates the results by sequentially processing from the initial stage of the signal. Level of the estimated value of, the signal processing device and signal processing method of the present invention can be calculated in a short time, and can be calculated with extremely high accuracy.

  In addition, since the signal processing apparatus and the signal processing method of the present invention process each measurement section, the time required for the estimated value calculation process after obtaining the PDF signal of the measurement section is the CPU (central processing unit) to be used. As the calculation speed increases, the processing can be performed in a shorter time. Therefore, the delay that has occurred in principle in the conventional signal processing apparatus and signal processing method can be eliminated extremely effectively.

  Furthermore, the signal processing apparatus and the signal processing method of the present invention use the upper limit value and the lower limit value of the amplitude values respectively present in the first PDF signal and the second PDF signal, and correlate PDF signals. Since the estimated value having the maximum value is obtained, the estimated value of the level of the desired signal can be extracted in a short time and with extremely high accuracy.

It is a schematic block diagram which shows the apparatus structure of the signal processing apparatus 100 which concerns on this invention. 2 is a schematic block diagram of PDF processing units 12 and 14 according to the present invention. FIG. It is a figure for demonstrating the probability density function PDF. It is a figure for demonstrating the process which calculates | requires the estimated value which is the maximum value of a correlation value. It is a figure for demonstrating the process which determines the estimated value which is the maximum value of a correlation value. It is a flowchart which shows an example of the software processing operation | movement which concerns on this invention. 4 is a diagram illustrating an example of a waveform of an input signal X. FIG. 6 is a diagram illustrating an example of a waveform of a first PDF signal C. FIG. It is a figure which shows the example of a waveform of the 2nd PDF signal D which added the average value parameter (beta). It is a figure which shows the example of a waveform of the correlation result of the 1st PDF signal C and the 2nd PDF signal D. It is a schematic block diagram which shows the apparatus structure of the conventional weighing | measuring apparatus.

  Hereinafter, embodiments for carrying out the present invention will be described in detail with reference to the accompanying drawings. It should be noted that the present invention is not limited by this embodiment, and all other forms, examples, operational techniques, etc. that can be implemented by those skilled in the art based on this form are included in the scope of the present invention. .

  In the present invention, as an example of a sensed physical quantity, for example, the amplitude of an input signal X output from a mechanical quantity sensor loaded with a measurement object is used as a random variable, and its probability density function PDF (Probability Density Function) or cumulative probability distribution is used. Since it is a signal processing apparatus that performs processing by obtaining APD (Amplitude Probability Distribution), first, a probability density function PDF and a cumulative probability distribution APD will be briefly described.

  The probability density function (also referred to as probability density distribution) PDF takes the time signal as shown in FIG. 3 as an example, and the occurrence frequency of the amplitude value extracted at an appropriate sampling time within the measurement time T can be obtained with the required measurement accuracy. It is obtained by counting for each quantized amplitude level.

  When this PDF value is integrated (accumulated) with a random variable, it becomes an APD value. However, when the definition of APD value “time probability that the envelope signal of signal amplitude exceeds a certain threshold level” is adopted, signal amplitude r (t) For the measurement time T, the occupation time Wi (xk) at the signal amplitude threshold R is cumulatively added, and when divided by T, the APD value at xi is expressed by the following equation (1). It is required as follows.

APD (xi) = ΣWi (xk) / T (1)
Here, the symbol Σ represents the total sum from i = 1 to N (xk).

  Conversely, if this APD value is subjected to a difference calculation with respect to a random variable, the PDF value can be obtained.

[Configuration example and operation example of signal processing apparatus]
Based on the above preparation, an operation example of the signal processing apparatus 100 of the present invention will be described. As an example of the sensed physical quantity, the signal processing apparatus 100 of this example superimposes the desired signal Y, which is an analog substantially DC component output from a mechanical quantity sensor loaded with a measurement object, on the desired signal Y, for example. It is an apparatus for outputting an estimated value of the level of a desired signal Y for an input signal X including a harmful noise component. The estimated value of the level of the desired signal Y extracted by the signal processing apparatus 100 of the present invention can be applied to an apparatus for measuring the weight of a measurement object, for example.

  First, the apparatus configuration of the signal processing apparatus 100 according to the present invention will be described with reference to FIG. The signal processing apparatus 100 of this example includes an A / D conversion unit 10, a delay unit 11, a first PDF processing unit 12, a filter 13, a second PDF processing unit 14, a correlation value calculation unit 15, and an estimated value calculation unit 16. It has.

  The signal processing apparatus 100 according to the present invention is realized by hardware such as a DSP (Digital Signal Processor) or FPGA (Field Programmable Gate Array) and its peripheral devices.

  The A / D converter 10 samples the analog input signal X at a predetermined sampling period, converts the analog signal to a digital signal, and performs A / D conversion including almost all signal bands of the input signal X. An input signal A is generated.

  The delay device 11 is a delay device that delays the A / D converted input signal A while maintaining its phase characteristics for a time equivalent to the time required for the filtering process of the filter 13. The output signal of the delay device 11 is a signal obtained by A / D converting a signal including a noise signal component and a substantially DC component of the input signal X composed of the desired signal Y. In the present invention, substantially direct current refers to alternating current with a frequency as low as about 20 Hz from direct current.

  The A / D converted input signal A delayed by the delay unit 11 is input to the first PDF processing unit 12. The first PDF processing unit 12 performs a PDF (Probability Density Function) process, and calculates a first PDF signal C that is a probability density function of the input signal A subjected to A / D conversion.

  On the other hand, the filter 13 is a filter that extracts a predetermined band signal B that is a predetermined band from the A / D converted input signal A, and performs a filtering process by digital processing. The predetermined band here is a band that includes a noise signal component included in the input signal X and does not include a substantially DC component, that is, a band that does not include the desired signal Y. The frequency characteristic of the filter 13 is configured to be a high-pass filter or a band-pass filter. In the case of a high-pass filter, the pass band of the filter 13 is, for example, a frequency of 20 Hz or more, and in the case of a band-pass filter, the pass band is, for example, a frequency of 20 Hz to 400 Hz. When the high-pass filter is adopted, the band to be processed becomes wider and the amount of computation increases than when the band-pass filter is adopted. Therefore, the amount of computation is smaller when the band-pass filter is adopted. The filter 13 is configured to be a linear phase filter.

  The predetermined band signal B subjected to the filtering process performed by the filter 13 is input to the second PDF processing unit 14. The second PDF processing unit 14 performs a PDF process to calculate a second PDF signal D that is a probability density function of the predetermined band signal B.

  An example of the internal configuration of the first PDF processing unit 12 and the second PDF processing unit 14 is shown in FIG. More specifically, referring to FIG. 2, the A / D converted input signal A delayed by the delay unit 11 is input to the normalization processing unit 30 of the first PDF processing unit 12, and negative data or the like is input. When the error occurs, a shift process to convert the data into positive data as a whole is performed, the normalized value is logarithmically converted by the logarithmic conversion unit 31, and the logarithmic value is quantized in units of accuracy to be measured. The logarithmic conversion unit 31 is for facilitating the processing of a large number of data, and this configuration can be omitted. The same operation is performed for the second PDF processing unit 14.

  The logarithmic value is input to the PDF calculation unit 32. The PDF calculation unit 32 obtains each appearance frequency (histogram) of the input logarithmic value, and divides this by the total number of data input at a certain time Tp, thereby obtaining a first PDF that is a normalized probability density function. The signal C is calculated. The same operation is performed for the second PDF processing unit 14.

  By the way, the desired signal Y is equivalent to a statistical average value not including a noise signal component. The profile of the PDF waveform of the signal of the A / D converted input signal A (here, the first PDF signal C) and a predetermined band are extracted from the A / D converted input signal A, and the statistical average value is excluded. It is considered that the profile of the PDF waveform shape (here, the second PDF signal D) of the predetermined band signal B composed of the noise signal component is the same.

  From this, when the PDF waveform shape comparison between the profile of the first PDF signal C and the profile of the second PDF signal D is performed, a measure of the identity of the shape is defined, and the value is maximized. That is, the level of the amplitude of the desired signal Y can be estimated by obtaining an estimated value that maximizes the scale value of the identity of the PDF curve. The identity of the shapes when the PDF waveform shapes are compared can be determined by calculating a correlation value by performing a cross-correlation operation.

  The obtained first PDF signal C and second PDF signal D are input to the correlation value calculation unit 15 in order to perform PDF waveform shape comparison. The correlation value calculation unit 15 includes a first PDF signal C that is a probability density function obtained by the first PDF processing unit 12 and a second PDF that is a probability density function obtained by the second PDF processing unit 14. Based on the PDF signal D, a cross-correlation operation is performed to calculate a correlation value.

  The operation of the estimated value calculation unit 16 will be described. Since the estimated value to be obtained is a level of amplitude proportional to the desired signal Y including the substantially DC component of the first PDF signal C, if the average value parameter to be obtained for the second PDF signal D is β, the cross-correlation calculation is performed. The signal used for is the average value parameter β + second PDF signal D. From this, the estimated value to be obtained indicating the identity of the shape when the PDF waveform shape comparison is performed is the maximum value of the correlation value between the first PDF signal C and the second PDF signal D. .

  A process for obtaining an estimated value that is the maximum correlation value will be described with reference to FIG. The lower limit value and the upper limit value of the level where the PDF signal exists are used. In the case of the first PDF signal C, as shown in FIG. 4A, the lower limit value of the level where the PDF signal exists is PL_1, and the upper limit value is PM_1. In the case of the second PDF signal D, as shown in FIG. 4B, the lower limit value of the level where the PDF signal exists is PL_2, and the upper limit value is PM_2.

  Next, the lowest lower limit PL_12 which is the lowest of the lower limit PL_1 of the amplitude value present in the first PDF signal or the lower limit PL_2 of the amplitude value present in the second PDF signal, and the first PDF signal Within an interval between the upper limit value PM_1 of the amplitude value existing and the upper limit value PM_2 of the upper limit value PM_2 of the amplitude value existing in the second PDF signal, the average value parameter β is set to nΔ steps (n is Each time, the function f1 of the first PDF signal C and the function f2 of the second PDF signal D are multiplied to obtain a correlation value, and an average value parameter that maximizes the correlation value Let β be the estimated value to be obtained.

  Next, the process of determining the estimated value that is the maximum correlation value will be further described with reference to FIG. In equation (2), one of the average value parameters β is a candidate value of the estimated value to be obtained, and the average value parameter β that maximizes ν (β) is the estimated value to be obtained. The accuracy for obtaining the average value parameter β is Δ, and the initial value is β0. The initial value β0 may be an intermediate value of the first PDF signal C or the second PDF signal D, for example. A search for β0 + −nΔ is performed, and processing for driving into the range of + −Δ is performed so that the correlation value becomes maximum. In this method, the estimated value is determined with β0 + −nΔ as the most probable value.

……（２） The cross-correlation operation ν uses the following equation (2). It is assumed that the first PDF signal C is a function f1, and the second PDF signal D is a function f2. Here, k is a random variable and represents an amplitude component.

(2)

  The estimated value estimated by the estimated value calculation unit 16 has a substantially DC component indicating the level of the amplitude of the desired signal Y. Based on the above processing, the estimated value calculation unit 16 estimates the level of the desired signal Y based on the maximum value of the correlation value and outputs it as an estimated value. Thereby, the signal processing device 100 is configured.

  Moreover, the estimated value of the level of the several desired signal Y is output continuously, The desired signal Y without attenuation is extracted from the input signal X by connecting each estimated value with the estimated value before and behind that estimated value. The signal processing apparatus 100 that extracts the desired signal Y can be configured.

  As described above, according to the signal processing device 100 of the present invention, since the measurement result is extracted for each section regardless of where the process is started, the estimated value of the level of the desired signal Y is obtained. Can be calculated in a short time.

  Furthermore, according to the signal processing apparatus 100 of the present invention, the PDF signals can be obtained by using the upper limit value and the lower limit value of the amplitude values respectively present in the first PDF signal C and the second PDF signal D. Therefore, the estimated value of the level of the desired signal Y can be extracted in a short time and with extremely high accuracy.

[Example of processing operation in signal processing apparatus]
Next, an example of the processing operation relating to the signal processing apparatus of the present invention will be described with reference to FIG.

  First, an analog input signal X is sampled at a predetermined sampling period and converted from an analog signal to a digital signal, and an A / D converted input signal A including almost all signal bands of the input signal X is generated. (S101).

  Next, the input signal A subjected to A / D conversion is subjected to PDF (Probability Density Function) processing by the first PDF processing, and the first PDF which is a probability density function of the input signal A subjected to A / D conversion. A signal C is generated (S102).

  While generating the first PDF signal C, the A / D converted input signal A including substantially the entire signal band of the input signal X includes a noise signal component and has a predetermined frequency band higher than the desired signal. A predetermined band signal B is extracted (S103). Specifically, a filtering process is performed by digital processing to perform a high-pass filter operation or a band-pass filter operation.

  Next, PDF processing is performed on the predetermined band signal B by the second PDF processing to generate a second PDF signal D that is a probability density function of the predetermined band signal B (S104).

  Based on the first PDF signal C and the second PDF signal D, a cross-correlation operation is performed to calculate a correlation value (S105).

  Next, the lower limit PL_1 and the upper limit PM_1 of the amplitude value existing in the first PDF signal, and the lower limit PL_2 and the upper limit PM_2 of the amplitude value existing in the second PDF signal are obtained (S106).

……（２） In the estimated value calculation step, the lowest lower limit PL_12 that is the lowest of the lower limit PL_1 of the amplitude value present in the first PDF signal or the lower limit PL_2 of the amplitude value present in the second PDF signal, and the first An upper limit value PM_1 of the amplitude value existing in the PDF signal or an upper limit value PM_2 of the upper limit value PM_2 of the amplitude value existing in the second PDF signal, and an average value parameter β including the candidates of estimated values to be obtained And a random variable k representing the amplitude component, a function f1 of the first PDF signal C, and a function f2 of the second PDF signal D, and performing a cross-correlation operation ν using the following equation (2): .

(2)

  Based on the maximum correlation value, an estimated value of the level of the desired signal Y is calculated (S107).

  As described above, an estimated value of the level of the desired signal Y without attenuation is calculated from the input signal X, and the signal processing is completed. In the signal processing of the present invention, a delay process corresponding to the delay unit 11 may be used as appropriate.

[Example of processing result by signal processor]
Next, an example of processing results of an actual signal processing device will be shown. The input signal X composed of a noise signal component and an analog desired signal to which the present invention is actually applied is a sensor signal, for example, an electric signal output from a mechanical quantity sensor. Specifically, for example, a strain sensor such as a load cell, in which a strain sensor, which is a kind of mechanical quantity sensor, is attached to the back of a cantilever structure base. The strain sensor converts a mechanical quantity, which is a mechanical deflection due to a load caused by the weight of an object to be measured, into an electrical signal by, for example, a piezoelectric phenomenon.

  As described above, the strain sensor outputs an electrical signal due to the load caused by the weight of the measurement object, and the desired signal Y to be obtained is a substantially direct current component due to the load caused by the weight of the measurement object.

  However, in addition to the desired signal Y to be obtained, the natural vibration noise of the mechanical system to which the strain sensor is attached and the vibration system noise generated at the base of the strain sensor are detected by the strain sensor, and the electrical noise from the power source is mixed in as a result. A signal X is generated.

  In order to clarify and evaluate the measurement object, simulation was performed using simulated data. FIG. 7 shows a waveform example of the input signal X. The simulation data used will be described as an example in which the weight of the object to be measured is about 582.84 g, and the true value of the DC amplitude level of the desired signal Y is 64535/2 (90.1753 dB in the case of common logarithmic conversion). An AC signal having a frequency of 10 Hz and an amplitude of 1000 is superimposed on the desired signal Y as a vibration component corresponding to the noise signal component. That is, the input signal X is a composite waveform of the desired signal Y and the noise signal component.

  FIG. 8 shows a waveform example of the first PDF signal C obtained by performing PDF processing on the A / D converted input signal A in the 700 to 750 msec measurement section of the input signal X shown in FIG. In addition, the measurement section here is a section in which 700 to 750 msec has elapsed from the load start timing when the measurement object is loaded onto the strain sensor, for example, and the natural vibration of the scale generated at the load start timing has converged to a certain extent. It can be regarded as a section.

  By the way, the value to be obtained is a level of a substantially direct current amplitude proportional to the desired signal Y, but the direct current component corresponding to the average value parameter β does not exist in the second PDF signal D. Therefore, when performing cross-correlation between the first PDF signal C and the second PDF signal D, the signal used for cross-correlation is β + second PDF signal D, and a DC shift that is DC-shifted by an average value parameter β. Correction is necessary.

  In consideration of the above-described DC shift correction, the second PDF signal D obtained by performing PDF processing on the predetermined band signal B corresponding to the measurement interval of 700 to 750 msec of the input signal X shown in FIG. FIG. 9 shows a waveform example to which 2 (90.3088 dB in the case of common logarithmic conversion) is added.

  FIG. 10 shows a waveform example of a result of cross-correlation calculation between the first PDF signal C and the second PDF signal D. The correlation value takes the maximum value when the waveform of the first PDF signal C shown in FIG. 8 matches the shape of the second PDF signal D waveform added with the average value parameter β shown in FIG. . This maximum value is 64866/2 (90.2197 dB when converted to common logarithm), and is very close to the value of 64535/2 (90.1753 dB when converted to common logarithm), which is the true value of the DC amplitude level of the desired signal Y. It can be seen that the estimated value is calculated with high accuracy.

  Moreover, it was tried whether the signal processing apparatus and signal processing method of this invention can track the fluctuation | variation of the desired signal Y. FIG. Here, the waveform example of the input signal X including the desired signal Y in FIG. 7 can be changed by adjusting the DC component by multiplication. If the processing result follows the DC component to be multiplied, it can be verified that the fluctuation of the desired signal Y is followed. As a result of this verification, when the value of the DC component to be multiplied is, for example, 64535, the processing result indicates 64866/2 (90.2197 dB when converted to common logarithm), and the processing result is, for example, 66535 as the value of the DC component to be multiplied. Indicates 66768/2 (90.848 dB in the case of common logarithmic conversion), and completely follows the value of the DC component multiplied by the input signal X.

  Therefore, according to the signal processing device and the signal processing method of the present invention, it is possible to calculate the estimated value of the level of the desired signal Y with extremely high accuracy, and even when the input signal X including the desired signal Y is varied. Since the tracking can be performed very accurately, the estimated value can be accurately calculated regardless of the level of the desired signal Y.

  As described above, the signal processing apparatus and the signal processing method according to the present invention, as an example of a sensed physical quantity, for example, a desired signal Y that is an analog substantially DC component output from a mechanical quantity sensor loaded with a measurement object. When calculating the estimated value of the level of the desired signal Y from the input signal X including the harmful noise component superimposed on the desired signal Y, there is an effect that the estimated value can be calculated in a short time and with extremely high accuracy. The signal processing apparatus and signal processing method of the present invention can be applied to an apparatus for measuring the weight of a measurement object, for example.

DESCRIPTION OF SYMBOLS 10 ... A / D conversion part 11 ... Delay device 12 ... 1st PDF process part 13 ... Band pass filter 14 ... 2nd PDF process part 15 ... Correlation value calculation part 16 ... Estimated value calculation part 30 ... Normalization process part 31 ... Logarithmic conversion unit 32 ... PDF calculation unit 80 ... A / D conversion unit 81 ... Vibration period calculation unit 82 ... Waveform condition storage unit 83 ... Correction waveform generation unit 84 ... Correction unit 85 ... Determination unit 86 ... Operation display unit 100 ... Signal processing device 200 ... weighing device (conventional example)
A ... A / D converted input signal B ... predetermined band signal C ... first PDF signal D ... second PDF signal X ... input signal Y ... desired signal

Claims (4)

A signal processing apparatus for outputting an estimated value of a level of a desired signal from an input signal (X) composed of a noise signal component and an analog desired signal (Y),
An A / D converter (10) for A / D converting the input signal to generate an A / D converted input signal (A);
A first PDF processing unit (12) for generating a first PDF signal (C) by performing PDF (Probability Density Function) processing indicating amplitude distribution on the A / D converted input signal (A); ,
A filter (13) for extracting a predetermined band signal (B) that includes the noise signal component and is a predetermined frequency band higher than the desired signal from the A / D converted input signal;
A second PDF processing unit (14) for performing a PDF process indicating an amplitude distribution on the predetermined band signal to generate a second PDF signal (D);
A correlation value calculation unit (15) for calculating a correlation value by performing a cross-correlation operation between the first PDF signal and the second PDF signal;
A signal processing apparatus comprising: an estimated value calculation unit (16) that estimates a level of the desired signal based on the maximum value of the correlation value and outputs the estimated value as an estimated value. The estimated value calculation unit includes a lower limit value (PL_1) and an upper limit value (PM_1) of an amplitude value present in the first PDF signal, and a lower limit value (PL_2) of an amplitude value present in the second PDF signal. The upper limit value (PM_2) is obtained,
A lower limit value (PL_12) which is a lower limit among the lower limit value of the amplitude value present in the first PDF signal or the lower limit value of the amplitude value present in the second PDF signal;
A maximum upper limit value (PM_12) which is a maximum upper limit value among the upper limit value of the amplitude value present in the first PDF signal or the upper limit value of the amplitude value present in the second PDF signal;
An average value parameter (β) including candidates of estimated values to be obtained;
A random variable (k) representing the amplitude component;
A function (f1) of the first PDF signal C;
When the function (f2) of the second PDF signal D is used,

The signal processing apparatus according to claim 1, wherein the average value parameter β that maximizes the correlation value v (β) is obtained based on and the β is an estimated value. A signal processing method for outputting an estimated value of a level of a desired signal from an input signal (X) comprising a noise signal component and an analog desired signal (Y),
A / D conversion step (S101) for A / D converting the input signal to generate an A / D converted input signal (A);
A first PDF processing step (S102) for generating a first PDF signal (C) by performing PDF (Probability Density Function) processing indicating amplitude distribution on the A / D converted input signal;
A band extraction step (S103) for extracting a predetermined band signal (B) that includes the noise signal component and is a predetermined frequency band higher than the desired signal from the A / D converted input signal;
A second PDF processing step (S104) for generating a second PDF signal (D) by performing a PDF process indicating an amplitude distribution on the predetermined band signal;
A correlation value calculating step (S105) for calculating a correlation value by performing a cross-correlation operation between the first PDF signal and the second PDF signal;
A signal processing method comprising: an estimated value calculating step (S107) that estimates a level of the desired signal based on the maximum value of the correlation value and outputs the estimated signal as an estimated value. A lower limit value (PL_1) and an upper limit value (PM_1) of an amplitude value present in the first PDF signal, and a lower limit value (PL_2) and an upper limit value (PM_2) of an amplitude value present in the second PDF signal. A further upper and lower limit value calculating step (S106) to be obtained;
In the estimated value calculating step, the lower limit value of the amplitude value existing in the first PDF signal (or the lower limit value (PL — 12) which is the lowest limit among the lower limit values of the amplitude value existing in the second PDF signal). )When,
A maximum upper limit value (PM_12) which is a maximum upper limit value among the upper limit value of the amplitude value present in the first PDF signal or the upper limit value of the amplitude value present in the second PDF signal;
An average value parameter (β) including candidates of estimated values to be obtained;
A random variable (k) representing the amplitude component;
A function (f1) of the first PDF signal C;
When the function (f2) of the second PDF signal D is used,

The signal processing method according to claim 3, wherein the average value parameter β that maximizes the correlation value v (β) is obtained based on and the β is an estimated value.

Images