JPS62115500A - Fundamental cycle detector for voice signal - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

DETAILED DESCRIPTION OF THE INVENTION (a) Field of Industrial Application The present invention relates to a fundamental period detection circuit device for audio signals.

(b) Conventional technology The fundamental period of an audio signal can be easily derived as the reciprocal of the fundamental frequency of the audio signal. The present applicant has published Japanese Patent Application Laid-Open No. 58-30 in connection with circuit means for extracting the fundamental frequency of an audio signal.
No. 795 (GIOH1100) (filed date August 1982)
May 17th) and JP-A No. 59-220794 (GIO
HIloo) (filed on May 30, 1982).

In Japanese Patent Application Laid-Open No. 58-30795, the audio signal is converted to a predetermined frequency (
The sample interval time at which the autocorrelation function of the sampled main sequence becomes the maximum, or
The sample interval time when the difference function between specific sample values shows the minimum value is derived as the fundamental period of the audio signal.

In the example of JP-A-59-220794, JP-A-58-
In addition to the method for deriving the fundamental period of an audio signal disclosed in No. 30795, the audio signal is converted into a digital signal of "i" and "o" at the AC intersection point, and the autocorrelation is calculated for the sample value sequence sampled based on a predetermined clock. and derive the fundamental period.

In these conventional devices, in order to accurately derive the fundamental period of a voice signal, the a-voice signal (or the digital signal of this AC crossover point) is numbered at fine intervals.
It is necessary to perform calculations to derive the fundamental period, but in this case: (a) The sampling main sequence for calculating fA becomes large, and many means for storing it are required for calculation processing.

(b) As the sample value sequence becomes larger, the amount of calculation increases, and a high-speed calculation circuit is required.

It is convenient to use a microcomputer to store these calculations and the sample values for the calculations, but in this case, if a large number of storage means such as RAM are used, the circuit configuration becomes large, and the In order to complete the calculation within a certain amount of time, it is necessary to keep the sample values within a predetermined number, which has the disadvantage that the detection accuracy of the fundamental period of the audio signal becomes rough.

(c) Problems to be Solved by the Invention In order to solve the above-mentioned drawbacks, the present invention is capable of solving the fundamental period (or fundamental frequency) of an audio signal with high accuracy without increasing the capacity of the storage means for arithmetic processing. ) is provided.

(d) Means for solving the problem The present invention converts an audio signal into a binary signal of "1°"0",
A 2 fI7i converting circuit that outputs the binary signal of the converted audio color circuit A clock circuit that generates a constant frequency clock, and a clock circuit that counts the output of the clock circuit and calculates the timing when the polarity of the binary signal of the audio is reversed. 5] A numerical circuit that outputs a numerical value related to 5], and an arithmetic means that inputs the binary code of the audio and the output of the counting circuit and detects the fundamental period of the input audio signal.

This calculation means detects the fundamental period of the audio signal according to the following procedure.

(1) Time series of polarity reversal of audio signal [Ti, Ti-+, Ti-z,...Ti-
+1] (however, Ti is at the time of the last capture), <2) Assuming the interval from the 8th previous polarity reversal as L(N), L(N) = I Ti-Ti-5I, (L(N) is the absolute value), (3) M times the value of L(N) (M -2, 3,...-M
o) and the interval error of MN details from Ti, E(NM
)=l (M-L(N)-|Ti-T open 1) 1, (4>The integrated value of the error obtained by integrating this error E(NM) from M-2 to M-Mo is 5(N) As, 8(N)maroΣ
Find E (NM) M=2 and calculate 5) N when the cumulative value of this error 5(N) is the minimum.
L(N) corresponding to is detected as the fundamental period of the input audio signal.

(E) Effect Since the present invention is configured as described above, even if the oscillation frequency of the clock circuit is set high, the amount of polarity inversion time series data stored for calculating the fundamental period of the audio signal is increased. Therefore, this storage means can also be configured to be small. In other words, there is an effect that the fundamental period of the audio signal can be detected with high accuracy.

(f) Example Next, embodiments will be described with reference to the drawings.

FIG. 1 is a block diagram of a fundamental period detection circuit for an audio signal according to the present invention, in which (1) is an input terminal for an audio signal, (2) is a filter circuit, and (3) is a band limited by the filter circuit (2). (5) is a clock generation circuit; (6)
(4) is a counting circuit that divides the frequency of the clock, and (4) is an arithmetic circuit that inputs the audio binary code and the output of the counting circuit, calculates the fundamental period of the audio signal, and outputs the result. Arithmetic circuit (
The fundamental period of the audio signal calculated and detected by 4) is output to the output terminal (0+"15e) of the calculation circuit (4), and when connected to an electronic musical instrument, etc., it constitutes an electronic musical instrument that can be played by voice. The arithmetic circuit 4) is RA
It can be composed of a microcomputer with storage means such as M. Further, the clock generation circuit (5) and the counting circuit (6) can also be realized using the timer function of a microcomputer.

FIG. 2 is a waveform diagram for explaining the function of the binarization circuit (3) of FIG. 1. In the figure, (a) is an example of the audio signal waveform applied to the audio signal input terminal (1), and (
at) to (a2) is this fundamental period. In order to reduce fluctuations near the AC crossover point during this fundamental cycle, a filter (2) passes low frequency components. In the case of FIG. 2, the fundamental period of the waveform (c) obtained by converting this band-limited audio signal waveform (b) into a binary code by the binarization circuit (3) is between (C+) (c 5), (c2) (c6) between,
((!3) It is clear that it is between (c7).

When considering a human singing voice, the fundamental frequency of the voice for an adult male is 100H, although it varies depending on age and gender.
In many cases, it is more than z. On the other hand, the duration of a musical scale is approximately 174 to 178 seconds in a musical score with a sixteenth note length, and therefore, an audio signal waveform in which the fundamental period of the audio does not vary more than 10 times can be expected. For women and children, the fundamental period is even shorter, and therefore the number of times the fundamental period does not fluctuate is even longer. Considering these characteristics, the fundamental period of an audio signal is detected by detecting a pattern of periodically repeated binary codes. The specific method will be described below.

FIG. 3(a) shows a diagram in which the input audio signal is converted into a digital signal (No. 8) of "1" and "O" by the binarization circuit (3). Each time the polarity of this binary code is reversed, The output of the counting circuit (6) is stored as timing information.From this polarity reversal time series [Ti, Ti-1, . . . ], the audio signal is The fundamental period is detected. That is, this detection may be performed every rising edge or every other rising edge.

Detection of the fundamental period when the polarity reversal timing (Ti) is stored will be described below.

Let the time interval from Ti to Ti-5 be L:(N).

L(N)=l Ti-Ti-吋1 Since the two main cylinders of this audio are band-limited by the filter (2), the basics of the audio signal are that the polarity is reversed every 2 to 6 times (an even number). There is a cycle. If L(N) is the fundamental period, the interval from the time of Ti is M times L(N) (M=2.3
．． . . . Mo) must also undergo polarity reversal. In addition, if the influence of noise etc. is weak, the number of polarity inversions in the fundamental period interval should be equal, so the interval (l Ti
-Ti-1NI) and the error E(
NM) as E (NM)=I (M-L (N)-|Ti-τ
1-M5lN, calculate the integrated value of this error 5(N),
M.

5 (No) = ΣMoM = 2E (NM) M = 2 Let N be No when 5 (N) of O is the minimum, and this N
Find o.

S(N,)=M i n 5(N) (When N in S(N) is an odd number, the polarity inversion circuit is an odd number and cannot be the fundamental frequency, so NO is always an even number.) It is clear that L(N) corresponding to this NO is the fundamental period to be sought. If there is a fundamental period at every No. of polarity reversals, the interval between every N polarity reversals is the same as the interval between adjacent ones, the interval with the previous Nth one from Ti, and the interval with every Nth one from now on. Since they should match with a small error, in this case, let the cumulative error value 5(N) be lTi-+ -Ni-+-NI),
The same thing can be done by finding the case where the accumulated value 5(N) of the errors is the minimum.

In addition, in the above explanation, it was stated that the timing value series used to calculate the fundamental period is stored every time the polarity is reversed, but even if the timing value series is stored every time the polarity is reversed, the calculation accuracy will decrease slightly, but the calculation accuracy will be reduced accordingly. can be sped up.

Also, the calculation of this basic period (L(No)) does not necessarily have to be done every time the polarity is reversed, but every time the polarity is reversed,
It may be set every other time.

As explained above, the arithmetic circuit (4) calculates the basic same frequency L (No) of the audio signal and outputs information corresponding to this L (No) to the 01 to 08 terminals. When driving, it is sufficient to quantize it in semitone units or scale units and output it.

According to such a configuration, the present invention provides a clock circuit (5)
Even if the frequency of the clock generated by is set high and the time precision of the timing value at the time of polarity inversion is set high, the time required for calculation and the amount of storage elements required will not become larger than necessary.

(G) Effects of the Invention According to the present invention, the time error can be extremely reduced when detecting the fundamental period of an audio signal, and there is no need to increase the time required for calculation or the amount of storage elements required. .

[Brief explanation of drawings]

FIG. 1 is a block diagram of a fundamental period detection circuit for an audio signal according to the present invention, FIG. 2 is an explanatory diagram of a binarization circuit, and FIG. 3 is a waveform example of two main cylinder numbers. <1) is the audio signal input terminal, (2) is the filter circuit,
(3) is a binarization circuit, (4) is an arithmetic circuit, (5) is a clock generation circuit, and (6) is a counting circuit.

Claims (3)

[Claims] (1) (a) A binarization circuit that converts the applied audio signal into a binary signal and outputs it; (b) A clock circuit that generates a clock of a predetermined frequency; (c) Counts the output of the clock circuit. 2 of the audio signal
a counting circuit that outputs a numerical value related to the timing of polarity reversal of the value signal; (d) inputting the output binary signal of the binarization circuit and the output of the counting circuit to detect the fundamental period of the input audio signal; (d-1) When the polarity of the binary signal of the audio signal is reversed (only at the time of rising or falling, or even when the polarity is reversed every predetermined number of times, A predetermined number (K) of timing value sequences (Ti, Ti_-_1, T
i_-_2,...Ti_-_K_+_1), (
However, Ti is the last imported timing value and Ti_-
_1 is the value imported one time ago, Ti_-_2 is the value imported two times ago)
and every time the polarity is reversed, N pieces (N=1, 2,
...) Let the interval with the previous timing value be L(N), L(N)=|Ti-T(i_-_N)|(L(N) is the absolute value), and (d-3 ) M times the value of L(N) (M・L(N): M=2
, 3...Mo) and the error E(NM) in the interval between Ti and Ti_-_M_N is E(NM)=|{M・L(N)-|Ti-Ti_-_M
_N|}|, (d-4) Let S(N) be the integrated value of this error E(NM), find S(N)=Σ＾M＾o_M_=_2E(NM), (d-5) A fundamental period detection device for an audio signal, characterized in that L(N) corresponding to N when the accumulated value S(N) of errors is minimized is detected as the fundamental period of an input audio signal. (2) The calculation means calculates the accumulated error value S(N) from the timing value series of polarity reversal, S(N)=Σ^N^(^M^o^-^1^)_j_=_
N｜{｜Ti＿−＿j−Ti＿−＿j＿−＿N｜−｜T
i−Ti_−_N|}|, |Ti−Ti_ when the S(N) is the minimum
-_N| is detected as the fundamental period of the input audio signal. 2. The fundamental period detection device for an audio signal according to claim 1, wherein -_N| is detected as the fundamental period of the input audio signal. (3) The calculation means calculates the accumulated error value S(N) from the timing value series of polarity reversal, S(N)=Σ^N^(^M^o^-^1^)_j_=_
N｜{｜Ti＿−＿j＿＋＿N−Ti＿−＿j｜−｜T
i_−_j−Ti_−_j_−_N|}| as the S(
2. The fundamental period detection device for an audio signal according to claim 1, wherein |Ti−Ti_−_N| when N) is the minimum is detected as the fundamental period of the input audio signal.