JPH0744198A - Long-period predictive delay search method - Google Patents

Abstract

PURPOSE:To decrease the throughput requird for a search and the amount of information on long-period predictive delay to be sent by limiting the range of the search through the use of the correlation property of the long-period predictive delay and using past long-period predictive delay. CONSTITUTION:At ST1, an input signal X is inputted at every frame, and at ST2, long-period predictive delay plag of a previous frame is inputted. At ST3, the search range is limited to an N1+N2+N3 point nearby plag, 2*plag/3, plag/2, or plag*3 by a using the long-period predictive delay plag of the previous frame. Then synthetic filtering is peformed at ST4 for a signal a1g traced back by delay 1g in the search range from a long-period predictive filter state wherein a past driving signal is stored to generate a synthesized speech b1g. At ST5, the distance between the input signal X and synthesized speech b1g is calculated and 1g which minimize the distance d1g corresponding to delay in the search range is outputted as the ling-period predictive delay of the frame at ST6.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a long-term prediction delay searching method in a speech coding method for coding a digital signal such as a speech signal.

[0002]

2. Description of the Related Art A conventional long-term prediction filter delay searching method will be described with reference to FIG. FIG. 3 is a flowchart of a conventional long-term prediction delay search method.

As shown in FIG. 3, steps (hereinafter referred to as ST
11), the input signal x is input at a certain fixed time interval (frame, M samples), while the delay of 1 g from the long-term prediction filter state in which past drive signals are stored.
In ST12, the synthetic filtering is performed on the long-term predicted signal a _1g traced back by only, and synthetic speech b _1g is generated.
In ST13, the distance d between the input signal x and the synthesized voice b _1g
1g is calculated by the formula of (Equation 1) shown below,

[0004]

[Equation 1]

Distance d for delays in all ranges (N points)
_1g that minimizes _1g is set as the long-term prediction delay of this frame and is output in ST14.

[0006]

However, in the above-mentioned conventional long-term prediction delay search method, since the distance from the input signal is obtained for the delay of the entire range, there is a problem that the processing amount becomes very large. Was. In addition, there is a problem that a large amount of information of long-term prediction delay is transmitted.

The present invention solves the above-mentioned conventional problems, and provides an excellent long-term prediction delay search method capable of performing a long-term prediction delay search by significantly reducing the processing amount and further reducing the amount of information to be transmitted. With the goal.

[0008]

In order to achieve the above-mentioned object, the present invention utilizes the fact that the long-term prediction delay has correlation between frames, and accordingly, the search range is appropriately limited. The processing amount required and the information amount of the long-term prediction delay to be transmitted are reduced.

[0009]

Therefore, according to the present invention, by utilizing the fact that the long-term prediction delay has a correlation between frames, the search range is appropriately limited, so that the processing amount required for the search and the long-term prediction to be transmitted are transmitted. This has the effect of reducing the amount of delay information.

[0010]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to FIG. FIG. 1 is a flowchart of a long-term prediction delay search method according to an embodiment of the present invention.

First, in ST1, the input signal x is input for each frame, and in ST2, the long-term prediction delay plag of the previous frame is input. Next, in ST3, the search range is set as shown in FIG. 2 by using the long-term prediction delay plag of the previous frame as shown in FIG.
N ₁ + N ₂ + N around lag / 2 or plug * 3
Limited to ₃ points. Then, in ST4, synthesis filtering is performed on the signal a _1g that is traced back by the delay 1g within the search range from the long-term prediction filter state in which the past drive signals are stored, and synthetic speech b _1g is generated. In ST5, the distance between the input signal x and the synthesized voice b _1g is calculated by (Equation 1), and _{1g at} which the distance d _1g with respect to the delay within the search range is the minimum is the long-term predicted delay of this frame, and ST6
To output.

As described above, according to the present embodiment, in the search for a long-term prediction delay by the long-term prediction filter, the search range is searched from the N points in the entire range to N ₁ + N ₂ + N using the long-term prediction delay of the previous frame. By limiting to _three points, the amount of processing required for the search is (N ₁ + N ₂ + N ₃ ) / N, and the amount of information of the long-term predicted delay to be transmitted is log ₂ N / log ₂ (N ₁ + N ₂ + N ₃ ). Can be reduced.

[0013]

As described above, according to the present invention, in the search for a long-term prediction delay in a long-term prediction filter, the past long-term prediction delay is used by utilizing the correlation of the long-term prediction delay and the search range is expanded. The present invention provides an excellent long-term prediction delay search method capable of limiting the amount of processing required for search and the amount of information of long-term prediction delay to be transmitted.

[Brief description of drawings]

FIG. 1 is a flowchart of a long-term prediction delay search method according to an embodiment of the present invention.

FIG. 2 is an explanatory diagram of search range limitation in the embodiment of the present invention.

FIG. 3 is a flowchart of a conventional long-term prediction delay search method.

[Explanation of symbols]

 1 Input of signal x 2 Input of past long-term prediction delay 3 Setting of search range 4 Synthesis filtering process 5 Distance minimization process 6 Output of long-term prediction delay

Claims (2)

[Claims] 1. A long-term prediction delay search method using a long-term prediction filter, wherein a search range is limited by using past long-term prediction delays, and a processing amount required for the search and an information amount of a long-term prediction delay to be transmitted are set. A long-term prediction delay search method characterized by reduction. 2. The long-term prediction according to claim 1, wherein the search range of the long-term prediction delay is limited to the past long-term prediction delay and an area which is an integral multiple of the past long-term prediction delay and a multiple of 1 / integer. Lazy search method.