JPH02230297A - Period detecting method - Google Patents

Abstract

PURPOSE:To securely detect a specific period in a short time by finding a 1st peak within a specific range, detecting a 2nd peak corresponding to the 1st peak by a search from a detection start point corresponding to the level of the 1st peak, and deciding whether or not a period determined based upon the 1st and 2nd peaks is appropriate. CONSTITUTION:The time t0 (when a MAX point is obtained within a specific range and a peak point is obtained is detected. A peak value is detected after a signal level drops below the half of the signal level at the time t0) and rises above the 1/2, and a next peak point WP1 is detected from a level variation state almost based upon the detected peak value, thereby estimating WP1-WP0 as its period temporarily. Another next peak is searched for within a range BW+DELTAB in accordance with the estimated period and the adequacy of the estimated period is checked in accordance with the time intervals of WP0, WP1, and WP2 to decide that the estimated period is correct at the time of (WP1-WP0) (WP2-WP1). Consequently, whether or no there is the periodicity and its period are known speedily and securely.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a method for detecting the period of a periodic signal portion included in an audio signal.

[Summary of the invention]

The present invention finds a first peak within a certain range of an audio signal waveform, and detects a second peak corresponding to the first peak by searching from a detection starting point according to the level of the first peak. , it is determined whether the period determined based on the first and second peaks is valid or not based on whether or not a predetermined peak value is obtained for each period, so that the predetermined II! This is designed to reliably detect JM. [Prior Art] For example, in order to encode an audio signal, it may be necessary to detect the period of a periodic part of the signal. In response to such requests, conventional methods have been used to obtain time data of the zero-crossing points of the audio signal and find the period for a periodic part from the time interval of the zero-crossing points indicated by this data. ing. [Invention tries to solve! II! I) However, according to this conventional method, in order to find regularity from the time interval of zero crossing points within a certain period, it is necessary to run a complicated program that includes many conditional judgment steps on a computer, which requires a large amount of time for detection. Since this method is time-consuming, it is not suitable for cases where it is desired to encode audio signals in real time with good responsiveness. Therefore, an object of the present invention is to provide an improved period detection method that can reliably detect the period of a periodic signal portion included in an audio signal using a relatively simple algorithm. be. (Means for Solving the Problem!!) A feature of the present invention for solving the above problem is that in a method for detecting a repetition period of a periodic part in an audio signal, A first step of detecting the time point t. at which the signal level is the maximum value and a local maximum value, and detecting the next peak value corresponding to the peak value according to the level of the peak value at 10, t1, and t2. The second step is to determine the level check starting point of t0, the third step is to detect the first peak after the starting point, and the first peak is determined to be the peak at t0 based on the level change in the vicinity of the first peak. A fourth step of determining whether the peak is a corresponding peak, and a fifth step of determining the time point t when the next peak corresponding to the peak at t. occurs by repeatedly performing the third and fourth steps. and t,−1. A sixth step of detecting at least one occurrence point 1. of another peak corresponding to the peak t0 within a range expected based on the values of t,−1.
jet F+ A seventh step of determining the period based on t0, t1, and t2. [Action J t in the first step. After is detected, in the fifth step, if it is detected that a peak considered to correspond to (. occurs at t1, t+-to is estimated to be a predetermined period. Therefore, t The time of occurrence of another peak corresponding to the peak of can be predicted with a certain degree of accuracy based on the value of 1.-t0.Therefore, based on the above idea, the time of occurrence of the peak can be predicted within a certain range, and the prediction The validity of whether 1. - 1. is a predetermined period is checked depending on whether a peak is detected in the sample. [Example] Hereinafter, an example of the present invention will be explained in detail. FIG. 1 shows an embodiment of a signal processing device 1 for detecting the period of a periodic part of an audio signal according to the method of the present invention. The audio signal S from the amplifier 3 is input to a period detector 4.The period detector 4 has a microphone 2 for converting the signal into It is configured as a microcomputer with a known configuration, and includes an I/O device 5, a CPU 6, and a ROM?.
The ROM consists of an R input M8 and a bus 9. A processing program for detecting the cycle is stored in advance in the internal memory, and the audio signal SP is processed according to the processing program to detect the cycle. FIG. 2 shows a flowchart showing this processing program. Below, while referring to the flowchart shown in Figure 2, the third
A method for detecting the period of the audio signal S shown in the figure according to the present invention will be explained. In step 11, the origin 0 on the time axis is set as the detection operation start point scan. Next, in step 12, a first process for detecting the maximum peak point PO within the 64 sample width is executed. The details of this first process are shown in the detailed flowchart of FIG. In step 31, 64
The sample point that gives the maximum value among the sample data (
time to) and its level L0 are detected, and the level L. A determination is made as to whether or not is greater than a predetermined level A. (Step 31). This level A is preferably selected to be about 2 times the peak level value of the audio signal S, and if the maximum value detected by this comparison is the maximum value of low-level noise, this can be effectively eliminated. Can be done. L. If ≦A, the flag F is cleared in step 33 and the process proceeds to the next step 13. If aLe>A, in step 34, each level of the two sample data following the sample data at time t0 is Level A
A determination is made as to whether the value is greater than or not. In this level check, at time t. If the sample data of is not pulse-like high level data, the levels of all three sample data should be greater than A. If the level check performed in this manner is OK, flag F is set in step 35, and the process proceeds to step 13. If the level check in step 34 is OK
Otherwise, the flag F is cleared in step 33, and then the process proceeds to step l3. The flag F indicates whether predetermined peak value data is included in the first 64 bits of sample data, and in the case of F-0, the determination result in step 13 is No, and it is detected in step 14. An impossible output signal is output from the period detector 4. When the first peak value is detected in step 12, F-1
Then, proceed to step l5. In step 15, the detection operation starting point scan for the next process is set to one P O (
= to) and set the loop counter value 100pnb to 0.
cent to. Thereafter, a second process is performed in step 16 to detect the next peak point 1 corresponding to the WPO. The second process will be described in detail with reference to FIG. 5. First, in step 41, the range of the audio signal level check is determined. In this example, scan
+16, that is, from the sampling point of WPO+16 to -PO
The check will be executed in the range up to the +106 sampling point. This check starting point can be determined as appropriate, but as described later, by determining it according to the level of the peak point (in this case WPO), it is possible to efficiently perform the specified check. becomes. In step 42, the level value of peak point ridge 0 <ypo>
At step 43, the point where the level of the audio signal decreases to the level of f1 is recognized, and furthermore, in step 43, the point where the level of the audio signal decreases to the level of (
Check the point at which the level rises to the level of A of HP O). In the method of the present invention, the level (
Based on the assumption that a peak corresponding to the peak at the WPO point does not occur in the above-mentioned range determined according to HP O'), that is, from the -PO point until the audio signal level reaches (WPo)/2 again, After the confirmation in step 43 is completed, processing for detecting a predetermined peak is started in step 44. That is, in order to detect the next peak WPI based on the level of each sampling value from time 1 onward in FIG. 3, first, the level Lx of the peak point X that occurs for the first time after tl is detected. In step 45, a check is performed to see whether this peak point X is the next expected peak point furrow 1 or not. This chaesok is the 6th
As shown in the figure, for the level Lx of the peak point X,
Multiple (for example, three) consecutive sampling points P
r, Pt, level LPI at P,. LP! ．．
Detect LPS. And these levels LP+,
LPx. If all LPs are smaller than the level Lχ, the peak point X is determined to be a predetermined local maximum point. In the example shown in FIG. 6, Lx>LP, but Lx<LP
Since I and Lx<LPs, it is determined that the peak point X is not a peak point corresponding to the peak point PO. When the peak check in step 45 is OK,
Proceed to step 46 and proceed to the third peak check for the next peak check.
The limit values used in the process are set. This set of limit values is shown in the table below. The table shows the check range and WP after ridge 1 that should be performed in searching for the next peak value for period detection based on the WPI-14PoO value, that is, the period value detected here.
The step range showing the number of sampling points counted from I to be ignored is shown in the table above.The step range is determined based on the table above. Of course, the data shown in the table above is just an example, and the data values can be changed as appropriate without being limited to this, but tnpi-t+p
It is important that the decision is made based on the value of o. Next, in step 47, a flag B indicating whether or not the ridge 1 has been detected is set, and the process advances to step 17. If the result of the peak check in step 45 is not OK, the process proceeds to step 48, where it is determined whether or not the predetermined end point (WPO + 106) has been exceeded for the check, and whether or not the peak check has yet to be completed is determined. If the peak point has not been exceeded, the process returns to step 44 and the process for detecting the peak point described above is performed again. Step 44. If the predetermined peak point is not successfully detected even after repeating step 45, the determination result in step 48 becomes YES, flag B is cleared in step 41, and step 17
Proceed to. Returning to FIG. 2, in step 17 it is determined whether or not flag B is set, and in the case of B-0, it is determined in step 1 that the cycle could not be detected.
4, and the execution of this program ends. On the other hand, in the case of B-1, the process proceeds to step 18, and the start point of the check for the next cycle detection, that is, the next predetermined peak point detection, is set at -P1. After that,
Proceeding to step 19, the third process for detecting the next peak point is executed according to the flowchart shown in FIG. To explain the flowchart of FIG. 7, in step 51, the start and end points of the check for determining the next peak point ridge 2 are newly set. Here, as shown in Figure 3, the WPI-WPOO value is
Then, based on the search range ±ΔB set according to the size of BW, the starting point is BW-ΔB and the ending point is BW.
+ΔB is determined. In step 52, the end point BW+ΔB set in this way is
A check is made to see if it exceeds the limit value U set in , and if it is BW+ΔBTU, the end point is scanned +
If BW+ΔB≦, the process proceeds to step 54 without changing the end point σ. In step 54, the sampling point of the maximum level between the start point and end point determined in this way is detected, and that point is set as 1P2. In step 55,
It is checked whether the level of WP2 is equal to or higher than the level A of -P1, and if the determination result is No, it is determined that the detected -22 is false, and in step 56 a predetermined peak is detected. Clearing flag C will display the result that detection could not be performed. On the other hand, if the determination result in step 55 is YES, further in step 57 the level of IIP2 is set to WPI.
A determination is made as to whether the level is less than twice the level of . If the determination result in step 57 is No, it is determined that WP2 is inappropriate or false due to the level abnormality, and the process proceeds to step 56, where flag C is cleared. The determination result in step 57 is Y.
If it becomes ES, the process proceeds to step 58, where it is checked whether the point of 1P2 is within -PL+V. If the determination result in step 58 is NO, proceed to step 56, and if the determination result is YES, -P
A final determination is made that P2 is the desired peak, and flag C is set in step 59 to indicate that P2 has been properly detected. That is, in the third process, based on the data of WPO and WPI, it is assumed that WP2 should be obtained in a period that is almost the same as the period BW between them.
Set the check range within the range of ΔB (step 5
1), it is assumed that the point with the maximum level obtained there is 蛛2. To determine whether this assumption is correct, it is checked whether the level is within a predetermined range (for example, between 2 and 2 times the level of -P1) (steps 55 and 57), and ['2
points are confirmed.

Returning to FIG. 2, in step 20, a check is made to see if the flag is C-1, and if C=1, the next starting point is set to scan or -P2 in step 21, and step 1
9, the process for detecting the next predetermined peak point ridge 3 is executed in step 22. In step 23, it is determined whether or not WP3 has been properly detected. If WP3 has been properly detected, in step 24 these peak values - PO, WPI,
The signal period is calculated based on lIP2 and WP3, data indicating the result is output, and the program execution ends. If it is determined in either step 19 or 22 that the predetermined peak cannot be detected, the determination result in step 20 or 23 becomes NO, and the process proceeds to step 25. In Step 25, the level of WPI (Kan Pi) is set to one PO.
A comparison is made with the level of (one PO), and step 26
Now, if (WPI) > (WPO), the program advances to step 14 and it is assumed that the cycle cannot be detected. On the other hand, if (WPI)≦(WPO), step 27
Step 28 increments the value of 100pnt0 by 1, and in step 28 it is determined whether Ioopnb=4.
If 1oopnb≠4, return to step 16 and execute steps 16 and subsequent steps again. When executed again, the peak value could not be detected and Ioopnb-4
If so, proceed to step 14, declare that the period cannot be detected, and terminate the program execution. According to the above detection method, as explained based on FIG. 4, first, the time to (sampling point) which is the MAX point and the peak value is detected within a predetermined range. This time t0, t1, t2
Using the signal level 2 as a reference, the peak value is detected after the signal level becomes 2 or less and then becomes A or more again, and the level in the vicinity of this detected peak is detected by the level check in step 34. Based on the change state, it is determined whether this peak corresponds to t0 or not.

In this way, the next peak point furrow 1 is detected, and WPI-W
Let's estimate FO as its period. Based on this estimated period, search for the next peak within the range of BW + ΔB, and
P.O. The validity of the estimated period is checked by the time interval of WPI, WP2, and if it is valid, that is, (WP
I-WPO) 1 (1P2-WPI), the estimated period is determined to be correct. In this way, you can quickly and reliably know the origin and period of periodicity. [Effects of the Invention] According to the present invention, periodic parts occurring in an audio signal can be detected quickly and reliably using a simple algorithm as described above.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an example of the configuration of an apparatus for detecting a period in an audio signal using the method of the present invention, and FIG.
A flowchart showing a computer program for signal processing according to the present invention for period detection executed in the apparatus shown in the figure, FIG. 3 is a waveform diagram of a signal to be subjected to period detection, and FIGS. 6 is an explanatory diagram for explaining the signal processing in the second processing step, and FIG. 7 is a detailed flowchart of a part of the flowchart shown in the figure.
It is a detailed flowchart of the third processing step shown in the figure. 1 ・ ・ ・ ・ 4 ・ ・ ・ ・ S ・ ・ ・ ・ WPO, WPI, scan Rumor Φ Fist U ・ ・ ・ ・ ・Signal processing device・Period detector・Audio signal WP2...Peak point・Start point・Check Range■ ・ ・ ・More than step range

Claims (1)

[Claims] In a method for detecting a repetition period of a periodic part in an audio signal, the first step is to detect a time point t_0 at which the level of the audio signal reaches a maximum value and a local maximum value within a predetermined range. step and the time t_
A second step of determining a level check starting point for detecting the next peak value corresponding to the peak value according to the level of the peak value at 0, and a third step of detecting the first peak after the starting point. and a fourth step of detecting a level change in the vicinity of the first peak and determining whether or not this first peak corresponds to the peak of t_0, and repeating the third and fourth steps. A fifth step of determining the occurrence time t_1 of the next peak corresponding to the peak of t_0 by executing the step; Period detection in an audio signal, comprising: a sixth step for detecting at least one occurrence time t_2; and a seventh step for determining a period based on time points t_0, t_1, and t_2. Method.