JP2010197645A - Note detector - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a note detector which can rapidly perform a note detection from an early time point of rising a musical signal, and is excellent in real time performance. <P>SOLUTION: An amplitude detecting section 20 detects an amplitude of an output signal from each of a plurality of complex band pass filters (BPF) 10, and an adjoining sound removal section 30 makes the amplitude signal from each amplitude detecting section 20 zero, or outputs it as it is. A comparison section 40 compares an output signal from the adjoining sound removal section 30 with a threshold, and outputs what exceeds the threshold. A sound region limit section 50 extracts and outputs the lowest note exceeding the threshold, and notes within nineteen halftones portion from this. The complex BPF creates two kinds of signals by multiplying each of two phase sine waves with the musical signal, and each of two kinds of created signals is independently passed through a low pass filter of the same configuration, and thereby, note detection is promptly performed from the early rising time point of the musical signal. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to an apparatus capable of detecting notes (sound names) of constituent sounds constituting a musical tone signal based on the inputted musical tone signal.

  As shown in FIGS. 7 (a), 7 (b), and 7 (c), for example, in the case of a keyboard instrument, even if a performance is performed so that only the fundamental tone is generated, an unnecessary harmonic sound is generated. FIG. 7A is an explanatory diagram of the generation state of the harmonic sound when the performance is performed so that only the note “E2” is pronounced, and the second harmonic, the third harmonic, and the fourth harmonic are generated with respect to the fundamental tone “E2”. ing. Similarly, FIG. 7B is an explanatory diagram of the generation state of the harmonic sound when the performance is performed so that only the note “B2” is generated, and the second harmonic and the third harmonic are generated with respect to the fundamental tone “B2”. In addition, as shown in FIG. 7C, even if the performance is performed so that only the note “G # 3” is sounded, the second overtone is generated. The triple overtone is shown as a non-playing sound because the note name is different from the played note and is particularly affected. Moreover, although there are actually more than five harmonics, they are omitted here. In view of this, there has been proposed an apparatus that eliminates such inconvenience by performing a third overtone removal process (see, for example, Patent Document 1). In this apparatus, the envelope detector detects and outputs envelope signals of output signals from a plurality of bandpass filters whose center frequencies of the respective passbands are shifted by a semitone frequency, and the third harmonic Since the removal unit outputs a signal obtained by removing the third harmonic of the entire envelope signal, and the pitch name extraction unit extracts the pitch names of the constituent sounds of the chord based on the output signal from the third harmonic removal unit. It has been possible to improve the detection accuracy of each note composing a chord by removing the influence of the third overtone having each component sound as a fundamental tone.

JP 2008-46321 A

  Certainly, according to the conventional apparatus as described above, the influence of the third harmonic is removed and the constituent notes can be accurately detected. However, when the envelope value of the third harmonic is large, it cannot be completely removed. There was a problem (see FIG. 8 (a)). Moreover, it cannot be denied that there is difficulty in responsiveness for various reasons. For example, in the conventional apparatus, the envelope detection unit is peak-held with a full-wave rectification unit that performs full-wave rectification on the input signal (BPF output), a peak hold unit that performs peak hold on the full-wave rectification signal, and the like. However, there is a problem that a signal processing delay occurs in the peak hold unit and the IP unit. In addition, there is a problem in operation stability because the peak value is updated at a half cycle of the signal during peak hold.

  The present invention has been made to solve such a conventional problem, and an object of the present invention is to provide a note detection device excellent in real-time property of note detection that can quickly detect a note from an early point of time when a musical sound signal rises. And

In order to achieve the above object, the present invention is an apparatus for detecting notes constituting a given musical tone signal,
A plurality of band pass filters in which the center frequency of each pass band is shifted by a semitone frequency;
A plurality of amplitude detecting means provided for each bandpass filter for detecting the amplitude of an output signal from each bandpass filter;
Determine whether the amplitude signal from each amplitude detection means is zero or whether to output as it is, and adjacent sound removal means for outputting the determined amplitude signal;
Comparing means for comparing the output from the adjacent sound removing means with a preset threshold value and outputting a value exceeding the threshold value;
An amplitude signal that exceeds the threshold, and includes a lowest note that is a note with the lowest frequency, and a range restriction unit that extracts and outputs notes within 19 semitones from the lowest note;
Each of the plurality of bandpass filters is
Two kinds of signals are generated by multiplying the given tone signal by each of two-phase sine waves, and each of the two kinds of generated signals is individually passed through a low-pass filter having the same configuration. It was made to be characterized by having comprised. Here, it is assumed that the minimum note is included in 19 semitones.

  According to the present invention, the amplitude detecting means is provided for each of a plurality of bandpass filters whose center frequencies of the respective passbands are shifted by a semitone frequency, and the amplitude of the output signal from each bandpass filter The adjacent sound removing means sets the amplitude signal from each amplitude detecting means to zero or outputs it as it is. Further, the comparing means compares the output from the adjacent sound removing means with a preset threshold value, and outputs the output exceeding the threshold value as it is. The sound range limiting means extracts and outputs the lowest note, which is an amplitude signal exceeding the threshold value, and a note within 19 semitones from the lowest note. This eliminates the need for subtraction, addition, and the like in the triple overtone removal process and the fundamental tone emphasis process as in the conventional apparatus, and makes it possible to remove the unnecessary sound only by comparing the magnitudes of the amplitude signals from the three amplitude detection means. Note detection can be performed accurately.

  Further, in this apparatus, each of the plurality of bandpass filters generates two types of signals by multiplying a given musical tone signal by each of two-phase sine waves, and generates two types of signals. Since each of the plurality of amplitude detecting means is individually passed through the same low-pass filter, each of the plurality of amplitude detecting means is first squared with the first signal that has passed through the first low-pass filter. , A second square unit that squares the second signal that has passed through the second low-pass filter, an adder unit that adds signals from each square unit, and an addition output of the addition unit And an amplitude calculation unit that obtains the positive square root to obtain an amplitude signal. Therefore, in the conventional apparatus, envelope detection is performed by the full wave rectification unit, the peak hold unit that performs peak hold on the full wave rectified signal, and the IP unit that performs interpolation processing on the peak held signal, and delay due to signal processing. However, this is solved and it is possible to realize note detection rich in real time.

  Then, the adjacent sound removing means corresponds to two band pass filters in which the amplitude signal from the amplitude detecting means corresponding to a certain band pass filter has a center frequency adjacent to the center frequency of the certain band pass filter. Only when the amplitude signal is larger than the amplitude signals from the two amplitude detection means, the amplitude signal from the amplitude detection means corresponding to the certain band-pass filter is output as it is. In other cases, the amplitude signal is set to zero. If the processing to be output is performed on the amplitude signals from the amplitude detection means corresponding to all the bandpass filters, the amplitude detection means shifts the center frequency of each pass band by a semitone frequency. For each of the plurality of band-pass filters, a wide pass band can be obtained, and a response rich in real time can be obtained.

  Further, in the conventional apparatus, when the full-wave rectification unit is used, the period fluctuation included in the full-wave rectified signal output remains, so that a correct value cannot be obtained, and each of the bandpass filters is prevented in order to hinder the stable operation of the adjacent sound removing means. The passband had to be narrowed, but this is eliminated and it is possible to realize note detection rich in real time.

  According to the present invention, it is possible to provide a note detection device that can quickly detect a note from an early point of time when a musical sound signal rises and is excellent in real-time note detection.

1 is a configuration diagram of a note detection device 1. FIG. FIG. 4 is an explanatory diagram of a center frequency setting state of each complex bandpass filter 10. 2 is a configuration diagram of a complex bandpass filter 10 and an amplitude detection unit 20. FIG. 3 is a configuration diagram of an adjacent sound removing unit 30. FIG. 6 is an explanatory diagram of the operation of the adjacent sound removal unit 30. FIG. It is explanatory drawing of the other structural example. It is explanatory drawing of a prior art. It is explanatory drawing of a prior art and a sound range restriction | limiting process.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

  FIG. 1 is a configuration diagram of a note detection apparatus 1 according to an embodiment of the present invention. The note detection device 1 is a device having a function of detecting notes constituting a given musical tone signal, and a plurality (30 in this example) of complex BPFs 10 (“BPF” means a bandpass filter). 30 amplitude detectors 20 provided for each complex BPF 10 to detect the amplitude of the output signal from each complex BPF 10 and output as an amplitude signal, and zero the amplitude signal from each amplitude detector 20 The adjacent sound removing unit 30 that determines whether to output the sound as it is, or the comparator 40 that compares the output from the adjacent sound removing unit 30 with a preset threshold value and outputs the signal exceeding the threshold value as it is. And a minimum note which is an amplitude signal exceeding the threshold and which is a note having the lowest frequency, and a range restriction unit 50 which extracts and outputs a note within 19 semitones from the lowest note. It is configured. The number of complex BPFs 10 is not limited to 30.

  The complex BPF1 (10), the complex BPF2 (10),..., And the complex BPF30 (10) are for the notes “D #”, “E”,. The center frequency of the pass band of BPF1 to complex BPF30 is set so as to be shifted by a semitone frequency. FIG. 2 is an explanatory diagram showing a part of the center frequency setting state, in which the center frequency f1 of the complex BPF1 (10) and the center frequency f2 of the complex BPF2 (10) are set to be shifted by a semitone. I understand. This is because the complex BPF1 (10) corresponds to the note “D #”, and the bandpass filter for the note “E”, which is a semitone higher than the note “D #”, is the complex BPF2 (10). Correspond. Note that the arrangement of semitones assumed in this apparatus is shown on the horizontal axis of FIG. Further, the amplitude detection unit 1 (20), the amplitude detection unit 2 (20),..., The amplitude detection unit 30 (20) are converted into complex BPF1 (10), complex BPF2 (10),..., Complex BPF30 (10), respectively. On the other hand, it is provided separately in the subsequent stage.

  FIG. 3 is a configuration diagram of one complex BPF 10 and an amplitude detection unit 20 provided in the subsequent stage. Each complex BPF 10 and the amplitude detector 20 have the same configuration, but the cutoff frequency of the LPF (“LPF” means a low-pass filter) is individually set for each complex BPF 10. In FIG. 3, reference numerals 11 and 12 denote multipliers, and reference numerals 13a and 13b denote LPFs having the same configuration. The multiplication unit 11 multiplies the given tone signal by “cos (−2πFct)” (cos is a cosine function, Fc is the center frequency of the BPF, and t is time), and the multiplication unit 12 is multiplied by the given tone signal. “Sin (−2πFct)” (sin is a sine function, Fc is the center frequency of the BPF, and t is time). That is, each complex BPF 10 generates two types of signals by multiplying the musical sound signal by two-phase sine waves (sin waves and cos waves having a phase different from “π / 2”). Each of the two types of signals is individually configured to pass through the LPFs 13a and 13b having the same configuration. Therefore, there is no response delay due to peak hold and interpolation processing as in the prior art, and note detection can be performed quickly from the time when the tone signal rises quickly. In each complex BPF 10, the relationship between the cutoff frequency Flpf of the LPFs 13a and 13b and the center frequency Fc of the BPF is “Flpf = Fc / (2Q)” (where “Q” is the Q value of the filter, for example, about 50). Become. Further, two types of signals are generated by multiplying the tone signal by each of the two-phase sine waves, and the two types of generated signals are individually passed through the LPFs 13a and 13b having the same configuration. With this configuration, there is an effect of shifting the center frequency Fc to DC (direct current component) by a two-phase sine wave, and the LPFs 13a and 13b have a BPF function because the band is limited to the DC center.

  The amplitude detection unit 20 includes a multiplication unit 21 (first square unit) that squares the signal LA (first signal) that has passed through the LPF 13a (first low-pass filter), and an LPF 13b (second low-pass filter). A multiplication unit 22 (second square unit) that squares the signal LB (second signal) that has passed through the filter), and an addition unit 23 that adds signals from the multiplication units 21 and 22 (square unit). And an amplitude calculation unit 24 that obtains the positive square root of the addition output of the addition unit 23 and outputs it as an amplitude signal. Therefore, in the conventional apparatus, a full-wave rectification unit that performs full-wave rectification on the input signal (BPF output), a peak hold unit that performs peak hold on the full-wave rectification signal, and an IP that performs interpolation processing on the peak-held signal. There is a problem that the envelope detection is performed and a delay occurs due to signal processing in the full-wave rectification unit and the peak hold unit. However, the configuration shown in FIG. This makes it possible to realize note detection rich in real-time characteristics.

  FIG. 4 is a configuration diagram of the adjacent sound removing unit 30. Reference numerals 31 and 32 are buffers, reference numerals 33 and 34 are adders, reference numerals 35 and 36 are arithmetic units, reference numeral 37 is an AND part, reference numeral 38 is a switching part, and reference numeral 39 is a zero part. The calculation unit 35 outputs “true signal (“ 1 ”in positive logic”) ”to the AND unit 37 only when the addition result of the addition unit 33 is“ positive (greater than 0) ”. Only when the addition result of the adding unit 34 is “0 or less (less than or equal to 0)”, the “true signal (“ 1 ”in positive logic”) ”is output to the AND unit 37. The AND unit 37 connects the switch of the switching 38 to the “b” side with respect to the switching unit 38 when the “true signal” is given from both of the calculation unit 35 and the calculation unit 36. In this case, a control signal is transmitted to the switching unit 38 so that the value “0” is output from the zero unit 39 by connecting the switch to the a side. When the switch is connected to the b side, the value of the buffer 32 (the value of the amplitude signal) is output as it is.

  Then, in the state where the amplitude signal to be processed is stored in the buffer 32, which is the target amplitude detection unit 20, in other words, whether the amplitude signal from the amplitude detection unit 20 is zero or is output as it is. The buffer 31 stores an amplitude signal from the amplitude detection unit 20 for the complex BPF 10 having a center frequency on the low frequency side lower than the center frequency of the complex BPF 10 for the amplitude detection unit 20 of interest by a semitone, and removes the adjacent sound. What is newly input to the unit 30 is an amplitude signal from the amplitude detection unit 20 for the complex BPF 10 having a center frequency on the high frequency side higher by a semitone than the center frequency of the complex BPF 10 for the amplitude detection unit 20 of interest. . As a result, the switch is moved to the “b” side only when the amplitude signal of the noted amplitude detection unit 20 is larger than the output of both the amplitude detection units 20 on the semitone low frequency side and the semitone high frequency side. When connected, the amplitude signal of the noted amplitude detector 20 is output as it is, while in other cases, the value “0” is output from the zero section 39 as an amplitude signal. In FIG. 4, it is described that there is one input line of the adjacent sound removing unit 30, but there are actually as many complex BPFs 10 as there are amplitude signals from all the amplitude detecting units 20. On the other hand, it is determined and output whether the amplitude signal is output as it is or zero is output.

  The upper part of FIG. 5 shows the signal from each amplitude detector 20 input to the adjacent sound removing unit 30, and the lower part of FIG. 5 shows the signal output from the adjacent sound removing unit 30 to the comparing unit 40. Is. In addition, the upper symbols (1), (2), (3),... (30) in FIG. 5 are “complex BPF1 (10) and amplitude detector 1 (20)”, “complex BPF2 (10)”, respectively. And “amplitude detection unit 2 (20)”, “complex BPF 3 (10) and amplitude detection unit 3 (20)”,..., “Complex BPF 30 (10) and amplitude detection unit 30 (20)”. a, b, c,..., g mean amplitude signals output from the respective amplitude detectors 20. That is, each amplitude signal is output from the amplitude detector 20 corresponding to the complex BPF 10 whose center frequency is shifted by a semitone.

  Now, by the operation of the adjacent sound removing unit 30 shown in FIG. 4, a process of setting the amplitude signal from each amplitude detecting unit 20 to zero or outputting it as it is is executed. The adjacent sound removing unit 30 includes two complex signals having an amplitude signal from the amplitude detection unit 20 (the attention amplitude detection unit 20) corresponding to a certain complex BPF 10 having a center frequency adjacent to the center frequency of the certain complex BPF 10. From two amplitude detectors 20 (two amplitude detectors 20 sandwiching the target amplitude detector 20 on the frequency domain) corresponding to the BPF 10 (in other words, two complex BPFs 10 on the low frequency side and the high frequency side) The amplitude signal from the amplitude detection unit 20 corresponding to the certain complex BPF 10 is output as it is only when the amplitude signal is larger than the amplitude signal of the complex BPF 10, while in other cases, the process of setting the amplitude signal to zero is performed for all the complex signals. The operation applied to the amplitude signal from the amplitude detector 20 corresponding to the BPF 10 is performed. In the specific example of FIG. 5, when attention is paid to the amplitude signal indicated by the symbol “a”, there is no low-frequency side amplitude signal (“0” when it does not exist), and the high-frequency side amplitude signal is indicated by the symbol “b”. As shown by “”, the amplitude signal of the code “a” is larger than the amplitude signal of the code “a”.

  Similarly, when attention is paid to the amplitude signal indicated by the symbol “b”, it is larger than the low frequency side amplitude signal “a”, but the high frequency side amplitude signal is larger than the amplitude signal of the symbol “b” as indicated by the symbol “c”. , The amplitude signal of the code b is set to “0”. When attention is paid to the amplitude signal indicated by the symbol “c”, the amplitude signal indicated by the symbol “c” is output as it is because it is larger than both the low frequency side amplitude signal “b” and the high frequency side amplitude signal “d”. Next, paying attention to the amplitude signal indicated by the symbol “d”, it is larger than the high frequency side amplitude signal “e”, but the low frequency side amplitude signal is larger than the amplitude signal of the symbol “d” as indicated by the symbol “c”. , The amplitude signal of the code “d” is set to “0”. That is, only when the amplitude signal of interest is larger than the two amplitude signals sandwiching it in the frequency domain, the amplitude signal of interest is output as it is, and in other cases, the amplitude signal is set to zero. Thus, even when the envelope value of the adjacent sound exceeds the threshold value, it becomes possible to detect stably, and thereby the bandwidth of the BPF can be designed wide. You can do it. In addition, the threshold value of the comparator 40 can be set low, and this also makes it possible to quickly detect a note from the time when the tone signal rises early. It should be noted that the equal sign “=” determination function of the calculation unit 36 of FIG.

  Next, the comparing unit 40 compares the output from the adjacent sound removing unit 30 with a preset threshold value, and outputs a value exceeding the threshold value. For example, as shown in the lower part of FIG. 5, the amplitude signals “c” and “f” exceeding the threshold are output to the sound range restriction unit 50.

  Then, the range restriction unit 50 extracts and outputs the lowest note that is an amplitude signal that exceeds the threshold used in the comparator 40 and that is within 19 semitones from the lowest note. For example, in the example of FIG. 5, the lowest note and a note within 19 semitones from the lowest note are extracted and output (the extracted note is indicated by “0N”). FIG. 8 (b) describes this in more detail. In FIG. 8B, note “E2” is the lowest note, and this lowest note and notes within 19 semitones from this lowest note (specifically, notes up to “A # 3”) are surrounded by a dotted line in a square shape. ) Note that the reason for extracting and outputting the notes within “19 semitones” from the lowest note is that, based on various experiments, it is empirically obtained that “19 semitones” can be detected by avoiding non-performance sounds. This is because the.

  As described above, according to the note detection device 1, it is possible to perform note detection quickly, and as a result, it is possible to perform note detection rich in real time. Note that a circuit as shown in FIG. 6 may be provided before the complex BPF 10. In other words, the anti-aliasing filter 60 removes the aliasing noise from the musical sound signal, and the downsampling unit 62 down-dumps the signal from which the aliasing noise has been removed, for example, at an interval of 1/40, and stores it in the buffer 64. In addition, it is good also as a structure supplied to each complex BPF10, and the structure which supplies a musical sound signal which gave the processing speed by the side of the note detection apparatus 1 sufficient.

  As described above, according to the present invention, it is possible to provide an apparatus having a note detection function suitable for application in the music field.

1 Note Detection Device 10 Complex BPF (Band Pass Filter)
20 Amplitude detection unit 30 Adjacent sound removal unit 40 Comparator 50 Sound range restriction unit

Claims (3)

A device for detecting notes constituting a given musical sound signal,
A plurality of band pass filters in which the center frequency of each pass band is shifted by a semitone frequency;
A plurality of amplitude detecting means provided for each bandpass filter for detecting the amplitude of an output signal from each bandpass filter;
Determine whether the amplitude signal from each amplitude detection means is zero or whether to output as it is, and adjacent sound removal means for outputting the determined amplitude signal;
Comparing means for comparing the output from the adjacent sound removing means with a preset threshold value and outputting a value exceeding the threshold value;
An amplitude signal that exceeds the threshold, and includes a lowest note that is a note with the lowest frequency, and a range restriction unit that extracts and outputs notes within 19 semitones from the lowest note;
Each of the plurality of bandpass filters is
Two kinds of signals are generated by multiplying the given tone signal by each of two-phase sine waves, and each of the two kinds of generated signals is individually passed through a low-pass filter having the same configuration. A note detection apparatus characterized by comprising the above. The apparatus of claim 1.
Each of the plurality of amplitude detection means includes:
A first squaring unit that squares the first signal that has passed through the first low-pass filter, a second squaring unit that squares the second signal that has passed through the second low-pass filter, and 2 A note detection apparatus comprising: an adder that adds signals from a multiplication unit; and an amplitude calculator that obtains a positive square root of the addition output of the adder to obtain an amplitude signal. The device according to any one of claims 1 and 2,
The adjacent sound removing means includes
An amplitude signal from the amplitude detection means corresponding to a certain band pass filter is obtained from two amplitude detection means corresponding to two band pass filters having a center frequency adjacent to the center frequency of the certain band pass filter. Only when the amplitude signal is larger than the amplitude signal, the amplitude signal from the amplitude detection means corresponding to the certain band-pass filter is output as it is. In other cases, the process of outputting the amplitude signal as zero is performed for all bands. A note detection apparatus characterized in that it is a means for applying to an amplitude signal from an amplitude detection means corresponding to a pass filter.

Images