JP5276334B2 - Parent-child relationship determining device, parent-child relationship determining method, parent-child relationship determining program, and recording medium - Google Patents

Description

  The present invention relates to a parent-child relationship determination apparatus and a parent-child relationship determination for reducing a total energy of each signal sequence by expressing a signal sequence of a plurality of channels as a difference between a signal sequence of a child channel and a signal sequence of a parent channel. The present invention relates to a method, a parent-child relationship determination program, and a recording medium.

The technology to reduce the total energy of each signal sequence by expressing the signal sequence of multiple channels with the difference between the signal sequence of child channels and the signal sequence of parent channel is a technology that compresses and encodes signals of multiple channels. It is known as a part (Patent Document 1).
International Publication No. 2006/019117 Pamphlet

  In the technique of searching for the parent-child relationship in Patent Document 1, a parent channel (hereinafter referred to as “root”) that does not become a child of any channel must first be determined, and the child channel of the parent channel must be searched in order. This is because, in the technique of Patent Document 1, only when the parent channel is connected to the root (when the parent channel is traced in the direction of the parent channel and finally reaches the root), the parent channel is used. This is because the child channel was determined to be connectable.

  In such a process, the processing time increases rapidly as the number of channels to be analyzed increases. For example, in order to compress a 512-channel MEG signal (cerebral magnetic field measurement signal), about 1000 times the real time is required.

  The present invention has been made to solve such a problem, and an object of the present invention is to provide a parent-child relationship determination device in which the processing time does not increase abruptly even if the number of analyzed channels increases. .

The parent-child relationship determination apparatus of the present invention includes a cross-correlation / energy calculation unit, a quotient calculation unit, and a parent-child relationship search unit. The cross-correlation / energy calculation unit calculates the cross-correlation C _{m1, m2} between signal sequences for all two channel combinations in M channels (where M is an integer of 3 or more) (where m1 and m2 are 1). The integer E of M or less and m1 ≠ m2) and the energy E _m of the signal train of all channels m (where m is an integer of 1 to M) are obtained. The quotient calculation unit calculates the square of the cross-correlation C _{j, ρ (j)} (where j is an integer of 1 to M and ρ (j) is an integer of 1 to M other than j) as energy E _{ρ (j).} A value C _{j, ρ (j)} ² / E _{ρ (j)} divided by is obtained for all combinations of channel j and channel ρ (j). Parent-child relationship search unit, the value _{C j,} select the descending order from the ^{_{ρ (j) 2 / E ρ}} (j) , selected values _{C j,} corresponding to ^{_{ρ (j) 2 / E ρ}} (j) Check whether the parent channel of channel j is determined. If channel ρ (j) is not a descendant of channel j, channel ρ (j) is determined as the parent channel of channel j. Or, if channel ρ (j) is a descendant of channel j, the process of selecting the next largest value C _{j, ρ (j)} ² / E _{ρ (j)} is the parent of M−1 channels. Repeat until the channel is determined.

Alternatively, the parent-child relationship determination device of the present invention includes a cross-correlation / energy calculation unit, a vector generation unit, a vector element sorting unit, and a parent-child relationship search unit. The cross-correlation / energy calculation unit calculates the cross-correlation C _{m1, m2} between signal sequences for all two channel combinations in M channels (where M is an integer of 3 or more) (where m1 and m2 are 1). The integer E of M or less and m1 ≠ m2) and the energy E _m of the signal train of all channels m (where m is an integer of 1 to M) are obtained. The vector generation unit _calculates the square of the cross-correlation C _{j, ρ (j)} (where j is an integer from 1 to M and ρ (j) is an integer from 1 to M other than j) as energy E _{ρ (j).} The value C _{j, ρ (j)} ² / E _{ρ (j)} divided by is obtained for all combinations of channel j and channel ρ (j), and the obtained value C _{j, ρ (j)} ² / E A vector G having all of _{ρ (j)} as elements is generated. Vector elements sorting section rearranges the descending order vector element G, to generate the vector G _S after sorting. Parent-child relationship search unit selects the element of the vector G _S to the element order, to check the selected element C _{j, [rho (j)} the parent channel of the channel j corresponding to ₂ ^/ E _{ρ (j)} is determined, If not determined and channel ρ (j) is not a descendant of channel j, channel ρ (j) is the parent channel of channel j and is determined, or channel ρ (j) is a descendant of channel j In the case of, the process of selecting the next largest value C _{j, ρ (j)} ² / E _{ρ (j)} is repeated until the parent channels of M−1 channels are determined.

  The parent-child relationship determination method of the present invention is based on the condition that the signal sequence energy of the child channel is expressed as a difference from the signal sequence of the parent channel under the condition that the parent channel of any channel is not a descendant of the channel. It is characterized by finding a parent-child relationship that minimizes the sum.

  The parent-child relationship determining apparatus and the parent-child relationship determining method of the present invention obtain a combination having the highest correlation between channels under the condition that a parent-child relationship loop is not formed (the parent channel of an arbitrary channel is not a descendant of the channel). select. Therefore, there is no need to determine the parent-child relationship in order from the root channel. Therefore, even if the number of channels increases, a rapid increase in processing time can be avoided.

  First, what conditions are necessary to reduce the total energy of each signal sequence by expressing the signal sequence of multiple channels and the difference between the signal sequence of the child channel and the signal sequence of the parent channel? To consider.

It is assumed that a signal sequence composed of N signals is input for M channels. The n-th signal (where n is an integer from 1 to N) of the channel m (where m is an integer from 1 to M) is expressed as s _m (n). Here, if the parent channel of the channel m is the channel ρ (m), the differentially expressed signal _m s ^~ _m (n) of the channel m is
^{_{s ~ m (n) = s}} m (n) -γ · s ρ (m) (n) (1)
It becomes. Where γ is a weighting factor. The energy E ^~ _m of the signal sequence of the child channel expressed as a difference is expressed as follows:

となる。そして、このエネルギーＥ^〜 _ｍが最小となる重み係数γは、

It becomes. The weighting coefficient γ that minimizes the energy E ^to _m is

のように計算できる。式（１）〜（３）より、差分表現された子チャネルの信号列のエネルギーＥ^〜 _ｍは、

It can be calculated as follows. From the equations (1) to (3), the energy E ^~ _m of the signal sequence of the child channel that is differentially expressed is

となる。なお、Ｅ_ｍはチャネルｍの信号列のエネルギー、Ｅ_ρ（ｍ）はチャネルρ（ｍ）の信号列のエネルギー、Ｃ_{ｍ，ρ（ｍ）}はチャネルｍの信号列とチャネルρ（ｍ）の信号列との相互相関であり、

It becomes. E _m is the energy of the signal train of channel m, E _{ρ (m)} is the energy of the signal train of channel ρ (m), and C _{m, ρ (m)} is the signal train of channel m and channel ρ (m). Cross-correlation with the signal sequence,

である。ロスレス符号化では、エネルギーを符号量として近似できるので、できるだけエネルギーを小さくするように親子関係を探索したい。つまり、チャネルｍにとって、差分表現（重み付き減算）した信号のエネルギーＥ^〜 _ｍを最小にしてくれる親チャネルρ（ｍ）を求めたい。

It is. In lossless coding, energy can be approximated as a code amount, so it is desirable to search for a parent-child relationship so as to make the energy as small as possible. That is, for the channel m, it is desired to obtain the parent channel ρ (m) that minimizes the energy E ^to _m of the differentially expressed (weighted subtraction) signal.

In the case of a stereo signal (signal with 2 channels), when channel 1 is a parent channel and channel 2 is a child channel, the total energy E _ref1 is
E _ref1 = E ₁ + E ₂ − (C _2,1 ) ² / E ₁ (6)
It becomes. Conversely, if channel 2 is the parent channel and channel 1 is the child channel, the total energy E _ref2 is
E _ref2 = E ₂ + E ₁ − (C _1,2 ) ² / E ₂ (7)
It becomes. Comparing the equations (6) and (7), it can be seen that since C _2,1 = C _1,2 , the total energy is smaller when the channel having the lower energy is used as a parent.

The above-described study on stereo signals is generalized when the number of channels is M. At least one channel is a root. If the root channel is J, the total energy E _total is

と表現できる。式（８）のＨを大きくすれば、エネルギーが減少するので、符号量を縮小でき、圧縮率が向上する。したがって、以下で説明する親子関係決定装置は、式（８）のＨを最大にする親子の組合せを求める。また、以下の説明では、本発明の親子関係決定装置の具体的な使用例として、親子関係決定装置を備えた符号化装置を示しながら説明する。

Can be expressed. If H in Expression (8) is increased, the energy is reduced, so that the code amount can be reduced and the compression rate is improved. Therefore, the parent-child relationship determination device described below finds a parent-child combination that maximizes H in Equation (8). Moreover, in the following description, it demonstrates, showing the encoding apparatus provided with the parent-child relationship determination apparatus as a specific usage example of the parent-child relationship determination apparatus of this invention.

  In the following description, components having the same function and process steps for performing the same process are assigned the same number to avoid duplication of description, and description thereof is omitted.

[First Embodiment]
In the first embodiment, a case where the number of channels is 3 will be described using a specific example in order to facilitate understanding of the present invention. FIG. 1 shows a functional configuration example of the encoding apparatus according to the first embodiment. FIG. 2 shows a processing flow example of the encoding apparatus according to the first embodiment. The encoding apparatus 100 is an apparatus that encodes a three-channel signal. The encoding apparatus 100 includes a frame buffer unit 110, a cross-correlation / energy calculation unit 120, a vector generation unit 130, a vector element sorting unit 140, a parent-child relationship search unit 150, a difference signal sequence generation unit 160, and an encoding unit 170.

The frame buffer unit 110 divides the input signal s _m (n) of each channel into a predetermined number (N), and outputs a signal sequence S _m composed of N signals for each channel (however, , M is an integer of 1 to 3, and n is an integer of 1 or more) (S110). The cross-correlation / energy calculation unit 120 performs mutual correlation between signal sequences for combinations of all two channels m1 and m2 in three channels (where m1 and m2 are integers of 1 to M and m1 ≠ m2). Correlations C _{m1 and m2} and energy E _m of the signal sequences of all channels m are obtained (S120). _{Incidentally,} since _{C m1, m2 = C m2,} m1, to obtain the cross-correlation _{C m1, m @ 2} for all combinations such that m1 <m @ 2 calculates a cross-correlation _{C m1, m @ 2} for all combinations m1 ≠ m @ 2 Is the same. Therefore, the cross-correlation / energy calculation unit 120 may calculate and output C _1,2 , C _1,3 , C _2,3 , E ₁ , E ₂ , E ₃ . For example, it is assumed that C _1,2 = 4, C _1,3 = 1, C _2,3 = 1/10, E ₁ = 60, E ₂ = 48, and E ₃ = 20.

The vector generation unit 130 _calculates the square of the cross-correlation C _{j, ρ (j)} (where j is an integer of 1 to 3 and ρ (j) is an integer of 1 to 3 other than j) as the energy E _{ρ (j )} Divided by C _{), ρ (j)} ² / E _{ρ (j)} is obtained for all combinations of channel j and ρ (j), and the obtained values C _{j, ρ (j)} ² / A vector G having all of E _{ρ (j)} as elements is generated (S130). In the case of the above example, for example, G = (C _2,1 ² / E ₁ , C _1,2 ² / E ₂ , C _3,1 ² / E ₁ , C _1,3 ² / E ₃ , C _{3, 2} ² / E ₂ , C _2,3 ² / E ₃ ) = (1/15, 1/12, 1/60, 1/20, 1/480, 1/200). Note that the order of the elements of the vector G does not matter.

Vector elements sorting unit 140 rearranges the descending order vector element G, to generate the vector _{G S} after sorting (S140). In the case of the above example, G _S = (C _1, ² / E ₂ , C _2,1 ² / E ₁ , C _1,3 ² / E ₃ , C _3,1 ² / E ₁ , C _2,3 ² / E ₃ , C _3,2 ² / E ₂ ). In the present embodiment, first create a vector G agnostic order of elements (S130), and generates a vector G _S are rearranged in descending order (S140), generates a vector G _S which are arranged in descending order from the beginning May be. In this case, instead of the vector generation unit 130 and the vector element sorting unit 140 (steps S130 and S140), the vector generation unit 130 ′ is arranged in descending order from the output of the cross-correlation / energy calculation unit 120 (step S120). it generates a vector _{G S} (S130 ').

FIG. 3 shows an example of the processing flow of the parent-child relationship search unit 150. First, the processing of the parent-child relationship search unit 150 is specifically shown using the above example as follows. 1 is substituted into i, and 3 is substituted into c (S151). It is confirmed whether c = 1 (S152). c ≠ 1, so to verify that determined parent channel of the channel 1 corresponding to the first element _C ^1,2 ₂ / _E 2 vectors _{G S} (S153). Since the parent channel of channel 1 is not decided, it is confirmed whether channel 2 is not a descendant of channel 1 (S154). Since channel 2 is not a descendant of channel 1, channel 2 is set as the parent channel of channel 1 (S155). 2 is substituted for c (S156). 2 is substituted for i, and the process returns to step S152 (S157). It is confirmed whether c = 1 (S152). c ≠ 1, so to verify that decided the second element _C ^2,1 2 / _E parent channel of the corresponding channel 2 to ₁ vector _{G S} (S153). Since the parent channel of channel 2 is not determined, it is confirmed whether channel 1 is not a descendant of channel 2 (S154). Since channel 1 is a child of channel 2, 3 is substituted for i, and the process returns to step S152 (S157). It is confirmed whether c = 1 (S152). c ≠ 1, so to verify that determined parent channel of the channel 1 corresponding to the third element _C ^{1, 3} 2 / _{E 3} vectors _{G S} (S153). Since the parent channel of channel 1 is determined to be channel 2, 4 is substituted for i, and the process returns to step S152 (S157). c ≠ 1, so to verify that determined parent channel of the channel 3 corresponding to the fourth element _C ^3,1 2 / _{E 1} of the vector _{G S} (S153). Since the parent channel of channel 3 has not been determined, it is confirmed whether channel 1 is not a descendant of channel 3 (S154). Since channel 1 is not a descendant of channel 3, channel 1 is set as the parent channel of channel 3 (S155). 1 is substituted into c (S156). 5 is substituted for i, and the process returns to step S152 (S157). It is confirmed whether c = 1 (S152). Since c = 1, it is output that channel 2 is the root, ρ (1) = 2, ρ (3) = 1, and the process of step S150 is terminated. In this example, a tree structure of 2 ← 1 ← 3 is created.

The process of the parent-child relationship search unit 150 (S150) is generally shown as follows. 1 is substituted into i, and 3 is substituted into c (S151). It is confirmed whether c = 1 (S152). If step S152 is Yes, the searched parent-child relationship is output, and step S150 is terminated. If step S152 is No, confirms whether the decided i th element _{C j, [rho ^(j)} the parent channel of the channel j corresponding to 2 / E ρ _(j) of the vector _{G S} (S153). When step S153 is Yes, it progresses to step S157. When step S153 is No, it is confirmed whether the channel ρ (j) is not a descendant of the channel j (S154). Note that “confirming that it is not a descendant” means repeatedly confirming whether it is not a child channel, a child channel of a child channel, or a child channel. By confirming in this way, it can be confirmed that a parent-child relationship loop is not created (the parent channel of an arbitrary channel is not a descendant of the channel). In step S154 of FIG. 3, “Does the connection from channel j to channel ρ (j) form a loop?” Indicates that “channel ρ (j) is not a descendant of channel j?” Means the same. This confirmation prevents a loop from occurring in the obtained parent-child tree structure. When step S154 is Yes, it progresses to step S157. If step S154 is No, channel ρ (j) is set as the parent channel of channel j (S155). Then, c-1 is substituted for c (S156). That is, c indicates the number of channels whose parents are not determined. In step S157, i + 1 is substituted for i, and the process returns to step S152 (S157). That is, the process proceeds to processing for the next element of the vector G _S (next highest element).

Based on the obtained parent-child relationship, the difference signal sequence generation unit 160 obtains and outputs a difference between the signal sequence of the child channel and the signal sequence of the parent channel (S160). Specifically, the weighting coefficient γ for each signal sequence is obtained by Equation (3), the difference signal δ _m (n) of each signal is obtained by Equation (1), and the difference signal sequence Δ _m = (δ _m (1) ,..., Δ _m (N)). Note that the original channel may be output as it is for the root channel.

  The encoding unit 170 encodes the difference signal sequence of all channels (where the signal sequence of the root channel is the original signal sequence) (S170). If the encoding apparatus 100 transmits information indicating the parent-child relationship with the code D, the decoding apparatus (not shown) obtains a differential signal sequence by decoding the code, and determines each channel based on the information indicating the parent-child relationship. Can be returned to the original signal sequence.

  In the present embodiment, the configuration of the encoding device has been described. However, the parent-child relationship determination device includes a cross-correlation / energy calculation unit 120, a vector generation unit 130, a vector element sorting unit 140, and a parent-child relationship search unit 150. It corresponds to. Since the parent-child relationship is determined by such processing, it is not necessary to determine the parent-child relationship in order from the root channel. Therefore, even if the number of channels increases, a rapid increase in processing time can be avoided.

[Second Embodiment]
Although the number of input channels is 3 in the first embodiment, the number of input channels is generalized in this embodiment. FIG. 4 shows a functional configuration example of the encoding apparatus according to the second embodiment. FIG. 2 shows a processing flow example of the encoding apparatus according to the second embodiment, and FIG. 3 shows a processing flow example of the parent-child relationship search unit 250. The encoding apparatus 200 is an apparatus that encodes an M channel signal (where M is an integer of 3 or more). The encoding apparatus 200 includes a frame buffer unit 210, a cross-correlation / energy calculation unit 220, a vector generation unit 230, a vector element sorting unit 240, a parent-child relationship search unit 250, a difference signal sequence generation unit 260, and an encoding unit 270.

The frame buffer unit 210 divides the input signal s _m (n) of each channel into a predetermined number (N), and outputs a signal sequence S _m composed of N signals for each channel (however, , M is an integer of 1 or more and M or less, and n is an integer of 1 or more) (S210). Correlation energy calculating unit 220, all the two channels in the M channels m1, m @ 2 (provided that, m1, m @ 2 is 1 or more _{M k} an integer, and mk1 ≠ mk2) between the signal sequence for the combination of The cross-correlation C _{m1 and m 2} and the energy E _m of the signal train of all channels m are obtained (S220). _{Incidentally,} since _{C m1, m2 = C m2,} m1, to obtain the cross-correlation _{C m1, m @ 2} for all combinations such that m1 <m @ 2 calculates a cross-correlation _{C m1, m @ 2} for all combinations m1 ≠ m @ 2 Is the same.

The vector generation unit 230 _converts the square of the cross-correlation C _{j, ρ (j)} (where j is an integer from 1 to M and ρ (j) is an integer from 1 to M other than j) to energy E _{ρ (j )} Divided by C _{), ρ (j)} ² / E _{ρ (j)} is obtained for all combinations of channel j and ρ (j), and the obtained values C _{j, ρ (j)} ² / A vector G having all of E _{ρ (j)} as elements is generated (S230). Note that the order of the elements of the vector G does not matter. Vector elements sorting unit 240 rearranges the descending order vector element G, to generate the vector _{G S} after sorting (S240). Similar to the first embodiment, a vector generation unit 230 ′ may be provided instead of the vector generation unit 230 and the vector element sorting unit 240 (steps S230 and S240). In this case, the vector generator 230 'is, from the output of the cross-correlation and energy calculator 220 (step S220), and generates a vector _{G S} arranged in a forward element is large (S230').

  The parent-child relationship search unit 250 substitutes 1 for i and substitutes M for c (S251). Subsequent processing (steps S152 to S157) of the parent-child relationship search unit 250 is the same as that of the parent-child relationship search unit 150 of the first embodiment (see FIG. 3).

The difference signal sequence generation unit 260 calculates and outputs a difference between the signal sequence of the child channel and the signal sequence of the parent channel based on the determined parent-child relationship (S260). Specifically, the weighting coefficient γ for each signal sequence is obtained by Equation (3), the difference signal δ _m (n) of each signal is obtained by Equation (1), and the difference signal sequence Δ _m = (δ _m (1) ,..., Δ _m (N)). Note that the original channel may be output as it is for the root channel.

  The encoding unit 270 encodes the differential signal sequences of all channels (where the signal sequence of the root channel is the original signal sequence) (S270). If the encoding device 200 transmits information indicating the parent-child relationship with the code D, the decoding device (not shown) obtains a differential signal sequence by decoding the code, and determines each channel based on the information indicating the parent-child relationship. Can be returned to the original signal sequence.

  In the present embodiment, the configuration of the encoding device has been described, but the parent-child relationship determination device corresponds to a device having a cross-correlation / energy calculation unit 220, a vector generation unit 230, a vector element sorting unit 240, and a parent-child relationship search unit 250. To do. Since the parent-child relationship is determined by such processing, it is not necessary to determine the parent-child relationship in order from the root channel. Therefore, even if the number of channels increases, a rapid increase in processing time can be avoided.

  In the above description, the difference signal sequence is obtained for the signal sequences of all the channels input to the encoding apparatus 200. However, encoding may be performed after determining a parent-child relationship for some of the channels to be encoded and obtaining a differential signal sequence. In this case, M may be the number of channels for which the parent-child relationship is determined.

[Third Embodiment]
In the second embodiment, the parent-child relationship of the signal sequence is obtained once, and the difference signal sequence is obtained once. In the present embodiment, an example in which a parent-child relationship between difference signal sequences is further obtained and a difference signal sequence between the difference signal sequences is obtained is shown. Since such a process can be repeated many times, the parent channel of the channel j obtained for the first time is represented by ρ ⁽¹⁾ (j), and the difference signal sequence obtained for the first time is represented by Δ _m ⁽¹⁾ . Similarly, a parent channel of the channel j obtained at the p-th time is represented by ρ ^(p) (j), and a difference signal sequence obtained at the p-th time is represented by Δ _m ^(p) .

  FIG. 5 shows a functional configuration example of the encoding apparatus according to the third embodiment. FIG. 6 shows an example of a processing flow of the encoding apparatus according to the third embodiment. The encoding apparatus 300 is an apparatus that encodes an M channel signal (where M is an integer of 3 or more). The encoding apparatus 300 includes a frame buffer unit 210, a cross-correlation / energy calculation unit 320, a vector generation unit 230, a vector element sorting unit 240, a parent-child relationship search unit 250, a difference signal sequence generation unit 360, and an encoding unit 370.

The processing of the frame buffer unit 210 (S210) is the same as in the second embodiment. Correlation energy calculating unit 320, all the two channels in the M channels m1, m @ 2 (provided that, m1, m @ 2 is 1 or more _{M k} an integer, and mk1 ≠ mk2) between the signal sequence for the combination of The cross-correlation C _{m1 and m 2} and the energy E _m of the signal train of all channels m are obtained (S320).

The processes (S230 to S250) of the vector generation unit 230, the vector element sorting unit 240, and the parent-child relationship search unit 250 are the same as those in the second embodiment. The difference signal sequence generation unit 360 obtains and outputs the difference between the signal sequence of the child channel and the signal sequence of the parent channel based on the obtained parent-child relationship (S360). Then, it is confirmed whether or not a predetermined condition is satisfied. If the condition is satisfied, the process proceeds to step S370. If not satisfied, the process returns to step S320 (S365). Note that the predetermined condition includes a case where the total number of energy is equal to or less than a threshold when the number of repetitions is set to P in advance. In step S320, the signal sequence Δ _m ^(p) output from the differential signal sequence generation unit is used as the input signal sequence, and the cross-correlation between the signal sequences and the energy for each signal sequence are obtained (S320).

  The encoding unit 370 encodes the differential signal sequences of all channels (however, the signal sequence of the root channel is the original signal sequence) (S370). If the encoding apparatus 300 transmits information indicating a parent-child relationship for a plurality of times with the code D, the decoding apparatus (not shown) obtains a differential signal sequence by decoding the code, and based on the information indicating the parent-child relationship Thus, the difference signal sequence of each channel can be returned to the original signal sequence.

  In the present embodiment, the configuration of the encoding device has been described. However, the parent-child relationship determination device includes a cross-correlation / energy calculation unit 320, a vector generation unit 230, a vector element sorting unit 240, a parent-child relationship search unit 250, and a differential signal sequence generation. This corresponds to a device having the unit 360. Since the parent-child relationship is determined by such processing, it is not necessary to determine the parent-child relationship in order from the root channel. Therefore, even if the number of channels increases, a rapid increase in processing time can be avoided.

  In the above description, the difference signal sequence is obtained for the signal sequences of all the channels input to the encoding apparatus 300. However, encoding may be performed after determining a parent-child relationship for some of the channels to be encoded and obtaining a differential signal sequence. In this case, M may be the number of channels for which the parent-child relationship is determined. In addition, each time a parent-child relationship is obtained, a channel for which a parent-child relationship is obtained may be appropriately selected.

[Fourth Embodiment]
In the first to third embodiments, the parent-child relationship of the signal sequence of a set of three or more channels is determined. In the present embodiment, the parent-child relationship of signal sequences of three or more channels of K groups (where K is an integer of 1 or more) is determined.

  FIG. 7 shows a functional configuration example of the encoding apparatus according to the fourth embodiment. FIG. 8 shows a processing flow example of the encoding apparatus according to the fourth embodiment. The encoding apparatus 400 is an apparatus that encodes an M channel signal (where M is an integer of 3K or more). The encoding apparatus 400 includes a frame buffer unit 210, a cross-correlation / energy calculation unit 420, a vector generation unit 430, a vector element sorting unit 440, a parent-child relationship search unit 450, a difference signal sequence generation unit 260, and an encoding unit 270. The processing of the frame buffer unit 210 (S210) is the same as in the second embodiment.

The cross correlation / energy calculation unit 420 divides the M channel into K sets. Here, the number of channels in the k-th set (where k is an integer between 1 and K) is _Mk . Therefore, M = ΣM _k . Correlation energy calculating unit 420, all the two channels in each set of _{M k} number of channels mk1, mk2 (although, mk1, mk2 is a number of channels belonging to k-th group, one or more _{M k} Cross correlations C _{mk1 and mk2} between signal sequences for the following integers and combinations of mk1 ≠ mk2) and energy E _m of signal sequences of all channels m are obtained (S420).

The vector generation unit 430 has a cross-correlation C _{j, ρ (j)} (where j is an integer of 1 to M _k and ρ (j) is an integer of 1 to M _k other than j) for each set. A value C _{j, ρ (j)} ² / E _{ρ (j)} obtained by dividing the square by energy E _{ρ (j)} is obtained for all combinations of channel j and channel ρ (j) in the set. Then, the vector generation unit 430 generates a vector G _k having all of the obtained values C _{j, ρ (j)} ² / E _{ρ (j)} as elements for all sets (S430).

The vector element sorting unit 440 rearranges the elements in the descending order for each vector G _k and generates a sorted vector G _Sk (S440). Similar to the first embodiment, a vector generation unit 430 ′ may be provided instead of the vector generation unit 430 and the vector element sorting unit 440 (steps S430 and S440). In this case, the vector generation unit 430 ′ generates a vector G _{Sk in} which the elements are arranged in descending order from the output of the cross-correlation / energy calculation unit 420 (step S420) (S430 ′). The parent-child relationship search unit 450 obtains the parent-child relationship for each group by the same processing as the parent-child relationship search unit 250 (steps S251 to S157).

  The process of the difference signal sequence generation unit 260 (S260) and the process of the encoding unit 270 (S270) are the same as in the second embodiment. With such a configuration, the same effect as in the second embodiment can be obtained. In the present embodiment, the parent-child relationship search device corresponds to a device having a cross-correlation / energy calculation unit 420, a vector generation unit 430, a vector element sorting unit 440, and a parent-child relationship search unit 450.

[Modification]
In 4th Embodiment, the parent-child relationship of the signal sequence was calculated | required once, and the difference signal sequence was calculated | required once. In this modification, an example in which a parent-child relationship between difference signal sequences is further obtained and a difference signal sequence between the difference signal sequences is obtained is shown. Since the difference signal sequence of the difference signal sequence is obtained in this way, this modification is a combination of the fourth embodiment and the third embodiment. The functional configuration example of the encoding apparatus according to the present modification includes a dotted line portion in FIG. Further, the processing flow example of the encoding device of the present modification example is a processing flow including the dotted line portion of FIG. The encoding device 400 ′ includes a frame buffer unit 210, a cross-correlation / energy calculation unit 420 ′, a vector generation unit 430, a vector element sorting unit 440, a parent-child relationship search unit 450, a difference signal sequence generation unit 360, and an encoding unit 370. Prepare. The processing of the frame buffer unit 210 (S210) is the same as in the second embodiment.

The cross-correlation / energy calculation unit 420 ′ divides the M channel into K sets. Then, cross-correlation and energy calculator 420, all of the two channels in each set of _{M k} number of channels mk1, mk2 (although, mk1, mk2 is a number of channels belonging to k-th group, one or more M _k an integer, and the mk1 ≠ mk2) combined cross-correlation _{C mk1, mk2} between signal sequence for the obtained energy _{E m} of the signal sequence for all channels m (S420 '). The first iteration is the same as in the fourth embodiment. In the second and subsequent iterations, the differential signal sequence from the differential signal sequence generation unit 360 is input to divide the M channel into K groups, and the cross-correlation C _{mk1 and mk2 and} the energy E _m of the signal sequence of all channels m are obtained. (S420 ′). However, the number of sets may be changed for each repetition, the number of channels belonging to each set may be changed, and which channel belongs to which set may be changed. For example, assume that a 12-channel signal sequence is input, and initially, the first to fourth channels are set to the first set, the fifth to eighth channels are set to the second set, and the ninth to 12th channels are set to the third set. In this case, the second grouping may be the same grouping. Alternatively, the 3rd to 6th channels may be the first set, the 7th to 10th channels may be the second set, and the 11, 12, 1, and 2 channels may be the third set. Alternatively, the 1st to 6th channels may be the first set, and the 7th to 12th channels may be the second set.

  The processes of the vector generation unit 430, the vector element sorting unit 440, and the parent-child relationship search unit 450 are the same as those in the fourth embodiment. The processing of the difference signal sequence generation unit 360 and the encoding unit 370 is the same as that in the third embodiment.

  With this configuration and processing flow, the present embodiment can provide the same effects as those of the above-described embodiment.

  The case where the fourth embodiment and the fourth embodiment modification are effective will be described using a specific example. FIG. 9 shows an example of the parent-child relationship of a 32-channel signal sequence. The numbers in the circles are numbers that identify the channels, and the tip of the arrow is the parent (the lower side in the figure is the parent). In this example, channel 28 is the root (a parent channel that is not a child of any channel).

  Assume that the result of FIG. 9 is obtained as a result of searching for a parent-child relationship for a signal sequence in a certain time zone (a certain frame) of 32 channels of continuous signals. When the parent-child relationship for the signal sequence of the next frame also tends to obtain a result close to the result of the previous frame, the channels are grouped into several groups as in the fourth embodiment and the fourth embodiment. The method of dividing is effective.

In the example of FIG. 9, channel 8 and channel 3 are both parents with four children. In this case, it is divided between channel 8 and channel 18 (parent channel of channel 8). Further, the channel 3 and channel 1 (the parent channel of channel 3) are divided. That is, channel numbers 31, 10, 7, 19, 22, 29, 9, 12, 23, 26, and 8 are the first set. In addition, channel numbers 16, 21, 18, 14, 25, 13, 20, 5, 15, 11, 30, 0, 2, 4, 17, 3 are set as the second set. Further, channel numbers 27, 1, 24, 6, and 28 are assumed to be the third set. By dividing in this way, the number of combinations of channels that require calculation of the cross-correlation value C _{j, ρ (j)} and value C _{j, ρ (j)} ² / E _{ρ (j)} can be reduced. Can be speeded up. That is, the threshold value T is determined in advance, and the division may be performed at the parent portion having a child equal to or greater than the threshold value.

  Specifically, the processing is as follows. Since the channel 28 is found to be the root from the result of the parent-child relationship search in the previous frame, the channel 28 is marked as root (flagged). Then, from the root to the leaf (child channel), a parent having a child equal to or greater than the threshold T is searched. If a matching channel is found, mark it (root the channel). In this way, in the example of FIG. 9, channel 3 and channel 8 are marked (channel 3 and channel 8 are the roots of each set). In the next frame, the parent-child relationship is searched in three groups.

[Experiment]
A 512-channel MEG signal (cerebral magnetic field measurement signal) was input. The results of the compression coding experiment are shown in FIGS. FIG. 10 shows the relationship between the compression time and the code amount. The horizontal axis represents the ratio between the compression time and the real time (signal sequence time), and the vertical axis represents the ratio between the code amount after compression and the input signal amount. A ratio of compression time to real time of 100% indicates that the signal sequence can be compressed at the same speed as the signal sequence is input. A small ratio between the code amount after compression and the input signal amount indicates that the compression rate is high (the code amount is small). The solid line is the result of encoding by the encoding apparatus of the second embodiment, and the dotted line is the result of encoding by the conventional method. In both cases, the code amount decreases as the number of channels increases (the compression rate increases), but the compression time increases. In the conventional method, when the number of channels increases to 64, the compression time increases abruptly and processing in real time becomes impossible (exceeds 100%). On the other hand, in the case of the method of the second embodiment, the compression rate is not much different from the conventional method, but the compression time when the number of channels is large is significantly shortened. That is, when the parent-child relationship determining apparatus of the present invention is used, it can be seen that the compression time can be shortened while maintaining the compression rate, and this effect is particularly remarkable when the number of channels is large.

  FIG. 11 shows the relationship between the number of channels to be searched and the compression time. The horizontal axis indicates the number of channels to be searched, and the vertical axis indicates the ratio between the compression time and the real time (the time of the signal sequence) in the index display. When 512 channels are set as the search range, the method of the second embodiment shows that the processing can be performed about 1000 times faster than the conventional method. As a result, even with 512 channel coding, the method of the second embodiment can be processed almost in real time. As described above, by using the parent-child relationship determination device of the present invention, it is possible to perform processing in which the processing time does not increase rapidly even when the number of channels to be analyzed increases.

[Fifth Embodiment]
In the first to fourth embodiments, the value C _{j, ρ (j)} ² / E _{ρ (j)} obtained by dividing the square of the cross-correlation C _{j, ρ (j)} by the energy E _{ρ (j)} is used as an element. It generates a vector G which were generated vector G _S by sorting in descending order. And in the parent-child relationship search part, it processed in the element order. In other words, processing is performed from the largest value among the values C _{j, ρ (j)} ² / E _{ρ (j) obtained} by dividing the square of the cross-correlation C _{j, ρ (j)} by the energy E _{ρ (j).} Is going. In the present embodiment, as an embodiment a further generalization of the second embodiment, without limitation to the method of producing a vector G and the vector G _S, the cross-correlation C _j, energy squared _{ρ (j)} E _{ρ ( A} method of performing processing from the largest value among the values C _{j, ρ (j)} ² / E _{ρ (j)} divided by _j) will be described.

  FIG. 12 is a diagram illustrating a functional configuration example of the parent-child relationship determination device according to the present embodiment. Moreover, FIG. 13 is a figure which shows the processing flow of the parent-child relationship determination apparatus of this embodiment. The parent-child relationship determination device 700 corresponds to the parent-child relationship determination device 205 (cross-correlation / energy calculation unit 220, vector generation unit 230, vector element sorting unit 240, parent-child relationship search unit 250) of the encoding device 200 of the second embodiment. (See FIG. 4). The parent-child relationship determination apparatus 700 includes a cross-correlation / energy calculation unit 220, a quotient calculation unit 730, and a parent-child relationship search unit 750.

The process (S220) of the cross correlation / energy calculation unit 220 is the same as that of the second embodiment. The quotient calculation unit 730 calculates the square of the cross-correlation C _{j, ρ (j)} (where j is an integer from 1 to M and ρ (j) is an integer from 1 to M other than j) as the energy E _{ρ (j )} Divided by C _{), ρ (j)} ² / E _{ρ (j)} is obtained for all combinations of channel j and channel ρ (j) (S730).

The parent-child relationship search unit 750 selects the quotient (value C _{j, ρ (j)} ² / E _{ρ (j)} ) having the maximum value (S751). Next, the parent-child relationship search unit 750 confirms whether the parent channel of M-1 channels has been determined (S752). If step S752 is Yes, the searched parent-child relationship is output, and step S750 is terminated. When step S752 is No, it is confirmed whether the parent channel of the channel j corresponding to the selected value C _{j, ρ (j)} ² / E _{ρ (j)} is determined (S753). When step S753 is Yes, it progresses to step S757. If step S753 is No, it is confirmed whether the channel ρ (j) is not a descendant of the channel j (S754). In other words, make sure that you do not create a parent-child loop. If step S754 is Yes, the process proceeds to step S757. If step S754 is No, channel ρ (j) is set as the parent channel of channel j (S755). Then, the next largest quotient (value C _{j, ρ (j)} ² / E _{ρ (j)} ) is selected, and the process returns to step S152 (S157).

As one of methods for efficiently selecting the quotient (value C _{j, ρ (j)} ² / E _{ρ (j)} ) in the descending order, as shown in the first to fourth embodiments, a vector G and a method of producing a vector G _S. However, it is not limited to this method. In addition, there is a method in which the elements of the vector G are sorted in ascending order to generate the vector G _S ′, and the parent channel is searched from the element having the largest element order.

The first point of the present invention is that the parent-child relationship is not searched in order from “root to leaf”, but the parent-child relationship does not become a loop (the parent does not become a descendant of the child). It is to make a tree structure. The second point is to search for a parent-child relationship in descending order of the quotient (value C _{j, ρ (j)} ² / E _{ρ (j)} ). If the first point is satisfied, it is possible to perform processing without increasing the processing time rapidly even if the number of channels to be analyzed increases. Further, if the second point is satisfied, the tree structure can be made more efficiently.

  In the following modification, another example that satisfies the above-described points of the present invention will be shown. In these descriptions, in order to avoid redundant description, the cross-correlation / energy calculation units 120, 220, 320, 420, 420 ′, vector generation units 130, 230, 430, vector elements of the first to fourth embodiments are used. How to change the sorting units 140, 240, and 440 will be described.

[Modification 1]
The cross-correlation / energy calculation unit 1120 instead of the cross-correlation / energy calculation unit 120, 220, 320, 420, 420 ′ is configured so that all two channels m1 and m2 in the M channels (where m1 and m2 are 1). The energy E ^to _{m1, m2} after the subtraction process for the combination of the integers of M or less and m1 ≠ m2)

のように求める（Ｓ１１２０）。ただし、ｓ^〜 _{ｍ１，ｍ２}は、ｍ２チャネルを親チャネルとした場合のｍ１チャネルの減算処理後の信号である。なお、Ｅ^〜 _{ｍ１，ｍ２}は、ｍ１とｍ２とを入れ替えた場合異なる値となるので、Ｅ^〜 _{ｍ１，ｍ２}とＥ^〜 _{ｍ２，ｍ１}の両方を求める必要がある。

(S1120). Here, s ^to _{m1 and m2} are signals after the subtraction process of the m1 channel when the m2 channel is the parent channel. In addition, since E ^~ _{m1, m2} becomes a different value when m1 and m2 are replaced, it is necessary to obtain both E ^~ _{m1, m2} and E ^~ _{m2, m1} .

The vector generation unit 1130 instead of the vector generation units 130, 230, and 430 generates a vector G having all the energy E ^to _{m1 and m2} after the subtraction process as elements (S1130).

Vector elements sorting unit 1140 in place of the vector elements sorting unit 140,240,440 are rearranged in ascending order element of the vector G, to generate the vector _{G S} after sorting (S1140).

  Other components and the processing flow are the same as those in the first to fourth embodiments. Even if it processes in this way, the two points of the above-mentioned this invention are satisfied, and the same effect can be acquired.

[Modification 2]
The cross-correlation / energy calculation unit 1220 instead of the cross-correlation / energy calculation units 120, 220, 320, 420, and 420 ′ is configured so that all two channels m1 and m2 in M channels (where m1 and m2 are 1). The energy E ^to _{m1, m2} after the subtraction process for the combination of the integers of M or less and m1 ≠ m2)

のように求める（Ｓ１２２０）。ただし、γ_ａ、γ_ｂ、γ_ｃは重みである。なお、Ｅ^〜 _{ｍ１，ｍ２}は、ｍ１とｍ２とを入れ替えた場合異なる値となるので、Ｅ^〜 _{ｍ１，ｍ２}とＥ^〜 _{ｍ２，ｍ１}の両方を求める必要がある。

(S1220). However, γ _a , γ _b , and γ _c are weights. In addition, since E ^~ _{m1, m2} becomes a different value when m1 and m2 are replaced, it is necessary to obtain both E ^~ _{m1, m2} and E ^~ _{m2, m1} .

The vector generation unit 1130 instead of the vector generation units 130, 230, and 430 generates a vector G having all the energy E ^to _{m1 and m2} after the subtraction process as elements (S1130).

Vector elements sorting unit 1140 in place of the vector elements sorting unit 140,240,440 are rearranged in ascending order element of the vector G, to generate the vector _{G S} after sorting (S1140).

  Other components and the processing flow are the same as those in the first to fourth embodiments. Even if it processes in this way, the two points of the above-mentioned this invention are satisfied, and the same effect can be acquired.

In the calculation for _obtaining the energy E ^to _{m1 and m2} after the subtraction process, the first modification uses signals of the same time, and the second modification also uses signals of previous and subsequent times. However, it is not necessary to limit to these, and a signal from time na to n + b may be used. If a and b are integers greater than or equal to 0, the equations for _obtaining the energy E ^to _{m1 and m2} after the subtraction processing of the first and second modifications can be expressed as follows.

なお、ａ＝ｂ＝０のときが変形例１、ａ＝ｂ＝１のときが変形例２に相当する。また、例えば、ａ＝１、ｂ＝τ＋１の場合の重み（γ_―１，γ_０，γ_１，γ_２，…，γ_τ―２，γ_τ―１，γ_τ，γ_τ＋１）があるときに、（γ_２，…，γ_τ―２）は０であると制限した下で、Ｅ^〜 _{ｍ１，ｍ２}が小さくなるように、（γ_―１，γ_０，γ_１）と（γ_τ―１，γ_τ，γ_τ＋１）を求めてもよい。

A case where a = b = 0 corresponds to the first modification example, and a case where a = b = 1 corresponds to the second modification example. Also, for example, when there are weights (γ ₋₁ , γ ₀ , γ ₁ , γ ₂ ,..., Γ _τ−2 , γ _τ−1 , γ _τ , γ _{τ + 1} ) when a = 1 and b = τ + 1. _{to, (γ 2, ..., γ} τ-2) under which the limitations and is ^0, as _{E ~ m1, m @ 2} is _{reduced, (γ -1, γ 0,} γ 1) and (gamma _{tau- 1} , γ _τ , γ _{τ + 1} ).

[Modification 3]
The cross-correlation / energy calculation unit 1320 instead of the cross-correlation / energy calculation units 120, 220, 320, 420, and 420 ′ is configured so that all two channels m1 and m2 in M channels (where m1 and m2 are 1). E ^~ _{m1, m2} after subtraction processing for combinations of integers of M or less and m1 ≠ m2)

のように求める（Ｓ１３２０）。なお、Ｅ^〜 _{ｍ１，ｍ２}は、ｍ１とｍ２とを入れ替えた場合異なる値となるので、Ｅ^〜 _{ｍ１，ｍ２}とＥ^〜 _{ｍ２，ｍ１}の両方を求める必要がある。

(S1320). In addition, since E ^~ _{m1, m2} becomes a different value when m1 and m2 are replaced, it is necessary to obtain both E ^~ _{m1, m2} and E ^~ _{m2, m1} .

The vector generation unit 1230 instead of the vector generation units 130, 230, and 430 generates a vector G having all the elements E ^to _{m1 and m2} after the subtraction process as elements (S1130).

Vector elements sorting unit 1140 in place of the vector elements sorting unit 140,240,440 are rearranged in ascending order element of the vector G, to generate the vector _{G S} after sorting (S1140).

  Other components and the processing flow are the same as those in the first to fourth embodiments. Even if it processes in this way, the two points of the above-mentioned this invention are satisfied, and the same effect can be acquired.

[Modification 4]
The cross-correlation / energy calculation unit 1420 instead of the cross-correlation / energy calculation unit 120, 220, 320, 420, 420 ′ is configured so that all two channels m1 and m2 in M channels (where m1 and m2 are 1). E ^~ _{m1, m2} after subtraction processing for combinations of integers of M or less and m1 ≠ m2)

のように求める。なお、Ｅ^〜 _{ｍ１，ｍ２}は、ｍ１とｍ２とを入れ替えた場合異なる値となるので、Ｅ^〜 _{ｍ１，ｍ２}とＥ^〜 _{ｍ２，ｍ１}の両方を求める必要がある。また、全てのチャネルｍの信号列のＥ_ｍを

Seek like. In addition, since E ^~ _{m1, m2} becomes a different value when m1 and m2 are replaced, it is necessary to obtain both E ^~ _{m1, m2} and E ^~ _{m2, m1} . Also, the E _m of the signal train of all channels m

のように求める（Ｓ１４２０）。

(S1420).

A vector generation unit 1430 instead of the vector generation units 130, 230, and 430 is
E _j −E ^to _{j, ρ (j)}
Are obtained for all combinations of channel j and channel ρ (j), and a vector G having all the obtained values as elements is generated (S1430).

Vector elements sorting unit 1440 in place of the vector elements sorting unit 140,240,440 are rearranged in descending order of elements of the vector G, to generate the vector _{G S} after sorting (S1440).

  Other components and the processing flow are the same as those in the first to fourth embodiments. Even if it processes in this way, the two points of the above-mentioned this invention are satisfied, and the same effect can be acquired.

As in the first and second modifications, the signals from time na to n + b may also be used in the third and fourth modifications. In this case, an equation for _obtaining E ^to _{m1 and m2} after the subtraction processing can be expressed as follows.

なお、ａ＝ｂ＝０のときが変形例３、変形例４に相当する。

A case where a = b = 0 corresponds to the third modification and the fourth modification.

Further, the weight γ of the above-described modified examples 1 to 4 may not be limited to the expression (3). Equation (3) is
This is an equation for obtaining a weight that minimizes the energy of a child channel that is differentially expressed using a signal sequence of the same time (a signal sequence divided into the same time zone by the frame buffer unit). Instead of such weights, signal sequences having different time zones may be used. Or

の中央値（メジアン）を用いてもよい。また、複数の時刻の信号により減算処理後のＥ^〜 _{ｍ１，ｍ２}が小さくなるように重みγ_−ａ，…，γ_ｂ（ただし、ａとｂは０以上の整数）を求めてもよい。

The median (median) of may be used. Further, the weights γ _−a ,..., Γ _b (where a and b are integers of 0 or more) may be obtained so that E ^to _{m1 and m2} after subtraction processing are reduced by signals at a plurality of times.

[Modification 5]
The cross-correlation / energy calculation unit of this modification is the same as the cross-correlation / energy calculation unit 120, 220, 320, 420, 420 ′.

  The vector generation unit 1530 instead of the vector generation units 130, 230, and 430 is configured for all combinations of channel j and channel ρ (j).

を求め、求めた値の全てを要素とするベクトルＧを生成する（Ｓ１５３０）。

And a vector G having all the obtained values as elements is generated (S1530).

Vector elements sorting unit 1440 in place of the vector elements sorting unit 140,240,440 are rearranged in ascending order element of the vector G, to generate the vector _{G S} after sorting (S1540).

  Other components and the processing flow are the same as those in the first to fourth embodiments. Even if it processes in this way, the two points of the above-mentioned this invention are satisfied, and the same effect can be acquired.

  FIG. 14 shows a functional configuration example of a computer. In the encoding method and decoding method of the present invention, the recording unit 2020 of the computer 2000 reads a program for operating the computer 2000 as each component of the present invention, and the control unit 2010, the input unit 2030, the output unit 2040, etc. By operating, it can be executed by a computer. In addition, as a method of causing the computer to read, the program is recorded on a computer-readable recording medium, and the program recorded on the server or the like is read into the computer through a telecommunication line or the like. There is a method to make it.

The figure which shows the function structural example of the encoding apparatus of 1st Embodiment. The figure which shows the example of a processing flow of the encoding apparatus of 1st Embodiment. The figure which shows the example of a processing flow of the parent-child relationship search part 150. The figure which shows the function structural example of the encoding apparatus of 2nd Embodiment. The figure which shows the function structural example of the encoding apparatus of 3rd Embodiment. The figure which shows the example of a processing flow of the encoding apparatus of 3rd Embodiment. The figure which shows the function structural example of the encoding apparatus of 4th Embodiment. The figure which shows the example of a processing flow of the encoding apparatus of 4th Embodiment. The figure which shows the example of the parent-child relationship of the signal sequence of 32 channels. The figure which shows the relationship between compression time and code amount. The figure which shows the relationship between the number of channels to search, and compression time. The figure which shows the function structural example of the parent-child relationship determination apparatus of 5th Embodiment. The figure which shows the processing flow of the parent-child relationship determination apparatus of 5th Embodiment. The figure which shows the function structural example of a computer.

100, 200, 300, 400 Encoding device 105, 205, 305, 405, 406, 700 Parent-child relationship determination device 110, 210 Frame buffer unit 120, 220, 320, 420, 420 ′ Cross correlation / energy calculation unit 130, 130 ', 230, 230', 430, 430 'Vector generation unit 140, 240, 440 Vector element sorting unit 150, 250, 450, 750 Parent-child relationship search unit 160, 260, 360 Differential signal sequence generation unit 170, 270, 370 Code Encoding unit 300 encoding device 730 quotient calculation unit

Claims (16)

Between signal sequences for combinations of all two channels m1 and m2 (where m1 and m2 are integers of 1 to M and m1 ≠ m2) in M channels (where M is an integer of 3 or more) A cross-correlation C _{m1, m 2} and a cross-correlation / energy calculation unit for obtaining energy E _m of the signal sequence of all channels m (where m is an integer of 1 to M);
Value obtained by dividing the square of the cross-correlation C _{j, ρ (j)} (where j is an integer from 1 to M and ρ (j) is an integer from 1 to M other than j) by energy E _{ρ (j)} A quotient calculator for determining C _{j, ρ (j)} ² / E _{ρ (j)} for all combinations of channel j and channel ρ (j);
The value C _{j, ρ (j)} ² / E _{ρ (j)} is selected in descending order, and the parent channel of channel j corresponding to the selected value C _{j, ρ (j)} ² / E _{ρ (j)} is If the channel ρ (j) is not a descendant of the channel j, the channel ρ (j) is determined as the parent channel of the channel j, or the channel ρ ( If j) is a descendant of channel j, the parent-child repeats the process of selecting the next largest value C _{j, ρ (j)} ² / E _{ρ (j)} until the parent channels of M−1 channels are determined. A relationship search unit;
A parent-child relationship determining device. Between signal sequences for combinations of all two channels m1 and m2 (where m1 and m2 are integers of 1 to M and m1 ≠ m2) in M channels (where M is an integer of 3 or more) A cross-correlation C _{m1, m 2} and a cross-correlation / energy calculation unit for obtaining energy E _m of the signal sequence of all channels m (where m is an integer of 1 to M);
Value obtained by dividing the square of the cross-correlation C _{j, ρ (j)} (where j is an integer from 1 to M and ρ (j) is an integer from 1 to M other than j) by energy E _{ρ (j)} C _{j, ρ (j)} ² / E _{ρ (j)} is obtained for all combinations of channel j and channel ρ (j), and the obtained value C _{j, ρ (j)} ² / E _{ρ (j)} A vector generation unit that generates a vector G _S having elements arranged in descending order;
Selecting elements of the vector G _S in element order, checking whether the parent channel of the channel j corresponding to the selected element C _{j, ρ (j)} ² / E _{ρ (j)} is determined, and not determined; and If channel ρ (j) is not a descendant of channel j, then channel ρ (j) is the parent channel of channel j, and if channel ρ (j) is a descendant of channel j, then A parent-child relationship search unit that repeats the process of selecting a large value C _{j, ρ (j)} ² / E _{ρ (j)} to M−1 channel parent channels;
A parent-child relationship determining device. After subtraction processing for combinations of all two channels m1 and m2 (where m1 and m2 are integers of 1 to M and m1 ≠ m2) in M channels (where M is an integer of 3 or more) Using one or more weights γ _−a ,..., Γ _b (where a and b are integers of 0 or more) obtained by determining the energies E ^to _{m1 and m2} in a predetermined method,

The cross-correlation / energy calculation unit
A vector generation unit that generates a vector G _S having elements in which all of the energy E ^to _{m1, m2} are arranged in ascending order;
The elements of the vector G _S are selected in the element order, and it is confirmed whether the parent channel of the channel j corresponding to the selected elements E ^to _{j, ρ (j)} is determined, and the channel ρ (j) is not determined. If it is not a descendant of channel j, channel ρ (j) is the parent channel of channel j and is fixed, or if channel ρ (j) is a descendant of channel j, the next smallest E ^~ _j, a parent-child relationship search unit that repeats the process of selecting _{ρ (j)} until the parent channels of M−1 channels are determined;
A parent-child relationship determining device. Values E ^to _m1 for combinations of all two channels m1 and m2 (where m1 and m2 are integers of 1 to M and m1 ≠ m2) in M channels (where M is an integer of 3 or more) _{, M2} using one or more weights γ _−a ,..., Γ _b (where a and b are integers of 0 or more) obtained by a predetermined method,

The cross-correlation / energy calculation unit
A vector generation unit that generates a vector G having elements in which all of the values E ^to _{m1, m2} are arranged in ascending order;
The elements of the vector G _S are selected in the element order, and it is confirmed whether the parent channel of the channel j corresponding to the selected elements E ^to _{j, ρ (j)} is determined, and the channel ρ (j) is not determined. If it is not a descendant of channel j, channel ρ (j) is the parent channel of channel j and is fixed, or if channel ρ (j) is a descendant of channel j, the next smallest E ^~ _j, a parent-child relationship search unit that repeats the process of selecting _{ρ (j)} until the parent channels of M−1 channels are determined;
A parent-child relationship determining device. Values E ^to _m1 for combinations of all two channels m1 and m2 (where m1 and m2 are integers of 1 to M and m1 ≠ m2) in M channels (where M is an integer of 3 or more) _{, M2} using one or more weights γ _−a ,..., Γ _b (where a and b are integers of 0 or more) obtained by a predetermined method,

And calculate E _m of the signal trains of all channels m.

The cross-correlation / energy calculation unit
Vectors E j _-E ~ ^j, _[rho and _(j), calculated for the combination of all channels j and channel [rho (j), to generate the vector G _S with elements arranged in descending order all the values obtained A generator,
Select an element of the vector G _S to the element order, to check the selected elements E _j -E ~ ^j, the parent channel of the channel j corresponding to _{[rho (j)} is fixed, not fixed, and the channel [rho ( If j) is not a descendant of channel j, then channel ρ (j) is the parent channel of channel j and is fixed, or if channel ρ (j) is a descendant of channel j, the next largest E a parent-child relationship search unit that repeats the process of selecting _j −E ^to _{j, ρ (j)} until the parent channels of M−1 channels are determined;
A parent-child relationship determining device. Between signal sequences for combinations of all two channels m1 and m2 (where m1 and m2 are integers of 1 to M and m1 ≠ m2) in M channels (where M is an integer of 3 or more) A cross-correlation C _{m1, m 2} and a cross-correlation / energy calculation unit for obtaining energy E _m of the signal sequence of all channels m (where m is an integer of 1 to M);
For all combinations of channel j and channel ρ (j)

A vector generation unit that generates a vector G _S having elements in which all of the obtained values are arranged in ascending order;
Select an element of the vector G _S to the element order, to see if the parent channel of the channel j corresponding to the selected elements are determined, not determined, and, when the channel [rho (j) is not a descendant of channel j is , Channel ρ (j) is the parent channel of channel j, and if channel ρ (j) is a descendant of channel j, the process of selecting the element of the next smaller vector G _S is A parent-child relationship search unit that repeats until the parent channel of one channel is determined;
A parent-child relationship determining device. The parent-child relationship determining device according to any one of claims 1 to 6,
For the input multiple-channel signal sequence, the signal sequence of the channel for which the parent channel is determined is the difference signal sequence expressed by the difference from the signal sequence of the parent channel, and the signal sequence of the channel without the parent channel is unchanged Including a differential signal sequence generation unit that outputs the signal sequence as a differential signal sequence,
A parent-child relationship determination device that repeats the processing from the cross-correlation / energy calculation unit to the processing of the differential signal sequence generation unit by using the difference signal sequence as an input to the cross-correlation / energy calculation unit until a predetermined condition is satisfied. The parent-child relationship determining device according to any one of claims 1 to 6,
The M channels are a part of channels selected from the channels input to the parent-child relationship determining device. The parent-child relationship determining device. The parent-child relationship determining device according to claim 7,
The M channels are some or all of the channels selected from the channels input to the parent-child relationship determination device,
A parent-child relationship determination device characterized in that the number of M and the channel to be selected are variable for each repetition process. Between signal sequences for combinations of all two channels m1 and m2 (where m1 and m2 are integers of 1 to M and m1 ≠ m2) in M channels (where M is an integer of 3 or more) A cross-correlation / energy calculation step for _{obtaining the} cross-correlation C _{m1, m2} and the energy E _m of the signal sequence of all channels m (where m is an integer of 1 to M);
Value obtained by dividing the square of the cross-correlation C _{j, ρ (j)} (where j is an integer from 1 to M and ρ (j) is an integer from 1 to M other than j) by energy E _{ρ (j)} A quotient calculation step for determining C _{j, ρ (j)} ² / E _{ρ (j)} for all combinations of channel j and channel ρ (j);
The value C _{j, ρ (j)} ² / E _{ρ (j)} is selected in descending order, and the parent channel of channel j corresponding to the selected value C _{j, ρ (j)} ² / E _{ρ (j)} is If the channel ρ (j) is not a descendant of the channel j, the channel ρ (j) is determined as the parent channel of the channel j, or the channel ρ ( If j) is a descendant of channel j, the parent-child repeats the process of selecting the next largest value C _{j, ρ (j)} ² / E _{ρ (j)} until the parent channels of M−1 channels are determined. A relationship search step;
A method for determining a parent-child relationship. Between signal sequences for combinations of all two channels m1 and m2 (where m1 and m2 are integers of 1 to M and m1 ≠ m2) in M channels (where M is an integer of 3 or more) A cross-correlation / energy calculation step for _{obtaining the} cross-correlation C _{m1, m2} and the energy E _m of the signal sequence of all channels m (where m is an integer of 1 to M);
Value obtained by dividing the square of the cross-correlation C _{j, ρ (j)} (where j is an integer from 1 to M and ρ (j) is an integer from 1 to M other than j) by energy E _{ρ (j)} C _{j, ρ (j)} ² / E _{ρ (j)} is obtained for all combinations of channel j and channel ρ (j), and the obtained value C _{j, ρ (j)} ² / E _{ρ (j)} A vector generation step of generating a vector G _S having elements arranged in descending order,
Selecting elements of the vector G _S in element order, checking whether the parent channel of the channel j corresponding to the selected element C _{j, ρ (j)} ² / E _{ρ (j)} is determined, and not determined; and If channel ρ (j) is not a descendant of channel j, then channel ρ (j) is the parent channel of channel j, and if channel ρ (j) is a descendant of channel j, then A parent-child relationship search step that repeats the process of selecting a large value C _{j, ρ (j)} ² / E _{ρ (j)} for M−1 channels until a parent channel is determined;
A method for determining a parent-child relationship. 10. or may be a parent-child relationship determining method of mounting 11 SL,
For the input multiple-channel signal sequence, the signal sequence of the channel for which the parent channel is determined is the difference signal sequence expressed by the difference from the signal sequence of the parent channel, and the signal sequence of the channel without the parent channel is unchanged A signal sequence generation step for outputting the signal sequence as a differential signal sequence,
A parent-child relationship determination method in which the difference signal sequence is input to the cross-correlation / energy calculation step and the cross-correlation / energy calculation step to the difference signal sequence generation step are repeated until a predetermined condition is satisfied. 10. or may be a parent-child relationship determining method of mounting 11 SL,
The M channels are a part of channels selected from the input channels. The parent-child relationship determination method, wherein: A parent-child relationship determining method according to claim 12 Symbol mounting,
The M channels are some or all of the channels selected from the input channels.
A parent-child relationship determination method characterized in that the number of M and the channel to be selected are variable for each repetition process.   A parent-child relationship determination program for operating a computer as the parent-child relationship determination device according to claim 1. 15. Symbol mounting parentage determination program computer-readable recording medium recording a.

Images