CN101719812B - Method for reducing signal redundancy of wireless sensor network node - Google Patents

Abstract

The invention relates to a method that the first-order difference compression technology based on the Laplacian statistical distribution is used to ensure that the wireless sensor network node compresses the data thereof to occupy less storage resources and reduce the signal redundancy of wireless sensor network node. The method is characterized by comprising the following steps: obtaining first-order difference signal and performing digital quantization of the program; and performing first-order difference method and estimation program of probability density function of the Laplacian distribution and coding first-order difference signal by using Huffman coding.

Description

Reduce the method for signal redundancy of wireless sensor network node

Technical field

The invention belongs to wireless sensor network node signal processing method field, especially a kind of first-order difference compress technique based on Laplce's Distribution Statistics reaches the method for the minimizing signal redundancy of wireless sensor network node that takies less storage resources so that the data of wireless sensor network node by its transducer of compression.

Background technology

At present, common sensor node in wireless network signal processing method is the improvement coding method of standard huffman coding and various huffman codings, and the problem of existence is few, the large footprint of performance and to take storage resources many.

Summary of the invention

The purpose of this invention is to provide a kind of first-order difference compress technique based on Laplce's Distribution Statistics so that wireless sensor network node by compressing the data of its transducer, reaches the method for the minimizing signal redundancy of wireless sensor network node that takies less storage resources.

Technical scheme of the present invention is: reduce the method for signal redundancy of wireless sensor network node, it is characterized in that comprising the following steps:

Obtaining and the digital quantization program of A, first-order difference signal specifically comprises the following steps:

(A1) turn on sensor 24 hours or sampling in 48 hours obtain sensing data, calculate the first-order difference signal;

(A2) the dynamic range values D of calculating differential signal adopts data relatively to ask the limiting value method, obtains the maximum d of differential signal _MAXWith minimum value d _MIN, namely

D＝d _MAX-d _MIN；

(A3) dynamic range values D is carried out divided in equal amounts, its density of cutting apart is 2N equivalent minizone d,

d＝D/2N

(A4) differential signal is carried out digital quantization, quantization method is if d _I-1＜first-order difference signal＜d _i, then:

d _Q＝(d _i-1+d _i)/2

Wherein, d _i=d _MIN+ i*d, i=0,1 ..., 2N, d _MAX=d _MIN+ 2N*d;

The estimation program of the probability density function of B, first-order difference method and its laplacian distribution specifically comprises the following steps:

(B1) calculate first-order difference d _iHistogram numerical value, described first-order difference signal d (t) defers to the probability density function f (d) of laplacian distribution, namely

f(d)＝A Exp(-d/τ)

(B2) from the analytic expression of histogram derivation probability density function f (d), specifically comprise the following steps;

(B2.1) regular to the probability density function of laplacian distribution;

(B2.2) numerical value with dynamic range values D is mapped on the interval of [N, N], and being about to dynamic range D, to cut apart [N, N] be the individual isometric interval of 2N, at each interval [d _I-1, d _i] on, functional value is f (d _i), i=-N ... ,-1,0,1 ..., N;

(B2.3) reading [N, N] from step (B1) histogram is that 2N each interval data in the isometric interval are to d _iAnd f (d _i), carry out calculating parameter τ value by following formula:

$\mathrm{\τ}=\underset{i=-N}{\overset{N}{\mathrm{\Σ}}}\frac{{-d}_i}{\ln f\left(d_i\right)}$

C, employing huffman coding are encoded to the first-order difference signal, specifically comprise the following steps:

(C1) set up look-up table, set up first-order difference signal and the one-to-one relationship of its probability density function in table, this table is as follows:

Index	The first-order difference signal	The probability density function of first-order difference signal
			-N	d -N	f(d -N)
-N 10	d -N+1	f(d -N+1)

-N+2	d -N+2	f(d -N+2)
			…	…
N-1	d N-1	f(d N-1)
			N	d N	f(d N)

(C2) the first-order difference signal numerical value in will showing and the probability density function numerical ordering of first-order difference signal, the principle of ordering be from the numeral of maximum probability to, the minimum probability numeral forms Hofman tree to as terminal column;

(C3) with the Hofman tree based on first-order difference signal and its probability density function that forms in the step (C2), carry out the standard huffman coding.

Effect of the present invention is: the method that reduces signal redundancy of wireless sensor network node, based on transducer first-order difference signal and estimation laplacian distribution probability density function thereof, adopt the technology of huffman coding, thereby reduced sensor signal redundancy (Redundancy).For in the dt time period, first-order difference is relatively x (t) of 0 x (t+dt), and it does not contain new information.Thereby can simplify processing, the purpose of packed data is played in transmission, and the work of storage.Satisfy wireless sensor network node and have many performances, little footprint, take the requirement of less storage resources.

The present invention is described further below in conjunction with drawings and Examples.

Description of drawings

Fig. 1 is program flow chart of the present invention;

Fig. 2 is first-order difference d of the present invention _iHistogram.

Embodiment

Reduce the method for signal redundancy of wireless sensor network node, comprise the following steps:

Obtaining and the digital quantization program of A, first-order difference signal specifically comprises the following steps:

(A1) turn on sensor 24 hours or sampling in 48 hours obtain sensing data x (t), and wherein t is time variable.On this basis, calculate first-order difference signal d (t), this differential signal d (t) is the poor of two time intervals sensing data x (t) that is dt and x (t+dt), namely

d(t)＝x(t+dt)-x(t)...(1)

Notice that d this moment (t) is continuous function, the first-order difference method has reduced sensor signal redundancy (Redundancy).For not having Varying parameters in the dt time period, the one jump is divided into 0, and for the relative x (t) of x (t+dt), it does not contain new information.Thereby can simplify the work of processing, transmission and storage, play the purpose of packed data.

(A2) the dynamic range D of calculating d (t) adopts data relatively to ask the limiting value method, obtains maximum d _MAXWith minimum value d _MIN, namely

D＝d _MAX-d _MIN...(2)

(A3) dynamic range D is carried out divided in equal amounts.Its density of cutting apart is 2N equivalent minizone d,

d＝D/2N...(3)

(A4) d (t) is carried out digital quantization, quantized value is

d _Q＝m _i if d _i-1＜d(t)＜d _i...(4)

At this d _i=d _MIN+ i*d, i=0,1 ..., 2N, m _i=(d _I-1+ d _i)/2.So far finish obtaining and digital quantization of first-order difference signal, the digital quantization value represents with 16 bits.

B, set up the estimation program of the probability density function (probability density function) of first-order difference method and its laplacian distribution (Laplace Distribution), comprise the following steps:

(B1) set up first-order difference d _iHistogram (histogram), this first-order difference signal d (t) defers to the probability density function f (d) (referring to Fig. 2) of laplacian distribution, namely

f(d)＝A Exp(-d/τ)...(5)

(B2) derive the analytic expression of probability density function f (d) from histogram;

(B2.1) first to f (d) normalization (Normalization), even A=1;

(B2.2) with dynamic range D=d _MAX-d _MINBe mapped on the interval of [N, N] D=[d _MIN, d _MAX] to cut apart [N, N] be 2N isometric interval.At each interval [d _I-1, d _i] on, functional value is f (d _i), fori=-N ... ,-1,0,1 ..., N.Here

f(d _i)＝Exp(-d _i/τ)...(5*)

(B2.3) estimation f (d _i) the parameter value τ of function, thereby obtain f (d _i) analytic expression.On [N, N] was each interval in 2N the isometric interval, reading out data was to d in histogram _iAnd f (d _i).Then substitution following formula

$\mathrm{\τ}=\underset{i=-N}{\overset{N}{\mathrm{\Σ}}}\frac{{-d}_i}{\ln f\left(d_i\right)}...\left(6\right)$

Find the solution parameter τ.

So far obtained the probability density function (probability density function) of first-order difference signal and its laplacian distribution (Laplace Distribution).The lower step will adopt huffman coding that the first-order difference signal is encoded.

C, employing huffman coding carry out coded program to the first-order difference signal, comprise the following steps:

(C1) set up look-up table, set up first-order difference signal and the one-to-one relationship of its probability density function in table.This table is as follows:

Index	The first-order difference signal	The probability density function of first-order difference signal
			-N	d -N	f(d -N)
-N+1	d -N+1	f(d -N+1)
			-N+2	d -N+2	f(d -N+2)
…	…
			N-1	d N-1	f(d N-1)
N	d N	f(d N)

(C2) will show in the every pair of numeral and its probability (d _i, f (d _i)) again one by one ordering, form Hofman tree.The principle of ordering be from the numeral of maximum probability to beginning, make it as first row in showing; The numeral of inferior large probability is to secondary series in the conduct table; Once analogize, thereby the minimum probability numeral becomes the Hofman tree of standard to as terminal column.

(C3) to the Hofman tree based on first-order difference signal and its probability density function, carry out the standard huffman coding, until whole first-order difference signal d in the his-and-hers watches _iFinish huffman coding.Be able to whole realizations to this laplacian distribution first-order difference compress technique that is used for wireless sensor network.

Claims (1)

1. reduce the method for signal redundancy of wireless sensor network node, it is characterized in that comprising the following steps:

Obtaining and the digital quantization program of A, first-order difference signal specifically comprises the following steps:

(A1) turn on sensor 24 hours or sampling in 48 hours obtain sensing data, calculate the first-order difference signal;

(A2) the dynamic range values D of calculating differential signal adopts data relatively to ask the limiting value method, obtains the maximum d of differential signal _MAXWith minimum value d _MIN, namely

D＝d _MAX-d _MIN

(A3) dynamic range values D is carried out divided in equal amounts, its density of cutting apart is 2N equivalent minizone d,

d＝D/2N

(A4) differential signal is carried out digital quantization, quantization method is if d _I-1＜first-order difference signal＜d _i, then:

d _Q＝(d _i-1+d _i)/2

Wherein, d _i=d _MIN+ i*d, i=0,1 ..., 2N, d _MAX=d _MIN+ 2N*d;

The estimation program of the probability density function of B, first-order difference method and its laplacian distribution specifically comprises the following steps:

(B1) calculate first-order difference d _iHistogram numerical value, described first-order difference signal d (t) defers to the probability density function f (d) of laplacian distribution, namely

f(d)＝A Exp(-d/τ)

(B2) from the analytic expression of histogram derivation probability density function f (d), specifically comprise the following steps;

(B2.1) regular to the probability density function of laplacian distribution, even A=1;

(B2.2) numerical value with dynamic range values D is mapped on the interval of [N, N], and being about to dynamic range D, to cut apart [N, N] be the individual isometric interval of 2N, at each interval [d _I-1, d _i] on, functional value is f (d _i), i=-N ... ,-1,0,1 ..., N;

(B2.3) reading [N, N] from step (B1) histogram is that 2N each interval data in the isometric interval are to d _iAnd f (d _i), carry out calculating parameter τ value by following formula:

$\mathrm{\τ}=\underset{i=-N}{\overset{N}{\mathrm{\Σ}}}\frac{{-d}_i}{\ln f\left(d_i\right)}$

C, employing huffman coding are encoded to the first-order difference signal, specifically comprise the following steps:

(C1) set up look-up table, set up first-order difference signal and the one-to-one relationship of its probability density function in table, this table is as follows:

Index The first-order difference signal The probability density function of first-order difference signal -N d _-N f(d _-N) -N+1 d _-N+1 f(d _-N+1) -N+2 d _-N+2 f(d _-N+2) … … N-1 d _N-1 f(d _N-1) N d _N f(d _N)

(C2) the first-order difference signal numerical value in will showing and the probability density function numerical ordering of first-order difference signal, the principle of ordering be from the numeral of maximum probability to, the minimum probability numeral forms Hofman tree to as terminal column;

(C3) with the Hofman tree based on first-order difference signal and its probability density function that forms in the step (C2), carry out the standard huffman coding.

Images