JPS6086429A - Sailing sound analyzer of ship - Google Patents

Abstract

PURPOSE:To extract a ship resonance spectrum, a resonance spectrum and 3dB band width by extracting higher harmonic line spectrum removed at its noise due to an auxiliary machinery and having an explosion period through a comb filter to extract a spectrum envelope, and simultaneously, reducing fluid noise and propeller noise in a wide range through a filter. CONSTITUTION:A sailing sound signal inputted to an input terminal 101 is converted into a power spectrum by a power spectrum converter 3 through a low- pass filter 1 and an AD converter 2 and an auxiliary machinery sound in a range specified by a controller is removed by an auxiliary machinery sound remover 4. The removed frequency and amplitude of the auxiliary machinery and the existence of the auxiliary machinery sound are stored in a storage device 10. Pitches are extracted by a pitch extractor 5 and stored in the storage device 10 and the variation of the pitch frequency is detected by a pitch monitoring device 5'. If no variation is generated, power spectrums are cumulatively meaned by a cumulator 7 and the ship resonance spectrum envelope is extracted by a comb filter circuit 6. The ship resonance spectrum is obtained from a differential spectrum between the circuits 4 and 6.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Application of the Invention] The present invention relates to a signal analysis device that receives sound emitted by a ship traveling on the ocean, processes the signal, and extracts characteristics.

[Background of the invention]

Sources of radiated noise on ships are mechanical noise such as diesel engines,
It can be divided into propeller noise from the thruster and hydrodynamic noise from the hull. Among these, regarding propeller noise, it is described in Japanese Patent Laid-Open No. 54-9967 that information such as the number of propellers can be obtained.

Conventionally, in the analysis of navigation noise, there is an example of estimating its spectral envelope using a linear prediction method (Japanese Patent Laid-Open No. 51-1123).
(See Publication No. 77 and Utility Model Application No. 1983-076505). The former is characterized by performing linear predictive analysis after accumulating short-term autocorrelation functions in order to improve noise. The latter utility model is for noise removal.

The feature is that linear prediction analysis is performed using adjacent pitch autocorrelation coefficients (correlation coefficients near pitches where no noise exists).

However, if there are strong sound components in front of auxiliary equipment, such as blowers, generators, power supplies, etc. in the cruising sound, the above two methods do not take measures in front of these auxiliary equipment, so the predicted spectrum is corrected. It is not possible to extract the hull resonance spectrum due to the engine explosion sound as it is strongly influenced by the front of the aircraft.

[Purpose of the invention]

The object of the present invention is to improve the above-mentioned drawbacks.

[Summary of the invention]

In order to achieve this objective, in the present invention, when estimating the hull resonance spectrum due to the engine explosion sound, the area in front of the auxiliary equipment that will cause interference is first highlighted in the spectrum, and then the harmonics of the engine explosion period are It is characterized by extracting the spectrum using a comb filter and extracting the spectrum envelope using a linear prediction method, and at the same time reducing broadband fluid noise and propeller noise using the comb filter. PAFLCOR coefficients are used as feature parameters. A linear prediction coefficient is obtained from the PAFLCOR coefficient, and the resonant frequency and 3 dB bandwidth of the hull are obtained.

[Embodiments of the invention]

First, the present invention will be explained in detail. This principle applies to all ships that use diesel engines as their propulsion power source. In recent years, improvements have been made to the anti-vibration table that supports the engine in order to prevent noise, but still.
Vibration noise exists.

Among these are periodic vibrations caused by explosions within the cylinders of diesel engines. The period is called the cylinder ignition period. If the period is -gt (5ec), then
The cylinder ignition frequency is 1/Pt (Hz). Assuming that this periodic sound source is p(t) and the hull is driven by the periodic sound source p(t), the impulse response of the hull is v
(t), the signal g(t) we observe is given by g(t)=Otori p(τ)v(−τ)dτ (1). The Fourier transforms of g (t), p (t), and v (t) are expressed as G (f), P (f), and V
(f), then G (f) = P (f)・V (f
) (2) In the spectrum of equation (2), if H(f) is the representative two types of auxiliary loudness 1st complement, equation (2) becomes G(f) = P(f)V( f)+H(f) (3) A specific example before the auxiliary equipment is shown in Fig. 1(a). Briefly explaining FIG. 1(a), the broken line is the fundamental wave (pitch) and its harmonic components associated with engine explosion, and the ° that connects the tops of the broken line is the hull resonance spectrum envelope. In this spectrum, in addition to the 1st complementary frequency of the auxiliary sound, its harmonic frequencies (A,', A2'
, A3'.

...) exists. Further, the pre-auxiliary A may or may not overlap the cylinder ignition frequency and its harmonic spectrum, that is, the frequency/A(H2) of the pre-auxiliary may be an integral multiple of the cylinder ignition frequency. On the other hand, the frequency fB (H2) of the auxiliary engine sound B is unrelated to the cylinder ignition cycle, and has the property of fluctuating and moving toward a higher frequency side as the speed increases, depending on the ship speed. However, the cylinder ignition cycle, frequency before auxiliary equipment, and sound pressure level differ depending on the type of ship.

The sound pressure level in front of these auxiliary machines is hidden below the hull resonance spectrum envelope level and does not overlap within the hull spectrum envelope (
(not at an integer multiple of the fundamental frequency component), there is no problem when estimating the spectrum by linear prediction. In order to extract and remove the part before the auxiliary machine in equation (3) explained above, the present invention uses the method shown in FIG. 2. That is, a frequency range is arbitrarily set in the power spectrum, and the maximum amplitude level in that range before the auxiliary equipment is found, extracted, and removed. In Figure 2, the undulations of the mountains represent the spectral characteristics in the vicinity of the area in front of the auxiliary equipment. This will be explained in more detail below.

(1) Range A set in advance in Figure 2 (a)
Find the maximum value between 1B1 and set that position as point E.

(2) Take the local maximum values that are adjacent to each other in the direction of point B, where the frequency increases with point E as the center, and the difference between these maximum values is less than a predetermined value δ (dB) (approximately 0.5 to 3 dB). For a dog, let D be the minimum point between the maximum values.

(3) Next, find the minimum point C in the same manner from point E to point A (direction where the frequency decreases).

(4) If both cases meet the above conditions, the minimum point C%
If there is no D, it is determined that there is no auxiliary equipment between A and 8.

(5) 0% The area in front of the auxiliary machine is removed by a calculation that connects point D with a straight line as shown in FIG. 2(b).

(6) Perform this calculation according to the number of auxiliary machines. Auxiliary equipment front A
In the case of , harmonics also exist, so the search range is set to the same number of harmonics and removed.

Through the above calculation operations, the hull resonance spectrum envelope is extracted from the spectrum removed before the auxiliary equipment. At this time, a comb filter having amplitude characteristics as shown in FIG. 1(b) is used. (Suzuki et al. 2; Extraction of noise by digital filtering, IEICE Technical Report EA76-57 P43) In this case, the pitch frequency is required, but the previously mentioned patent application No. 1982-8
According to No. 8151, an appropriate analysis window length and power spectrum cumulative number are determined and easily obtained. If the comb filter shown in FIG. 3 is used, the engine hull resonance spectrum (spectral envelope characteristic due to the fundamental wave of the pitch frequency and its harmonic spectrum) can be obtained from the output terminal 3002. In FIG. 3, the input terminal 300
A power spectrum calculation unit 301 calculates a power spectrum from the acoustic signal input from 1, and a comb filter 302 calculates the resonance spectrum from the result. Adder 3 calculates the difference between the power spectrum and the resonance spectrum.
03 and output from the output terminal 3003. When there are multiple ships (diesel ships), the peak of the pitch is equal to the number of ships (T1, T2, T
3) Since it exists, apply a comb filter in descending order of the extracted pitch frequency to obtain a hull resonance spectrum. According to this method, when a plurality of ships exist and their pitch frequencies are integer multiples of each other, although there is a loss of spectral components at each double frequency in the spectrum, this adverse effect can be reduced. That is, the loss in the spectral envelope of a ship with a small pitch frequency is smaller than the loss in the spectral envelope of a ship with a high pitch frequency. Therefore, it goes without saying that there is no problem if the pitch frequencies are not integral multiples of each other.

On the other hand, from the output terminal 3003, there is a possibility that the running sound spectrum of the front of the auxiliary equipment other than those removed above, the unknown front of the auxiliary equipment that does not overlap with pitch harmonics, or the running sound spectrum of other vessels (turbine ships, etc.) may be obtained. There is.

An autocorrelation coefficient can be obtained by performing fast Fourier transform (FFT) on the hull resonance spectrum and the cruise sound spectrum. The autocorrelation coefficient is (r(n)=r (nΔ)In=0, 1
,...), then in the linear prediction method, D
urbin's solution (J, Durbin: The Fit
Ting of Time-series Mode
ls, Rev.

In5t,Int,5tatist,,28-3,23
3/234 (1960)), the following three equations are obtained. That is, αte)=α'("""r""(X:' 1kutkur
1゜α = -kr (5) ``τ=Er-1('''r) t'' o=r(o)
(6) However, k7 is the PARCOR coefficient and E7 is the prediction residual power. Starting from τ = 1, the order is increased sequentially, and when τ = p, the prediction coefficient a, = αl (P' (1 <
i<p) is obtained.

The estimated linear prediction spectrum P←) is, however, the pole in P(G)) is Z t = γ, @ e
” t t =1.

2. If p is set, the resonant frequency Pi is 3 dB
The bandwidth Bi is given by b t = log − (10) πΔ.

An embodiment according to the present invention will be described below.

FIG. 5 is a block diagram showing the configuration of an embodiment according to the present invention. When a cruise sound signal is inputted from an input terminal 101, it is inputted to a low frequency door filter 1 having a cutoff frequency suitable for the sampling frequency, and then sampled and quantized by an A/D converter 2. The power spectrum calculator 3 performs calculations using the circuit shown in FIG. 5(b). That is, the time series data input from the input terminal 301 is processed using the window length (
Cut out using the window cutter 31 of the frame length). The fast Fourier transformer 32 frequency-analyzes the output of the windower 31. Multiplier 33 converts it into a power spectrum. Next, the pre-auxiliary remover 4 removes the pre-auxiliary from the power spectrum. The range in which to search for the type in front of the auxiliary machine is specified in advance by the control device 11.

The frequency and amplitude of the removed pre-auxiliary machine, and the presence or absence of the pre-auxiliary machine are stored in the storage device 10. The pitch extractor 5 is based on the pitch extraction method of the cepstral method, and its configuration is shown in FIG. 5(C). That is, when a power spectrum is input to the input terminal 501, it is converted into a logarithmic value by the logarithmic converter 51, and after being input to the fast Fourier transformer 52, it is converted to a power cestrum. The pitch detector 53 detects a plurality of peaks of high quenching frequency that exceed a certain threshold value. At this time, the control device 11 instructs how many peaks to detect. The extracted pitch is input to the storage device 10. On the other hand, the pitch monitoring device 5' detects fluctuations in pitch frequency based on the pitch calculated by the pitch extractor 5. If there is a change in pitch, the controller 11 instructs the accumulator 7 to stop accumulating the spectrum, stops the addition, averages the accumulated power spectrum, and inputs it to the storage device 10.

If there is no change in the pitch frequency, the control device 11 operates to cumulatively average the spectra of a predetermined number of frames. The same is true when there is no pitch, and the power spectrum is cumulatively averaged a predetermined number of times. The comb filter circuit 6 extracts the hull resonance spectrum envelope from the stable power spectrum thus obtained. (The auxiliary equipment noise has already been removed.) If the data input to the input terminal 501 includes the navigation sounds of multiple ships, the storage device 10 stores multiple peaks, so the control device 11 With this instruction, the hull resonance spectra of multiple ships can be obtained using the spectrum obtained by subtracting the output of 6 from the output of 4. Also, other auxiliary equipment noise and spectra of other ships are extracted from the remaining spectrum after subtraction.

These power spectra are input to the autocorrelator 8. The autocorrelator 8 obtains autocorrelation coefficients by performing fast Fourier transform on the cumulative power spectrum.

The circuit configuration of the PARCOR analyzer 9 is shown in FIG. 5(d). That is, the autocorrelation coefficient input to the input terminal 901 is subjected to PARCOR analysis by the PARCOR calculator 91 to obtain a PARCOR coefficient. Then, the linear prediction calculator 92 calculates the linear prediction coefficient of the prediction order instructed by the control device 11 from the PAIE (, COB coefficient). The prediction spectrum calculator 93 calculates the prediction spectrum from the linear prediction coefficient. Resonance frequency calculation The device 94 calculates the resonant frequency and 3 dB bandwidth from the linear prediction coefficient.The PARCOR coefficient, linear prediction coefficient, resonant frequency, 3
dB bandwidth and hull resonance spectrum are controlled by the control device 11.
The data is stored in the storage device 10 via. The display device 12 is used to visually check the results of parameters such as the hull resonance spectrum, cepstrum, pitch, and auxiliary machine sound information stored in the storage device 10 in response to commands from the input device 13.

The storage device 10 is of a pack type that can be replaced with a new one when it is full of information.

〔Effect of the invention〕

As explained above, the present invention analyzes the sound of a ship running, removes the sound in front of the auxiliary equipment, extracts the pitch, and uses a spectral cumulative average that can change the cumulative number according to pitch fluctuations, and a pitch harmonic. An analysis device whose main purpose is to extract a spectral envelope, that is, a hull resonance spectrum for each ship, and to extract the hull resonance spectrum envelope, resonance frequency, and 3 dB bandwidth by PARCOR analysis is provided. With this analysis device, useful ship-specific characteristics can be obtained automatically.

[Brief explanation of drawings]

Figure 1 (al is a diagram showing an example of the power spectrum of a ship's running sound, Figure 1 (b) is a diagram showing an example of a comb filter with an amplitude characteristic of frequency 1/Pt (Hz), Figure 2 is a diagram showing an example of a comb filter with an amplitude characteristic of a frequency of 1/Pt (Hz), Figure 3 is a block diagram explaining the method for extracting the hull resonance spectrum, Figure 4 is a diagram explaining pitch extraction by cepstral analysis, and Figure 5 is an implementation of the present invention. There is an example block configuration diagram C. 3... Power spectrum calculator, 4... Auxiliary equipment pre-removal unit, 5... Pitch extractor, 5'... Pitch monitoring device,
6...Comb filter. Violet 1 Figure 22 QD Frequency C Whistle 3 Figure 4- Location Frenzy (Sec)

Claims (1)

[Claims] 1. A means for frequency-analyzing ship running sound to extract and remove auxiliary equipment noise, and cepstrum analysis of the auxiliary equipment noise power spectrum to extract the pitch, monitor pitch fluctuations, and extract the power spectrum. A means for obtaining a hull resonance spectrum for each ship from the accumulated power spectrum using a comb filter, and a means for performing PARCOR analysis on the power spectrum to obtain desired characteristic parameters. Running sound analyzer.