JP4733552B2 - PARCOR coefficient calculation device, PARCOR coefficient calculation method, program thereof, and recording medium thereof - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To calculate a PARCOR coefficient with computational quantity less than a conventional Burg method. <P>SOLUTION: A forward prediction error b'<SB>m</SB>in a model order m is calculated from a forward prediction error b'<SB>m-1</SB>, a backward prediction error b<SB>m-1</SB>and a PARCOR coefficient &nu;<SB>m-1, m-1</SB>in a model order m-1, and a backward prediction error b<SB>m</SB>in the model order m, is calculated from the forward prediction error b'<SB>m-1</SB>, the backward prediction error b<SB>m-1</SB>and the PARCOR coefficient &nu;<SB>m-1, m-1</SB>in the model order m-1. Moreover, an average energy A<SB>m</SB>in the model order m is calculated by multiplying an average energy A<SB>m-1</SB>in the model order m-1 by a value (an update coefficient) determined according to the model order m-1. Then, a PARCOR coefficient &nu;<SB>m, m</SB>in the model order m is calculated by dividing an inner product of the forward prediction error b'<SB>m</SB>and the backward prediction error b<SB>m</SB>in the model order m by the average energy A<SB>m</SB>. <P>COPYRIGHT: (C)2008,JPO&amp;INPIT

Description

  The present invention relates to a linear prediction analysis technique used for signal analysis and signal coding.

  A method for directly obtaining a PARCOR coefficient is known as one of all-pole linear prediction analysis. This calculation method is described in Non-Patent Document 1 as the Burg method.

  An outline of the Burg method will be described. In the Burg method, it is considered that the linear prediction model of model order m−1 is extended to a linear prediction model of model order m. That is, in the Burg method, the PARCOR coefficient of the linear prediction model is calculated from a recurrence relationship in which the PARCOR coefficient at a model order m that is one greater than the model order is calculated from the PARCOR coefficient at a certain model order m-1.

The given data _{x 1, x 2, ···,} and _{x N.} In this case, reference prediction error filter data (linear prediction model of model order _{m) x 1, x 2,} ···, average power _{P m} [wherein in the case through the case and backward forward pass _{x N} (1) . ] Is adopted on condition that it is minimized. However, γ _{m, k} (k = 1, 2,..., M) is a linear prediction coefficient of the model order m.

Here, γ _{m, k} (k = 1, 2,..., M) has a relationship represented by Expression (2), that is, Expression (3) [Levinson algorithm; see Non-Patent Document 1. ].

When the formula (1) is rewritten using the formula (2) or the formula (3), the formula (4) is obtained.

Here, b ′ _{m, i} is expressed by Equation (5). Let b ′ _{m, i} be called a forward prediction error.
In addition, b _{m, i} is expressed by Equation (6). Let b _{m, i} be called backward prediction error.

Note that the forward prediction error and the backward prediction error have a relationship represented by Expressions (7) and (8) [see Non-Patent Document 1]. ].

At this time, from the condition that the average output P _m is minimized, γ _{m, m} [so-called PARCOR coefficient. ], The formula (9) is obtained.

In other words, the PARCOR coefficient γ _{m, m} is the inner product of the forward prediction error b ′ _{m, i} (i = 1,..., N−m; the same applies hereinafter) and the backward prediction error b _{m, i} [of the equation (9) Corresponds to a molecule. ] Of the respective energies of the forward prediction error b ′ _{m, i} and the backward prediction error b _{m, i} [hereinafter referred to as “average energy”]. This corresponds to the denominator of Equation (9). ] Divided by].
Note that the linear prediction coefficient γ _{m, k} (k = 1, 2,..., M−1) when the model order is m (hereinafter also simply referred to as “order m”) is given by Equation (3). . The linear prediction coefficient γ _{m, k} (k = m) is given by the equation (9) as a PARCOR coefficient.
By Hino Mikio, “Spectral Analysis”, Asakura Shoten, 1979

As apparent from the equations (7), (8), and (9), in the conventional Burg method, every time the order of the PARCOR coefficient is increased by one, two types of prediction errors b ′ _{m, i} , b _{m, i} (i = 1,..., N−m; the same applies hereinafter) _, the inner product of the two types of updated prediction errors b ′ _{m, i} , b _{m, i} , and the average energy are calculated. There is a problem that the processing amount is very large according to the number of data N and the order m.

  In view of the above problems, an object of the present invention is to provide a PARCOR coefficient calculation apparatus, method, program, and recording medium that calculate a PARCOR coefficient with a smaller amount of calculation than the conventional Burg method.

In order to solve the above problems, the present invention is configured as follows. That is, when the PARCOR coefficient of the linear prediction model is calculated from the recurrence relationship for obtaining the PARCOR coefficient in the model order m-1 from the PARCOR coefficient in the model order m-1, the model energy is calculated as the average energy A _m-1 in the model order m-1. value determined in accordance with the order m-1 (hereinafter, referred to as. "update coefficient") includes a mean energy calculation means for calculating an average energy a _m in the model order m by multiplying, using the average energy a _m The PARCOR coefficient at the model order m is calculated.
That is, the model order from the forward prediction error b ′ _{m−1 in} the model order m−1, the backward prediction error b _{m−1 in the} model order m−1, and the PARCOR coefficients γ _{m−1 and m−1 in the} model order _{m−1. 'calculates m,} forward prediction error _b in model order _m-1' forward prediction error b in m _m-1, PARCOR coefficients in backward prediction error _{b m-1} and the model order m-1 in the model order m-1 gamma A backward prediction error b _m in the model order _m is calculated from _{m−1 and m−1} . Furthermore, the average energy Am in the model order _m is calculated by multiplying the average energy Am _-1 in the model order m-1 by a value (update coefficient) determined according to the model order m-1. Then, the PARCOR coefficient γ _{m, m} in the model order _m is calculated by dividing the inner product of the forward prediction error b ′ _m and the backward prediction error b _m in the model order m by the average energy A _m in the model order m.
In the conventional Burg method, every time the model order is increased by one, the average energy [corresponds to the denominator of equation (9). As described above, the average energy at the model order m is obtained by multiplying the average energy A _m-1 at the model order m-1 by a value (update coefficient) determined according to the model order m-1. _Am is calculated. In other words, the average energy A _m in the model order m, an approximation that depends on the average energy A _m-1 in the model order m-1.

Alternatively, the PARCOR coefficient may be calculated as follows. That is, when the PARCOR coefficient of the linear prediction model is calculated from the recurrence relationship for obtaining the PARCOR coefficient in the model order m from the PARCOR coefficient in the model order m−1, a preset condition (hereinafter referred to as “determination condition”) is calculated. And the average energy at the model order m by multiplying the average energy A _m−1 at the model order m−1 by a value (update coefficient) determined according to the model order m−1. error calculating an average energy calculating means for calculating the a _m, an average of the energy of the forward prediction error b _'m and backward prediction error b _m in the model order m (hereinafter, referred to as. "error dependent mean energy B _m') to A dependent average energy calculating means, and a forward prediction error b ′ _m and a backward prediction error in the model order m. The PARCOR coefficient γ _{m, m} at the model order _m is calculated by dividing the inner product with the difference b _m by the average energy A _{m at the} model order _m or the error-dependent average energy B _m according to the determination result of the determination means. Is possible.
That is, the model order from the forward prediction error b ′ _{m−1 in} the model order m−1, the backward prediction error b _{m−1 in the} model order m−1, and the PARCOR coefficients γ _{m−1 and m−1 in the} model order _{m−1. 'calculates m,} forward prediction error _b in model order _m-1' forward prediction error b in m _m-1, PARCOR coefficients in backward prediction error _{b m-1} and the model order m-1 in the model order m-1 gamma A backward prediction error b _m in the model order _m is calculated from _{m−1 and m−1} . Then, it is determined whether or not the determination condition is satisfied. If it is determined that the determination condition is satisfied, the PARCOR coefficient is calculated as follows. The average of the energy of the forward prediction error b _'m and backward prediction error b _m in the model order m (error depending average energy B _m) is calculated. Next, the PARCOR coefficient γ _{m, m} in the model order _m is calculated by dividing the inner product of the forward prediction error b ′ _m and the backward prediction error b _m in the model order m by the error-dependent average energy B _m in the model order m. To do. When it is determined that the determination condition is not satisfied, the PARCOR coefficient is calculated as follows. The average energy Am in the model order _m is calculated by multiplying the average energy Am _-1 in the model order m-1 by a value (update coefficient) determined according to the model order m-1. Next, the PARCOR coefficient γ _{m, m} in the model order _m is calculated by dividing the inner product of the forward prediction error b ′ _m and the backward prediction error b _m in the model order m by the average energy A _m in the model order m.
As described above, the calculation of the PARCOR coefficient by the conventional Burg method and the calculation of the PARCOR coefficient based on the approximation of the average energy may be combined depending on the success or failure of the determination condition.

Incidentally, the calculation of the average energy A _m is determined depending on the model orders m-1 the prediction error energy C _m-1 in the model order m-1, which corresponds to one of the energy of the forward prediction error or backward prediction errors calculating a prediction error energy C _m in model order m by multiplying the value (update coefficient), further, it is the average of the energy in the model order m of the prediction error energy C _m and the other of the prediction error in the model order m It may be replaced by a configuration of calculating the average energy a _m in model order m.
As described above, the prediction error energy corresponding to one of the forward prediction error and the backward prediction error is approximated by multiplying by the update coefficient.

Assuming that the above-described determination condition is that the model order is a preset order, the inner product of the forward prediction error b ′ _m and the backward prediction error b _m in the model order m is obtained, and the model order m satisfies the determination condition. If the model order m is not the error-dependent average energy B _m in the model order _m and the model order m does not satisfy the determination condition, the PARCOR coefficient γ _m in the model order m is obtained by dividing by the average energy A _m in the model order _{m. m} is calculated. That is, the PARCOR coefficient is calculated by the conventional Burg method with a preset model order.
Further, the “predetermined order” may be an order equal to or less than a predetermined threshold value or an order smaller than the threshold value. That is, in the low-order model order, the PARCOR coefficient is calculated by the conventional Burg method.
By introducing such a determination condition, accumulation of approximate errors is prevented, or deterioration in accuracy of the PARCOR coefficient in a low-order model order is prevented.

The update coefficient may be obtained by subtracting the square of the PARCOR coefficient γ _{m−1, m−1 in} the model order m−1 from 1. The basis for this will be described later.

  Further, the computer can be operated as a PARCOR coefficient calculation device by a PARCOR coefficient calculation program that causes the computer to function as the PARCOR coefficient calculation device of the present invention. Then, by using a computer-readable program recording medium that records the PARCOR coefficient calculation program, it is possible to cause another computer to function as a PARCOR coefficient calculation device or to distribute the PARCOR coefficient calculation program.

  According to the present invention, since the average energy for calculating the PARCOR coefficient is approximately calculated, the amount of calculation for calculating the PARCOR coefficient can be reduced. This is particularly effective when the model order is high or the number of data is large.

<< Approximate calculation of average energy >>
The gist of the present invention is to reduce the amount of calculation for calculating the PARCOR coefficient by approximating the average energy corresponding to the denominator of the equation (9). Thus, approximate calculation of average energy will be described.

The update of the prediction error represented by Expression (7) and Expression (8) acts in the direction of reducing each prediction error. Therefore, the forward prediction error after update depends on the forward prediction error before update and the backward prediction error before update, and the backward prediction error after update depends on the forward prediction error before update and the backward prediction error before update. Focusing on this, consider the approximation that the forward prediction error and the backward prediction error energy are expected to be approximately equal, and they match. In this case, the average energy A _m can be approximated represented as formula (10). Here, E [•] represents an expected value. That is, the average energy _Am is proportional to E [b _{m, i} ² ].

By the way, E [b _{m, i} ² ] has a relationship represented by the formula (11) [see Non-Patent Document 1 above. ].

From equation (10) and (11), obtaining a relation expressed by the formula (12) for the average energy _{A m.} However, this relationship is approximate.

Expression (9) is rewritten into Expression (13) using Expression (12).

In previous Burg method, was successively seeking m following average energy A _m from the forward prediction error and backward prediction error. That is, the sum of squares of the forward prediction error and the backward prediction error [see Equation (9). ].
On the other hand, according to the formula (12) which is the gist of the present invention, the m-th order average energy _Am can be obtained from the m-1 order average energy A _m-1 and the m-1 order PARCOR coefficient. it can.

The Burg method extends the linear prediction model of the order m−1 to the linear prediction model of the order m, and the m−1 order average energy A _m−1 and the m−1 order PARCOR coefficient are (m order). Is known at the stage of determining the PARCOR coefficient. Therefore, as represented by the equation (12), the m-th order average energy _Am can be obtained by one subtraction and two multiplications. This reduction in the amount of computation is directly linked to a reduction in the amount of computation required for calculating the m-th order PARCOR coefficient [see Equation (13). ].

<< First Embodiment >>
A first embodiment of the present invention will be described with reference to the drawings.
<PARCOR coefficient calculation apparatus according to the first embodiment>
As illustrated in FIG. 1, the PARCOR coefficient calculation device (1) includes an input unit (11) to which a keyboard or the like can be connected, an output unit (12) to which a liquid crystal display or the like can be connected, a CPU (Central Processing Unit; 14). [A cache memory or the like may be provided. RAM (Random Access Memory) (15), ROM (Read Only Memory) (16) and external storage device (17) which is a hard disk, and these input unit (11), output unit (12), A CPU (14), a RAM (15), a ROM (16), a bus (18) connected so as to exchange data between the external storage devices (17), and the like are provided. If necessary, the PARCOR coefficient calculation device (1) may be provided with a device (drive) that can read and write a storage medium such as a CD-ROM. A physical entity having such hardware resources includes a general-purpose computer.

  The external storage device (17) of the PARCOR coefficient calculation device (1) stores and stores a program for calculating the PARCOR coefficient, data necessary for processing of the program, and the like. Further, data obtained by the processing of these programs is appropriately stored and stored in the RAM (15) or the like.

Further, it is assumed that time discrete signals x ₁ , x ₂ ,..., X _{N obtained} by sampling the signals at predetermined time intervals are stored and stored in the external storage device (17) as data. Signals handled in each embodiment are acoustic signals such as human voice, music, and noise / noise. Of course, the signal is not limited to an acoustic signal, and may be a biological signal, for example.
Time discrete signal _{x 1,} x 2, · · ·, instead of a configuration _{in which x N} are stored previously stored in the external storage device (17), for example, the time from a storage medium such as a CD-ROM discrete signal _{x 1} , X ₂ ,..., X _N are read and stored in the external storage device (17), or the PARCOR coefficient calculation device (1) is provided with a microphone, and the collected sound signal obtained by this microphone is obtained. a / D converted and time discrete signals _{x 1, x 2, ···,} is considered such that the structure to obtain _{x N.} In any case, the constituent elements (means) necessary for executing A / D conversion from the collected sound signal to the acoustic signal and the data providing technology from the recording medium are achieved by conventional means of a known technique. Therefore, explanation and illustration are omitted.

Also in the external storage device (17) is time-discrete signal _{x 1,} x 2, · · ·, a program for setting an initial value from _{x N,} the program for calculating the PARCOR coefficients, and updates the forward prediction errors program for a program for updating the backward prediction errors, a program for calculating the average energy a _m is stored stored. In addition, a control program for controlling processing based on these programs is also stored as appropriate.

  In the PARCOR coefficient calculation device (1), each program stored in the external storage device (17) and data necessary for processing each program are read into the RAM (15) as necessary, and the CPU (14) Interpreted and processed. As a result, the CPU (14) realizes predetermined functions (initial value setting unit, PARCOR coefficient calculation unit, forward prediction error update unit, backward prediction error update unit, average energy calculation unit, control unit), so that the PARCOR coefficient Calculation is realized.

<PARCOR coefficient calculation process of the first embodiment>
Next, the flow of the PARCOR coefficient calculation process in the PARCOR coefficient calculation device (1) will be described descriptively with reference to FIGS. In the description of the PARCOR coefficient calculation process, the Burg method extends the linear prediction model of the order m-1 to the linear prediction model of the order m, so first, the process in the case of the first order (m = 1) will be described. Next, the process in the case of extending from the first order to the second order will be described, and then the process in the case of extending from the (m−1) th order to the mth order will be described (that is, an explanation similar to the so-called mathematical induction method) To do.) In this embodiment, the PARCOR coefficient of the Mth-order linear prediction model is calculated.

[When m = 1]
First, the initial value setting unit (140) reads the time discrete signals x ₁ , x ₂ ,..., X _N from the external storage device (17), sets m = 1, and the forward prediction error b ′ _{1, i} = x. _i , backward prediction error b _{1, i} = x _{i + 1} , average energy A ₁ = x ₁ ² + x ₂ ² +... + x _N ² are set to initial values, b ′ _{1, i} , b _{1, i} , A ₁ Each is stored in a predetermined storage area of the RAM (15) (step S1). Hereinafter, when the description “reads XX from RAM” is used, it means “read XX from a predetermined storage area in which XX is stored in the RAM”.
Here, it should be noted that the subscript i does not indicate a specific value but corresponds to the index i in the equation (13) [the same applies hereinafter. The same applies to the subscript i in the figure. ]. That is, the number of data corresponding to the number of values that can be taken by the index i exists.

Next, the PARCOR coefficient calculation unit (141) reads each of b ′ _{1, i} , b _{1, i} , A ₁ from the RAM (15), performs the calculation of Expression (13), and performs the primary PARCOR coefficient γ _{1. , 1} and the PARCOR coefficient γ ₁ , ₁ is stored in a predetermined storage area of the RAM (15) (step S2).

Following step S2, the control unit (190) determines whether m = M is successful (step S3a). If m ≠ M, the control unit (190) controls to perform the extension process from the primary to the secondary. If m = M, the control unit (190) controls to end the PARCOR coefficient calculation process. If M> 1, since m = 1 at this stage, the extension process from the primary to the secondary is performed.
If M = 1, the desired PARCOR coefficient is the first-order PARCOR coefficient γ _1,1 and can be terminated at this stage.

[Extension from primary to secondary]
The control unit (190) sets a value obtained by adding 1 to the current m as a new m (step S4). That is, at this stage, the control unit (190) performs control to perform the following processing with m = 2.

First, the forward prediction error updating unit (142), RAM (15) from _{b '1, i + 1,} b 1, i + 1, reads the gamma _{1, 1,} respectively, calculates the go and secondary forward prediction equation (8) The error b'2 _{, i} is calculated, and the forward prediction error b'2 _{, i} is stored in a predetermined storage area of the RAM (15) (step S5).

Next, the backward prediction error updating unit (143) reads b ′ _{1, i} , b _{1, i} , γ _1,1 from the RAM (15), performs the calculation of Expression (7), and performs the second backward The prediction error b _{2, i} is calculated, and the backward prediction error b _{2, i} is stored in a predetermined storage area of the RAM (15) (step S6).

Next, the average energy calculation unit (145) reads A ₁ and γ _1,1 from the RAM (15), calculates the secondary average energy A ₂ by performing the calculation of the equation (12), and calculates the average energy. A ₂ is stored in a predetermined storage area of the RAM (15) (step S7).

Next, the PARCOR coefficient calculation unit (141) reads each of b ′ _{2, i} , b _{2, i} , and A ₂ from the RAM (15), performs the calculation of Equation (13), and performs the second-order PARCOR coefficient γ _{2. , 2} are calculated, and the PARCOR coefficients γ ₂ , ₂ are stored in a predetermined storage area of the RAM (15) (step S8).

If necessary, the linear prediction coefficient γ ₂ , ₁ can be calculated following the process of step S8. In this case, the linear prediction coefficient calculation unit (146) reads γ _1,1 and γ _2,2 from the RAM (15), calculates the linear prediction coefficient γ _2,1 by performing the calculation of Expression (3). The linear prediction coefficients γ ₂ , ₁ are stored in a predetermined storage area of the RAM (15) (step S9). The linear prediction coefficient calculation unit (146) is realized by the CPU (14) interpreting and executing a program for calculating the linear prediction coefficient from the PARCOR coefficient.

At this stage, the second-order PARCOR coefficient γ _2,2 , the second-order forward prediction error b ′ _{2, i} , the second-order backward prediction error b _{2, i} , and the second-order average energy A ₂ Is stored in the RAM (15) (the linear prediction coefficients γ _{2, 1} are also stored if necessary).

Subsequent to the process of step S8 (the process of step S9 when the process of step S9 is executed as necessary), the same process as step S3a is performed. That is, the control unit (190) determines whether m = M is successful (step S3b). If m ≠ M, the control unit (190) controls to perform the expansion process from the secondary to the tertiary. If m = M, the control unit (190) controls to end the PARCOR coefficient calculation process.
In this way, if m ≠ M, the extension is performed to the next higher order. Therefore, in the following, a general description will be given in the case of the expansion from the m−1 order to the m order.

It should be noted that at the stage of calculating the m−1 order PARCOR coefficient, the RAM (15) stores the m−1 order PARCOR coefficient γ _{m−1, m−1} , m−1 order. The forward prediction error b ′ _{m−1, i} , m−1 updated to the backward prediction error b _{m−1, i} , m−1 order average energy A _m−1 is stored [ If necessary, m-2 linear prediction coefficients γ _{m−1, k} (k = 1, 2,..., M−2) are also stored. ].

[Extension from m-1 order to m order]
The control unit (190) performs control to perform the following processing with a value obtained by adding 1 to the current m as a new m (step S4).

First, the forward prediction error updating unit (142) reads b ′ _{m−1, i + 1} , b _{m−1, i + 1} , γ _{m−1, m−1} from the RAM (15), and calculates the equation (8). To calculate the _m-th forward prediction error b ′ _{m, i} and store the forward prediction error b ′ _{m, i} in a predetermined storage area of the RAM (15) (step S5).

Next, the backward prediction error update unit (143) reads b ′ _{m−1, i} , b _{m−1, i} , γ _{m−1, m−1} from the RAM (15), and calculates Equation (7). An m-th order backward prediction error b _{m, i} is calculated by performing calculation, and the backward prediction error b _{m, i} is stored in a predetermined storage area of the RAM (15) (step S6).

Next, the average energy calculation unit (145) reads each of A _m−1 , γ _{m−1, and m−1} from the RAM (15), performs the calculation of Expression (12), and calculates the mth average energy A _m. It is calculated and stored in the average energy _{a m} in a predetermined storage area of the RAM (15) (step S7).

Next, PARCOR coefficient calculation unit (141), RAM (15) from _{b 'm, i, b m} , i, is read, respectively _{A m,} PARCOR coefficient of m-th order by performing the operation of Equation (13) gamma _{m ,} M are calculated, and the PARCOR coefficient γ _{m, m} is stored in a predetermined storage area of the RAM (15) (step S8).

If necessary, following the process of step S8, linear prediction coefficients γ _{m, k} (k = 1, 2,..., M−1) can be calculated. In this case, the linear prediction coefficient calculation unit (146) reads from the RAM (15) γ _{m−1, k} (k = 1, 2,..., M−2), γ _{m−1, m−1} , γ. Each of _{m and m} is read and the calculation of equation (3) is performed to calculate m−1 linear prediction coefficients γ _{m, k} (k = 1, 2,..., m−1). The linear prediction coefficients γ _{m, k} (k = 1, 2,..., M−1) are stored in a predetermined storage area of the RAM (15) (step S9).

In this stage, the m-th order PARCOR coefficient γ _{m, m} , the m-th order forward prediction error b ′ _{m, i} , the m-th order backward prediction error b _{m, i} , and the m-th order average energy A _m Is stored in the RAM (15) [m−1 linear prediction coefficients γ _{m, k} (k = 1, 2,..., M−1) are also stored as necessary. ].

  Subsequent to the process of step S8 (the process of step S9 when the process of step S9 is executed as necessary), the same process as step S3a is performed. That is, the control unit (190) determines m = M (step S3b). If m ≠ M, the control unit (190) performs the process of expansion from the m-th order to the m + 1-order [actually, it is none other than the expansion from the (m-1) th order to the m-th order described above. ] To perform. If m = M, the control unit (190) controls to end the PARCOR coefficient calculation process.

When m = M holds, the RAM (15) stores the M-th order PARCOR coefficient γ _{M, M} , the M-th order forward prediction error b ′ _{M, i} , and the M-th order backward prediction error b. _{M, i} , M-th order average energy _AM is stored [if necessary, M−1 linear prediction coefficients γ _{M, k} (k = 1, 2,..., M−1) are also stored. Has been. ]. The PARCOR coefficients γ _{M and M} are stored and stored in the external storage device (17) under the control of the control unit (190).

<Modification of First Embodiment>
In the first embodiment, the average energy _Am is calculated according to the equation (12). Meanwhile, the process of deriving the equation (12), i.e., equation (10), as is clear from equation (11), the m-th order of the average energy A _m, can be obtained as follows. That is, the m-th order forward prediction error energy _{C m} approximate calculated from forward prediction error energy _{C m-1} and PARCOR coefficients in m-1 order as the formula (12), and the m-th order forward prediction error energy _{C m} The m-th order average energy _Am is calculated as the average energy with the m-th order backward prediction error energy. Specifically, the average energy _Am is calculated by the equation (14).

At this time, the m-th order PARCOR coefficient is given by Equation (15).

Of course, the m-th order backward prediction error energy is approximately calculated from the m-1st-order backward prediction error energy and the PARCOR coefficient as shown in Expression (12), and the m-th order backward prediction error energy and the m-th order forward prediction error energy it may calculate the m-th order of the average energy a _m as the average energy.

In the modification of the first embodiment, the process of step S7 of the first embodiment is changed as follows (see FIGS. 6 and 7).
That is, the average energy calculation unit (145) outputs the m−1 order forward prediction error energy C _m−1 , the m order backward prediction error coefficient b _{m, i} , and the m−1 order PARCOR coefficient γ from the RAM (15). Load the _{m-1, m-1,} respectively, to calculate the average energy _{a m} and m-order forward prediction error energy _{C m} of m-th order by performing the operation of equation (14), the average energy _{a m} and forward prediction error energy storing C _m in a predetermined storage area of the RAM (15) (step S7S).

Hereinafter, the second embodiment and the third embodiment will be described. Each is an improved form of the first embodiment or a modification of the first embodiment, but will be described as an improvement of the first embodiment for convenience of explanation.
In the second and third embodiments, the PARCOR coefficient is calculated based on the conventional Burg method without performing the approximation as in the present invention, depending on whether or not a preset condition (determination condition) is satisfied. It is a form to introduce a case.

<< Second Embodiment >>
A second embodiment of the present invention will be described with reference to the drawings.
<Outline of Second Embodiment>
In the first embodiment, the average energy _Am is calculated using the approximation of Expression (12). However, when the order is low, an approximation error becomes a problem, that is, a linear prediction model with low accuracy may be generated. Therefore, in the second embodiment, a certain order Ma or higher (for example, Ma is a value of about 4 to 10). And The average energy _Am is calculated using the approximation of equation (12). That is, below a certain order Ma, the PARCOR coefficient is calculated by the conventional Burg method.

<PARCOR coefficient calculation apparatus of the second embodiment>
The PARCOR coefficient calculation apparatus according to the second embodiment has the same hardware configuration as the PARCOR coefficient calculation apparatus (1) according to the first embodiment, and only different parts from the first embodiment will be described.
In the second embodiment, in addition to the program of the first embodiment, the error-dependent average energy B _m in the conventional Burg method is added to the external storage device (17) [see formula (16)]. ] Is also stored and stored. In addition, the control program can determine whether or not the current order is equal to a certain order Ma-1, and can control processing based on each program based on the determination result. The order Ma is stored and stored in the external storage device (17). The initial value B ₁ of the error-dependent average energy B _m given by the equation (16) may be given using the proviso of the equations (7) and (8).

  In the PARCOR coefficient calculation device (1) of the second embodiment, each program stored in the external storage device (17) and data necessary for the processing of each program are read into the RAM (15) as necessary, Interpretation is executed and processed by the CPU (14). As a result, the CPU (14) performs predetermined functions (initial value setting unit, PARCOR coefficient calculation unit, forward prediction error update unit, backward prediction error update unit, average energy calculation unit, error dependent average energy calculation unit, control unit). By implementing this, the calculation of the PARCOR coefficient is realized. The control unit also functions as a determination unit that determines whether or not the current order is equal to a certain order Ma-1.

<PARCOR coefficient calculation process of the second embodiment>
Next, with reference to FIGS. 8 and 9, the flow of the PARCOR coefficient calculation process in the second embodiment will be described descriptively. Here, a different part from the flow of the PARCOR coefficient calculation process in the first embodiment will be described.

In the PARCOR coefficient calculation process in the second embodiment, the process of the next step S10a is performed when m ≠ M in the process of step S3a in the first embodiment.
That is, the control unit (190) determines whether m = Ma-1 is successful (step S10a). If m = Ma-1, the control unit (190) performs control so as to perform the processes after step S4 described in the first embodiment [steps S4 to S8, S3b, and each process of S9 as necessary]. To do. If m ≠ Ma−1, the control unit (190) performs control so as to perform processing from step S4r described later.

  In the process of step S10a, if m ≠ Ma-1, the process of step S4r is performed. The process in step S4r is the same as the process in step S4 described in the first embodiment.

  Subsequently, the process of step S5r is performed. The process in step S5r is the same as the process in step S5 described in the first embodiment.

  Subsequently, the process of step S6r is performed. The process in step S6r is the same as the process in step S6 described in the first embodiment.

Subsequently, the process of step S7r is performed. That is, the error-dependent average energy calculation unit (148) reads each of b _{m, i} , b ′ _{m, i} from the RAM (15), performs the calculation of Expression (16), and performs the m-th order error-dependent average energy B _m. It is calculated and stored in the error-dependent average energy _{B m} in a predetermined storage area of the RAM (15) (step S7R).

Next, the PARCOR coefficient calculation unit (141) reads b ′ _{m, i} , b _{m, i} , and B _m from the RAM (15), performs the calculation of Expression (9), and performs the m-th order PARCOR coefficient γ _{m. ,} M are calculated, and the PARCOR coefficient γ _{m, m} is stored in a predetermined storage area of the RAM (15) (step S8r).

  Next, the process of step S9r is performed as necessary. The process in step S9r is the same as the process in step S9 described in the first embodiment.

  Subsequent to the process of step S8r (the process of step S9r when the process of step S9r is executed as necessary), the process of step S3r is performed. The process in step S3r is the same as the process in step S3a or step S3b described in the first embodiment. Specifically, the control unit (190) determines m = M (step S3r). If m ≠ M, the control unit (190) controls to perform the process of step S10b described later. If m = M, the control unit (190) controls to end the PARCOR coefficient calculation process.

If m ≠ M in the process of step S3r, the control unit (190) determines whether m = Ma−1 is successful (step S10b). If m ≠ Ma−1, the control unit (190) performs control so as to perform the processes after step S4r described above (steps S4r to S8r, S3r, S10b, and each process of S9r as necessary). If m = Ma-1, the control unit (190) performs control so as to perform the processes after step S4 described in the first embodiment [steps S4 to S8, S3b, and each process of S9 as necessary]. To do.
In addition, after it determines with m = Ma-1 by the process of step S10b, the average energy _Am-1 used by the process of step S7 when m = Ma is set by the process of step S4 is the latest step. This is the error-dependent average energy B _m−1 calculated by the process of S7r. Further, PARCOR coefficients γ _{m-1, m-1} to be used in the processing of the step S7 is the PARCOR coefficient γ _{m-1, m-1} calculated in the process of immediately preceding step S8r.

<Another form of the second embodiment>
In the second embodiment, the processes in steps S4r, S5r, S6r, and S10b are the same as the processes in steps S4, S5, S6, and S10a, respectively. Therefore, another form of the second embodiment is a PARCOR coefficient calculation process as follows. It can be the flow of. Here, with reference to FIG. 10, a different part from the flow of the PARCOR coefficient calculation process in the first embodiment will be described.

In the PARCOR coefficient calculation process in another form of the second embodiment, the process of the next step S10c is performed following the process of step S6 in the first embodiment.
That is, the control unit (190) determines whether m = Ma is successful (step S10c). The reason for determining whether m = Ma is different from the second embodiment is that the process of step S10c is performed after the process of step S4. If m = Ma, the control unit (190) performs control so as to perform the processing after Step S7 described in the first embodiment [Steps S7, S8, S3b, and each processing of S9 as necessary]. If m ≠ Ma, the control unit (190) performs control so as to perform the processing after step S7r described above (steps S7r, S8r, S3r, and each processing of S9 as necessary).
In the second embodiment, the process of step S10b is performed when m ≠ M in the process of step S3r. However, in another form of the second embodiment, the process of step S4 is performed when m ≠ M in the process of step S3r. The subsequent processing is performed (see FIG. 10).

<< Third Embodiment >>
A third embodiment of the present invention will be described with reference to the drawings.
<Outline of Third Embodiment>
In the first embodiment, the average energy _Am is calculated using the approximation of Expression (12). On the other hand, from the viewpoint of preventing accumulation of approximate errors, in the third embodiment, once every h times [for example, h is a value of about 4 to 10. The PARCOR coefficient is calculated by the conventional Burg method.

<PARCOR coefficient calculation apparatus according to the third embodiment>
The PARCOR coefficient calculation apparatus according to the third embodiment has the same hardware configuration as the PARCOR coefficient calculation apparatus (1) according to the second embodiment, and only different parts from the second embodiment will be described.
In the third embodiment, instead of the control program of the second embodiment, the control program determines whether or not the remainder obtained by dividing the current order m by h is equal to 0. It is possible to control the processing based on it. The value h is stored and stored in the external storage device (17).

  In the PARCOR coefficient calculation device (1) of the third embodiment, each program stored in the external storage device (17) and data necessary for the processing of each program are read into the RAM (15) as necessary, Interpretation is executed and processed by the CPU (14). As a result, the CPU (14) performs predetermined functions (initial value setting unit, PARCOR coefficient calculation unit, forward prediction error update unit, backward prediction error update unit, average energy calculation unit, error dependent average energy calculation unit, control unit). By implementing this, the calculation of the PARCOR coefficient is realized. The control unit also functions as a determination unit that determines whether or not the remainder obtained by dividing the current order m by h is equal to zero.

<PARCOR coefficient calculation process of the third embodiment>
Next, with reference to FIG. 11, the flow of PARCOR coefficient calculation processing in the third embodiment will be described descriptively. Here, a different part from the flow of the PARCOR coefficient calculation process in the first embodiment will be described.

In the PARCOR coefficient calculation process in the third embodiment, the process of the next step S11 is performed when m ≠ M in the process of step S3a in the first embodiment.
That is, the control unit (190) determines whether or not the remainder obtained by dividing the current order m by h, that is, m mod h is equal to 0 (step S11). If m mod h ≠ 0, the control unit (190) performs control so as to perform the processing after Step S4 described in the first embodiment [Steps S4 to S8, S3b, and each processing of S9 as necessary]. To do. If m mod h = 0, the control unit (190) performs control so as to perform the processing after Step S4r described in the second embodiment (Steps S4r to S8r, S3r, and if necessary, each processing of S9r). To do.

In the second embodiment, the process of step S10b is performed when m ≠ M in the process of step S3r. However, in the third embodiment, the control unit (190) performs the process of m ≠ M in the process of step S3r. Then, control is performed so as to perform the processing after step S11 (see FIG. 11).
Note that, after it is determined in step S3r that m ≠ M, the average energy A _m−1 used in the first step S7 is the error-dependent average energy B _m− calculated in the latest step S7r. ₁ . Further, PARCOR coefficients γ _{m-1, m-1} to be used in the processing of the step S7 is the PARCOR coefficient γ _{m-1, m-1} calculated in the process of immediately preceding step S8r.

<Another form of the third embodiment>
In the third embodiment, the processes of steps S4r, S5r, and S6r are the same as the processes of steps S4, S5, and S6, respectively. be able to. Here, with reference to FIG. 12, a different part from the flow of the PARCOR coefficient calculation process in 2nd Embodiment is demonstrated.

In the PARCOR coefficient calculation process in another form of the third embodiment, the process of step S11 described above is performed following the process of step S6 in the first embodiment.
That is, the control unit (190) determines whether m mod h = 0 is successful (step S11). If m mod h ≠ 0, the control unit (190) performs control so as to perform the processes after step S7 described in the first embodiment (steps S7, S8, S3b, and each process of S9 as necessary). To do. If m mod h = 0, the control unit (190) performs control so as to perform the processes after step S7r described above (steps S7r, S8r, S3r, and if necessary, each process of S9r).
In the third embodiment, the process of step S11 is performed when m ≠ M in the process of step S3r. However, in another form of the third embodiment, the process of step S4 is performed when m ≠ M in the process of step S3r. The subsequent processing is performed (see FIG. 12).

  In the third embodiment, the process of step S11 is a process of determining whether m mod h is equal to 0, but is not limited to such a process. For example, it is determined whether or not the current order is an order set arbitrarily in advance (a plurality of orders can be set). If the order is a set order, the processing after step S4r can be performed. From this point of view, the process of determining whether m mod h is equal to 0 is controlled so that the PARCOR coefficient is calculated by the conventional Burg method when the model order is a multiple of h. ing.

In addition to the above embodiments, the PARCOR coefficient calculation apparatus and method according to the present invention are not limited to the above-described embodiments, and can be appropriately changed without departing from the spirit of the present invention.
Further, the processing described in the above PARCOR coefficient calculation device / method is not only executed in time series in the order described, but also executed in parallel or individually as required by the processing capability of the device that executes the processing. It may be.

  When the processing functions in the PARCOR coefficient calculation device are realized by a computer, the processing contents of the functions that the PARCOR coefficient calculation device should have are described by a program. By executing this program on a computer, the processing functions in the PARCOR coefficient calculation device are realized on the computer.

  The program describing the processing contents can be recorded on a computer-readable recording medium. As the computer-readable recording medium, for example, any recording medium such as a magnetic recording device, an optical disk, a magneto-optical recording medium, and a semiconductor memory may be used. Specifically, for example, as a magnetic recording device, a hard disk device, a flexible disk, a magnetic tape or the like, and as an optical disk, a DVD (Digital Versatile Disc), a DVD-RAM (Random Access Memory), a CD-ROM (Compact Disc Read Only). Memory), CD-R (Recordable) / RW (ReWritable), etc., magneto-optical recording medium, MO (Magneto-Optical disc), etc., semiconductor memory, EEP-ROM (Electronically Erasable and Programmable-Read Only Memory), etc. Can be used.

  The program is distributed by selling, transferring, or lending a portable recording medium such as a DVD or CD-ROM in which the program is recorded. Furthermore, the program may be distributed by storing the program in a storage device of the server computer and transferring the program from the server computer to another computer via a network.

  A computer that executes such a program first stores, for example, a program recorded on a portable recording medium or a program transferred from a server computer in its own storage device. When executing the process, this computer reads the program stored in its own recording medium and executes the process according to the read program. As another execution form of the program, the computer may directly read the program from a portable recording medium and execute processing according to the program, and the program is transferred from the server computer to the computer. Each time, the processing according to the received program may be executed sequentially. Also, the program is not transferred from the server computer to the computer, and the above-described processing is executed by a so-called ASP (Application Service Provider) type service that realizes the processing function only by the execution instruction and result acquisition. It is good. Note that the program in this embodiment includes information that is used for processing by an electronic computer and that conforms to the program (data that is not a direct command to the computer but has a property that defines the processing of the computer).

  In this embodiment, the PARCOR coefficient calculation device is configured by executing a predetermined program on a computer. However, at least a part of these processing contents may be realized by hardware.

  The present invention is useful for signal analysis and signal encoding.

The figure which shows the hardware structural example of the PARCOR coefficient calculation apparatus concerning 1st Embodiment. The block diagram which shows the function structural example of the PARCOR coefficient calculation apparatus concerning 1st Embodiment (equivalent to the process in the case of order 1). The block diagram which shows the function structural example of the PARCOR coefficient calculation apparatus concerning 1st Embodiment (equivalent to the process in the case of extending from degree 1 to 2). The block diagram which shows the function structural example of the PARCOR coefficient calculation apparatus concerning 1st Embodiment (equivalent to the process in the case of extending from order m-1 to m). The figure which shows the processing flow of the PARCOR coefficient calculation process concerning 1st Embodiment. The block diagram which shows the function structural example of the PARCOR coefficient calculation apparatus concerning the modification of 1st Embodiment. The figure which shows the processing flow of the PARCOR coefficient calculation process concerning the modification of 1st Embodiment. The block diagram which shows the function structural example of the PARCOR coefficient calculation apparatus concerning 2nd Embodiment. The figure which shows the processing flow of the PARCOR coefficient calculation process concerning 2nd Embodiment. The figure which shows the processing flow of the PARCOR coefficient calculation process concerning another form of 2nd Embodiment. The figure which shows the processing flow of the PARCOR coefficient calculation process concerning 3rd Embodiment. The figure which shows the processing flow of the PARCOR coefficient calculation process concerning another form of 3rd Embodiment.

Explanation of symbols

1 PARCOR coefficient calculation device 141 PARCOR coefficient calculation unit 142 forward prediction error update unit 143 backward prediction error update unit 145 average energy calculation unit 146 linear prediction coefficient calculation unit 148 error-dependent average energy calculation unit 149 forward prediction error energy calculation unit 190 control unit

Claims (14)

The forward prediction error b ′ _m−1 when the model order of the linear prediction model (hereinafter simply referred to as “model order”) is _m−1 , the backward prediction error b _m−1 and the model order m−1 when the model order is m−1. Forward prediction error updating means for calculating a forward prediction error b ′ _m in the model order m from the PARCOR coefficients γ _{m−1, m−1 in} FIG.
From the forward prediction error b ′ _{m−1 in} the model order m−1, the backward prediction error b _{m−1 in the} model order m−1, and the PARCOR coefficient γ _{m−1, m−1 in the} model order m ₋₁ to the model order m A backward prediction error updating means for calculating the backward prediction error b _m ;
A PARCOR coefficient calculating device comprising PARCOR coefficient calculating means for calculating a PARCOR coefficient γ _{m, m} in a model order m using a forward prediction error b ′ _m and a backward prediction error b _m in the model order m,
The average energy _{A m-1} in the model order m-1, the average energy A in model order m the square of the PARCOR coefficient gamma _{m-1, m-1} to definitive model orders m-1 by multiplying a value obtained by subtracting from 1 an average energy calculating means for calculating _m ,
The PARCOR coefficient calculating means includes:
The PARCOR coefficient γ _{m, m} in the model order _m is calculated by dividing the inner product of the forward prediction error b ′ _m and the backward prediction error b _m in the model order m by the average energy A _m in the model order m. PARCOR coefficient calculation device. The forward prediction error b ′ _m−1 when the model order of the linear prediction model (hereinafter simply referred to as “model order”) is _m−1 , the backward prediction error b _m−1 and the model order m−1 when the model order is m−1. Forward prediction error updating means for calculating a forward prediction error b ′ _m in the model order m from the PARCOR coefficients γ _{m−1, m−1 in} FIG.
From the forward prediction error b ′ _{m−1 in} the model order m−1, the backward prediction error b _{m−1 in the} model order m−1, and the PARCOR coefficient γ _{m−1, m−1 in the} model order m ₋₁ to the model order m A backward prediction error updating means for calculating the backward prediction error b _m ;
PARCOR coefficient calculating means for calculating PARCOR coefficient γ _{m, m} in model order m using forward prediction error b ′ _m and backward prediction error b _m in model order m;
A PARCOR coefficient calculation device comprising:
The prediction error energy _{C m-1} in the model order m-1, which corresponds to one of the energy of the forward prediction error or backward prediction error, PARCOR coefficient gamma _m-1 definitive model orders _m-1, the square of the _m-1 the calculated prediction error energy C _m in model order m by multiplying the value obtained by subtracting from 1, further an average of the energy in the model order m of the prediction error energy C _m and the other of the prediction error in the model order m with an average energy calculating means for calculating an average energy a _m in a certain model order m,
The PARCOR coefficient calculating means includes:
The PARCOR coefficient γ _{m, m} in the model order _m is calculated by dividing the inner product of the forward prediction error b ′ _m and the backward prediction error b _m in the model order m by the average energy A _m in the model order m.
P ARCOR coefficient calculation device. The forward prediction error b ′ _m−1 when the model order of the linear prediction model (hereinafter simply referred to as “model order”) is _m−1 , the backward prediction error b _m−1 and the model order m−1 when the model order is m−1. Forward prediction error updating means for calculating a forward prediction error b ′ _m in the model order m from the PARCOR coefficients γ _{m−1, m−1 in} FIG.
From the forward prediction error b ′ _{m−1 in} the model order m−1, the backward prediction error b _{m−1 in the} model order m−1, and the PARCOR coefficient γ _{m−1, m−1 in the} model order m ₋₁ to the model order m A backward prediction error updating means for calculating the backward prediction error b _m ;
A PARCOR coefficient calculating device comprising PARCOR coefficient calculating means for calculating a PARCOR coefficient γ _{m, m} in a model order m using a forward prediction error b ′ _m and a backward prediction error b _m in the model order m,
Determination means for determining whether or not a preset condition (hereinafter referred to as “determination condition”) is satisfied;
The average energy _{A m-1} in the model order m-1, the average energy A in model order m the square of the PARCOR coefficient gamma _{m-1, m-1} to definitive model orders m-1 by multiplying a value obtained by subtracting from 1 mean energy calculating means for calculating _m ,
The average of the energy of the forward prediction error b _'m and backward prediction error b _m in the model order m (hereinafter, referred to as "error-dependent mean energy B _m".) And a error-dependent average energy calculating means for calculating,
The PARCOR coefficient calculating means includes:
By dividing the inner product of the forward prediction error b ′ _m and the backward prediction error b _{m in} the model order m by the average energy _Am or the error-dependent average energy B _m in the model order m according to the determination result of the determination means. A PARCOR coefficient calculation device capable of calculating a PARCOR coefficient γ _m, m in a model order m. The forward prediction error b ′ _m−1 when the model order of the linear prediction model (hereinafter simply referred to as “model order”) is _m−1 , the backward prediction error b _m−1 and the model order m−1 when the model order is m−1. Forward prediction error updating means for calculating a forward prediction error b ′ _m in the model order m from the PARCOR coefficients γ _{m−1, m−1 in} FIG.
From the forward prediction error b ′ _{m−1 in} the model order m−1, the backward prediction error b _{m−1 in the} model order m−1, and the PARCOR coefficient γ _{m−1, m−1 in the} model order m ₋₁ to the model order m A backward prediction error updating means for calculating the backward prediction error b _m ;
PARCOR coefficient calculating means for calculating PARCOR coefficient γ _{m, m} in model order m using forward prediction error b ′ _m and backward prediction error b _m in model order m;
A PARCOR coefficient calculation device comprising:
Determination means for determining whether or not a preset condition (hereinafter referred to as “determination condition”) is satisfied;
The prediction error energy _{C m-1} in the model order m-1, which corresponds to one of the energy of the forward prediction error or backward prediction error, PARCOR coefficient gamma _m-1 definitive model orders _m-1, the square of the _m-1 the calculated prediction error energy C _m in model order m by multiplying the value obtained by subtracting from 1, further an average of the energy in the model order m of the prediction error energy C _m and the other of the prediction error in the model order m and average energy calculating means for calculating an average energy a _m in a certain model order m,
The average of the energy of the forward prediction error b _'m and backward prediction error b _m in the model order m (hereinafter referred to as "error-dependent mean energy B _m".) And a error-dependent average energy calculating means for calculating,
The PARCOR coefficient calculating means includes:
By dividing the inner product of the forward prediction error b ′ _m and the backward prediction error b _{m in} the model order m by the average energy _Am or the error-dependent average energy B _m in the model order m according to the determination result of the determination means . It is possible to calculate the PARCOR coefficient γ _{m, m} at the model order m
P ARCOR coefficient calculation device. The above judgment condition is
The model order is a preset order,
The PARCOR coefficient calculating means is
The inner product of the forward prediction error b ′ _m and the backward prediction error b _m in the model order m is the error-dependent average energy B _m in the model order m when the model order m satisfies the above determination condition, and the model order m is the above PARCOR coefficient calculating apparatus according to claim 3 or claim 4 calculates a PARCOR coefficient gamma _{m, m} in the model order m by dividing the average energy _{a m} in the model order m if not satisfy the judgment conditions. The above "preset order" is
The PARCOR coefficient calculation device according to claim 5, wherein the order is equal to or less than a predetermined threshold or smaller than the threshold. A PARCOR coefficient calculation method for calculating a PARCOR coefficient of a linear prediction model from a recurrence relationship for obtaining a PARCOR coefficient in a model order m from a PARCOR coefficient in a model order m-1,
The forward prediction error update unit includes a forward prediction error b ′ _{m−1 in} the model order m−1, a backward prediction error b _{m−1 in} the model order m−1, and a PARCOR coefficient γ _{m−1, m} in the model order m−1. a forward prediction error update step of calculating a forward prediction error b _'m in the model order m _-1,
The backward prediction error update unit includes a forward prediction error b ′ _{m−1 in} the model order m−1, a backward prediction error b _{m−1 in} the model order m−1, and a PARCOR coefficient γ _{m−1, m} in the model order m−1. a backward prediction error update step of calculating the backward prediction error b _m in the model order m _-1,
Model by average energy calculating unit, the average energy _{A m-1} in the model order m-1, multiplied by a value obtained by subtracting the square of the PARCOR coefficient gamma _{m-1, m-1} to definitive model orders m-1 from 1 the average energy calculating step of calculating an average energy a _m in order m,
The PARCOR coefficient calculation unit divides the inner product of the forward prediction error b ′ _m and the backward prediction error b _m in the model order m by the average energy A _m in the model order m, whereby the PARCOR coefficient γ _{m, m} in the model order _m A PARCOR coefficient calculation method comprising: calculating a PARCOR coefficient. A PARCOR coefficient calculation method for calculating a PARCOR coefficient of a linear prediction model from a recurrence relationship for obtaining a PARCOR coefficient in a model order m from a PARCOR coefficient in a model order m-1,
The forward prediction error update unit includes a forward prediction error b ′ _{m−1 in} the model order m−1, a backward prediction error b _{m−1 in} the model order m−1, and a PARCOR coefficient γ _{m−1, m} in the model order m−1. a forward prediction error update step of calculating a forward prediction error b _'m in the model order m _-1,
The backward prediction error update unit includes a forward prediction error b ′ _{m−1 in} the model order m−1, a backward prediction error b _{m−1 in} the model order m−1, and a PARCOR coefficient γ _{m−1, m} in the model order m−1. a backward prediction error update step of calculating the backward prediction error b _m in the model order m _-1,
The average energy calculating unit, the prediction error energy _{C m-1} in the model order m-1, which corresponds to one of the energy of the forward prediction error or backward prediction error, PARCOR coefficient gamma _m-1 which definitive model orders m-1 _, the square of the _m-1 to calculate a prediction error energy C _m in model order m by multiplying the value obtained by subtracting from 1, further prediction in model order m error energy C _m and model order m of the other prediction error the average energy calculating step of calculating an average energy a _m in the model order m is the average of the energy in,
The PARCOR coefficient calculation unit divides the inner product of the forward prediction error b ′ _m and the backward prediction error b _m in the model order m by the average energy A _m in the model order m, whereby the PARCOR coefficient γ _{m, m} in the model order _m PARCOR coefficient calculation step for calculating
A PARCOR coefficient calculation method having : A PARCOR coefficient calculation method for calculating a PARCOR coefficient of a linear prediction model from a recurrence relationship for obtaining a PARCOR coefficient in a model order m from a PARCOR coefficient in a model order m-1,
The forward prediction error update unit includes a forward prediction error b ′ _{m−1 in} the model order m−1, a backward prediction error b _{m−1 in the} model order m−1, and a PARCOR coefficient γ _{m−1, m} in the model order m−1. a forward prediction error update step of calculating a forward prediction error b _'m in the model order m _-1,
The backward prediction error update unit includes a forward prediction error b ′ _{m−1 in} the model order m−1, a backward prediction error b _{m−1 in the} model order m−1, and a PARCOR coefficient γ _{m−1, m} in the model order m−1. a backward prediction error update step of calculating the backward prediction error b _m in the model order m _-1,
A determination step for determining whether or not the determination unit satisfies a preset condition (hereinafter referred to as “determination condition”);
When it is determined that the determination condition is satisfied in the determination step,
An error-dependent average energy calculation unit calculates an average of each energy of the forward prediction error b ′ _m and the backward prediction error b _{m in} the model order m (hereinafter referred to as “error-dependent average energy B _m ”). A calculation step;
The PARCOR coefficient calculation unit divides the inner product of the forward prediction error b ′ _m and the backward prediction error b _m in the model order m by the error-dependent average energy B _m in the model order m, thereby generating a PARCOR coefficient γ _m in the model order m. _, M, a first PARCOR coefficient calculation step;
When it is determined that the determination condition is not satisfied in the determination step,
Model by average energy calculating unit, the average energy _{A m-1} in the model order m-1, multiplied by a value obtained by subtracting the square of the PARCOR coefficient gamma _{m-1, m-1} to definitive model orders m-1 from 1 the average energy calculating step of calculating an average energy a _m in order m,
The PARCOR coefficient calculation unit divides the inner product of the forward prediction error b ′ _m and the backward prediction error b _m in the model order m by the average energy A _m in the model order m, whereby the PARCOR coefficient γ _m, in the model order _m. a second PARCOR coefficient calculating step for calculating _m ;
PARCOR coefficient calculation method have. A PARCOR coefficient calculation method for calculating a PARCOR coefficient of a linear prediction model from a recurrence relationship for obtaining a PARCOR coefficient in a model order m from a PARCOR coefficient in a model order m-1,
The forward prediction error update unit includes a forward prediction error b ′ _{m−1 in} the model order m−1, a backward prediction error b _{m−1 in the} model order m−1, and a PARCOR coefficient γ _{m−1, m} in the model order m−1. a forward prediction error update step of calculating a forward prediction error b _'m in the model order m _-1,
The backward prediction error update unit includes a forward prediction error b ′ _{m−1 in} the model order m−1, a backward prediction error b _{m−1 in the} model order m−1, and a PARCOR coefficient γ _{m−1, m} in the model order m−1. a backward prediction error update step of calculating the backward prediction error b _m in the model order m _-1,
A determination step for determining whether or not the determination unit satisfies a preset condition (hereinafter referred to as “determination condition”);
When it is determined that the determination condition is satisfied in the determination step,
An error-dependent average energy calculation unit calculates an average of each energy of the forward prediction error b ′ _m and the backward prediction error b _{m in} the model order m (hereinafter referred to as “error-dependent average energy B _m ”). A calculation step;
The PARCOR coefficient calculation unit divides the inner product of the forward prediction error b ′ _m and the backward prediction error b _m in the model order m by the error-dependent average energy B _m in the model order m, thereby generating a PARCOR coefficient γ _m in the model order m. _, M, a first PARCOR coefficient calculation step;
When it is determined that the determination condition is not satisfied in the determination step,
The average energy calculating unit, the prediction error energy _{C m-1} in the model order m-1, which corresponds to one of the energy of the forward prediction error or backward prediction error, PARCOR coefficient gamma _m-1 which definitive model orders m-1 _, the square of the _m-1 to calculate a prediction error energy C _m in model order m by multiplying the value obtained by subtracting from 1, further prediction in model order m error energy C _m and model order m of the other prediction error the average energy calculating step of calculating an average energy a _m in the model order m is the average of the energy in,
The PARCOR coefficient calculation unit divides the inner product of the forward prediction error b ′ _m and the backward prediction error b _m in the model order m by the average energy A _m in the model order m, whereby the PARCOR coefficient γ _m, in the model order _m. a second PARCOR coefficient calculating step for calculating _m ;
A PARCOR coefficient calculation method having : The above judgment conditions are:
The PARCOR coefficient calculation method according to claim 9 or 10 , wherein the model order is a preset order. The above "preset order" is
The PARCOR coefficient calculation method according to claim 11 , wherein the order is equal to or less than a predetermined threshold value or smaller than the threshold value. A PARCOR coefficient calculation program for causing a computer to function as the PARCOR coefficient calculation device according to any one of claims 1 to 6 . A computer-readable recording medium on which the PARCOR coefficient calculation program according to claim 13 is recorded.

Images