 |
Description  |
|
|
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention generally relates to notes detector which receives an
input of a musical instrument sound or a human voice, extracts its pitch,
determines its note and generates a musical sound electronically in
accordance with the result of the determination. More particularly, the
present invention relates to a note detector which receives a plurality of
musical instrument sounds and/or human voices, including a chord to
perform note detecting operations and to an electronic instrument with
such a note detector.
2. Description of the Related Art
Conventionally, techniques are provided of receiving an input of a musical
instrumental sound or a human voice, extracting its pitch, determining its
note, and printing out the result in the form of a score, or encoding and
recording the series of results of determination in order to execute an
automatic performance by generating other musical instrument sounds (for
example, as disclosed in Published Unexamined Japanese Patent Applications
Sho 57-692 and Sho 58-97179).
Actually, such prior arts can only handle an input of a single sound
basically and does not allow for an input of a plurality of sounds
including a chord.
It was proposed that the name of a chord is detected in response to
reception of an input of the chord and displayed in accordance with a
signal indicative of the name of the chord (Published Unexamined Japanese
Utility Model Application Sho 60-26091).
This application discloses provision of analog bandpass filters equal in
number to notes, holding the respective peak outputs from those filters,
and using the larger ones of the peak outputs as more powerful candidates
for the components of a chord using a level detector.
The technique of using such analog filters has the problem that the result
of the determination would be likely to fluctuate or be unstable due to
the ambient temperature or that the circuit structure would be
large-scaled, disadvantageously.
SUMMARY OF THE INVENTION
It is therefore an object of the present invention to provide a digitalized
note detector which obtains respective notes of one or more input sounds
in a short time, and operates in a stabilized manner on a small circuit
scale, and an electronic musical instrument using such note detector.
According to one aspect of the present invention, there is provided a note
detector comprising digital signal processing means for performing
different-characteristic digital filtering operations on a digital
waveform signal representing a given acoustic signal on a time divisional
basis to detect the respective levels of spectra of frequencies
corresponding to one or more notes contained in the given acoustic signal;
and means for detecting the one or more notes contained in the given
acoustic signal on the basis of the result of the digital filtering
operations performed by the digital signal processing means.
According to the note detector, signal processing is all performed in a
digital region.
More specifically, the digital filtering includes bandpass filtering with a
frequency corresponding to each note as the central frequency.
Alternatively, bandpass filtering may be performed by using two filters; a
high-pass filter and a low-pass filter.
In a preferred embodiment, the digital signal processing means may perform
a high-pass filtering operation with a predetermined characteristic and
sequentially perform on a time divisional basis low-pass filtering
operations one with a resonance having a peak at a frequency corresponding
to each note.
The digital signal processing means detects the level of a signal waveform
resulting from the digital filtering operation for each note. In an
illustrative structure to detect such level, the envelope of a waveform
signal resulting from each filtering operation is extracted. This
processing may include, for example, extraction of a peak level of the
waveform at predetermined intervals of time. According to a preferred
embodiment, the absolute value of the waveform signal resulting from the
filtering operation is taken and passed through a low-pass filter of a
resonance type with a peak at a frequency of 0.
Thus, when one or more notes are obtained on a real time basis, the results
may be used to in various applications. For example, each note may be
displayed and/or printed. Alternatively, the resulting signals may
sequentially be stored in a storage device, processed in a desired way,
displayed and/or printed in the form of a score (musical paper).
Alternatively, a tone based on the resulting signal may automatically be
played by using a musical sound generator in the form of a specified
musical sound, for example, a piano sound.
In an illustrative structure, a musical sound is generated from the musical
sound signal generating means on a real time basis in response to the
result of the detection. In this case, the musical sound may be output as
one having a predetermined timbre simultaneously with the generation of
the original acoustic signal with the same pitch as, or with a different
pitch from, the original acoustic signal after detuning or transposing.
Preferably, such electronic musical instrument has a plurality of musical
sound generating channels to which control means assigns notes on the
basis of the result of the detection.
According to one illustrative structure of an electronic musical instrument
of the present invention, when detecting means detects a note different
from those allocated already to the predetermined number of musical sound
generating channels, the control means allocates the present note to an
empty one of the musical sound generating channels to start generation of
the corresponding musical sound signal. When the detecting means detects
no more notes allocated to the channels, it instructs the musical sound
generating channel to stop the generation of the musical sound signal
corresponding to that note.
It will be obvious to those skilled in the art that the present invention
may have other structures, modifications and applications.
BRIEF DESCRIPTION OF THE DRAWINGS
Other objects and features of the present invention will be more clearly
understood by those skilled in the art by reading the following detailed
description of a preferred embodiment of the present invention, when taken
together with the accompanying drawings.
In the FIGURES:
FIG. 1 is a schematic indicative of the basic principles of one embodiment
of the present invention;
FIG. 2 is a frequency characteristic diagram of a bandpass filter H.sub.t
(z) of FIG. 1;
FIG. 3 is a schematic of improved versions of the basic principles of the
embodiment of FIG. 1;
FIG. 4 is a frequency characteristic diagram of cascaded high- and low-pass
filters H.sub.1 (z) and H.sub.2t (z) of FIG. 3;
FIG. 5 is a schematic of the high-pass filter H.sub.1 (z) of FIG. 3;
FIG. 6 is a frequency characteristic diagram of the high-pass filter
H.sub.1 (z) of FIG. 5;
FIG. 7 is a schematic of the low-pass filter H.sub.2t (z) of FIG. 3;
FIG. 8 is a frequency characteristic diagram of the low-pass filter
H.sub.2t (z) of FIG. 7;
FIG. 9 is a diagram indicative of a pole, a zero point, a polar vector and
a zero-point vector of the digital filter of FIG. 7;
FIG. 10 is a frequency characteristic diagram related to FIG. 9;
FIG. 11 is a schematic of the low-pass filter H.sub.E (z) of FIG. 3;
FIG. 12 is a frequency characteristic diagram of the low-pass filter
H.sub.E (z) of FIG. 11;
FIG. 13 illustrates the extraction of an envelope using the structure of
FIG. 11;
FIG. 14 is a schematic of the entire circuit of an embodiment;
FIG. 15 is a schematic of the internal circuit of a DSP of FIG. 14;
FIG. 16 is a flowchart indicative of the operation of the DSP of FIG. 14;
FIG. 17 is a flowchart indicative of the operation of a CPU of FIG. 14; and
FIG. 18 illustrates the operation of the embodiment.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT
A preferred embodiment of the present invention will be described in more
detail below with reference to the drawings.
Basic principles
The basic principles of the preferred embodiment will first be described.
FIG. 1 illustrates that a digital signal processor (below referred to as a
DSP) constitutes N bandpass filters one having a band width corresponding
to each note and also N envelope extracting circuits.
First, an input acoustic signal x(n) is converted to a digital signal (if
the acoustic signal is originally a digital signal, it is used as it is).
The signal x(n) is filtered by the N bandpass filters H.sub.t (z) by the
time divisional processing by the DSP. At this time, the respective
transfer functions of the bandpass filters are changed depending on the
corresponding notes of each of a plurality of octaves.
FIG. 2 shows the frequency characteristic of a Chebyshev-type bandpass
filter the transfer function of which is given by
##EQU1##
where t is a subscript designating each note.
If the bandpass filter with i=1 is constituted, the DSP performs the
following operation given by
Y.sub.t (n)=H.sub.t (0)(x(n)+b.sub.1t x(n-1)+b.sub.2t x(n-2))-(a.sub.1t
Y.sub.t (n-1)+a.sub.2t Y.sub.t (n-2))
If i.gtoreq.2, an operation similar to the above equation is performed
repeatedly.
The respective coefficients of each of the filters can numerically be
calculated.
As a specified example, if two cascaded digital bandpass filters for i=1, 2
at A.sub.4 =440 Hz are constituted under the following conditions (1-7 are
shown in FIG. 2),
1=1 dB
2=(sampling frequency f.sub.s)=10 KHz
3=12 dB or more
4=415 Hz
5=430 Hz
6=450 Hz and
7=466 Hz,
the following coefficient values of the digital filters are obtained:
H.sub.44Hz (0)=0.08192384
and
when i=1,
a.sub.1 (1)=-1.91442200776
a.sub.2 (1)=0.99333673
b.sub.1 (1)=-1.91105345727 and
b.sub.2 (1)=1:
when i=2,
a.sub.1 (2)=-1.9210712
a.sub.2 (2)=0.993606
b.sub.1 (2)=-1.93525314797 and
b.sub.2 (2)=1.
In this way, the operations for the bandpass filters are performed for each
note on a time divisional basis and as a result, signals Y.sub.t (n) where
t=1 to N are obtained.
The DSP then extracts the envelope of the signals Y.sub.t (n) on a time
divisional basis by calculating the respective peak levels (absolute
values) of waveforms Y.sub.t (n) at predetermined intervals of time (for
example, for the pitch periods corresponding to the respective notes).
Alternatively, the envelope is obtained by performing a particular digital
filtering operation on the absolute value signal .vertline.Y.sub.t
(n).vertline. as will be described later in more detail.
By such DSP time divisional processing, the respective envelope signals
E.sub.t (n) where t=1 to N for the corresponding notes are calculated, a
CPU such as a microcomputer determines the respective levels of the
outputs to detect one or more note signals contained in the original input
waveform signal x(n).
While the basic principles are to perform the filtering operations with the
bandpass filters having peaks for the respective notes on a time
divisional basis, the bandpass filtering may be realized by various types
of digital filters having similar functions without relying on Chebyshev
characteristic bandpass filters. Alternatively, the bandpass filters may
be realized by cascaded low- and high-pass filters.
One principle of reducing the number of multiplications when the filtering
operations are performed to facilitate filtering on a real time basis will
be described below. (Eight multiplications are required in the above
example of the Chebyshev type bandpass filters for A.sub.4 =440 Hz.)
Improved principles
FIG. 3 illustrates the improved principles on which the DSP performs a
digital filtering operation where the number of multiplications is
reduced.
First, an input acoustic signal x(n) in the form of a digital
representation is input to a high-pass digital filter H.sub.1 (z) where
x(n) is 0 at a frequency of 0 and maximum at a frequency of f.sub.s /2.
The structure of the filter will be described in more detail later. The
output Y(n) of filter H.sub.1 (z) is applied to low-pass filters H.sub.2t
(z) which act for the corresponding notes t on a time divisional basis.
Low-pass filters H.sub.2t (z) each have the characteristic of a resonance
type low-pass filter having a peak at a note frequency, as will be
described in more detail later.
Therefore, the frequency characteristic of a filter which includes cascaded
high- and low-pass filters H.sub.1 (z) and H.sub.2t (z) is a
pseudo-bandpass filter as shown in FIG. 4.
In FIG. 4, f.sub.1, f.sub.2, . . . , f.sub.N correspond to the note
frequencies, and N may be selected to be about 40-50 (three octaves to
four octaves). Of course, when a note is to be detected in a wider octave
range, either a high-speed DSP or a plurality of DSPs for concurrent
processing may be used for that purpose.
The outputs W.sub.t (n) of low-pass filters H.sub.2t (z) where t=1 to N are
applied to low-pass filters H.sub.E (z) which operate for the respective
notes on a time divisional basis. The characteristics of filters H.sub.E
(z) will be described in more detail later. The respective outputs E.sub.t
(n) will be the envelope signals for the corresponding notes. The
subsequent processing is similar to that described concerning the basic
principles.
The structures and characteristics of the respective digital filters of
FIG. 3 will be described below in more detail.
High-pass filter H.sub.1 (z)
FIG. 5 shows an illustrative structure of high-pass filter H.sub.1 (z)
which is a second-order FIR digital filter the transfer function of which
is given by
##EQU2##
In FIG. 5, the numeral 5-1, 5-2, each denote a delay element; 5-3, 5-4, 5-5
each, a multiplier; and 5-6, 5-7 each, an adder. When this high-pass
filter is realized by the operation of DSP, the following is executed:
##EQU3##
In this case, the multiplication of a coefficient and a signal can be
realized by a mere shift operation.
The frequency characteristic of the high-pass filter is given by
##EQU4##
where the characteristic is minimum at .OMEGA.=0 (at a frequency of 0 Hz)
and maximum at .OMEGA.=.pi. (f.sub.s /2 Hz). FIG. 6 shows such
characteristic.
Low-pass filter H.sub.2t (z)
FIG. 7 shows an illustrative structure of low-pass filter H.sub.2t (z)
which is a second-order IIR digital filter, the transfer function of which
is given by
##EQU5##
As will be understood later, .theta. and CY change depending on the
subscript t indicative of a note. The character r is a parameter
indicative of the intensity of the resonance (or the degree of a peak).
In FIG. 7, the numerals 7-1, 7-2 each denote a delay element; 7-3, 7-4, 7-5
each, a multiplier; and 7-6, 7-7 each, an adder. When this low-pass filter
is realized by the operation of the DSP, the following is executed:
W.sub.t (n)=CY.multidot.Y(n)+2r cos .theta.W.sub.t (n-1)-r.sup.2 W.sub.t
(n-2) (2)
The frequency characteristic of the low-pass digital filter is given by
##EQU6##
where the poles of the transfer function are at:
Z.sub.1 =re.sup.j.theta.,Z.sub.2 =re.sup.-j.theta.
and a double zero point lies at Z=0. FIG. 9 shows the disposition of the
poles, zero points, and the polar and zero-point vectors for
0<.theta.<.pi./2. As will be understood in FIG. 9, the length of vector
v.sub.2 decreases initially as .OMEGA. moves along a unit circle from
.OMEGA.=0 to .OMEGA.=.pi. and then increases. The minimum length of vector
v.sub.2 is in the vicinity of a pole (re.sup.j.theta.). It is known that
the magnitude of a frequency response at frequency .OMEGA. is the ratio in
length of zero-point vector v.sub.1 to vector v.sub.2 and that the phase
of the frequency response is the angle between the real axis and vector
V.sub.1 minus the angle between the real axis and vectors V.sub.2. FIG. 10
shows only the amplitude characteristic.
The magnitude of the frequency response (amplitude characteristic) is
inversely proportional to the magnitude of polar vector V.sub.2 and
maximum at .OMEGA. in the vicinity of .theta., which is also shown in FIG.
10. The sharpness of this peak is determined depending on the magnitude of
r. A filter having a sharp peak (resonance characteristic) in the vicinity
of r=1 is thus realized.
As will be clear from the above examination, if the value of .theta. which
is 2.pi.f.sub.t /f.sub.s is determined for each note, a low-pass digital
filter of a resonance type with a peak at a note frequency f.sub.t is
realized, as shown in FIG. 8. In this case, r and CY can experimentally or
mathematically be obtained such that r has a magnitude which does not
influence the level of the next note and that CY results in equal level
outputs W.sub.t (n) at the corresponding notes.
If, for example, the ratio in magnitude of the frequency response at the
note frequency f.sub.t of f to that at the note frequency f+.DELTA.f (or
f.sub.t+1) which is spaced from f.sub.t by .DELTA.f is required to be m:1,
the following quartic equation:
.vertline.H.sub.2 (e.sup.j2.pi.f/fs).vertline..sup.2 /.vertline.H.sub.2
(e.sup.j2.pi.(f+.DELTA.f)/fs).vertline..sup.2 =m.sup.2
is required to be solved with respect to r and one of the resulting r's
which satisfy 0<r<1 is selected to calculate the respective coefficients
-2r cos .theta. and r.sup.2. As a result of the calculation, for example,
if f.sub.s =5 KHz, f=440 Hz and m=4, -2r cos .theta.=-1.9773, r.sup.2
=0.9851 and CY =36.7. This process applies for other notes.
Low-pass filter H.sub.E (z)
FIG. 11 shows an illustrative structure of a low-pass filter H.sub.E (z)
which is a second-order IIR digital filter of the same type as the
low-pass filter H.sub.2t (z) described above. The transfer function of
filter H.sub.E (z) is given by
##EQU7##
This transfer function is obtained by substituting r=0.9 and .theta.=0
into the above low-pass filter H.sub.2t (z).
In FIG. 11, the numeral 11-1 denotes an absolute value circuit which
provides the absolute value of input signal W.sub.t (n) which is the
output signal from low-pass filter H.sub.2t (z) and the resulting output
.vertline.W.sub.t (n).vertline. is digitally filtered. The numerals 11-2,
11-3 each denote a delay element; 11-4, 11-5, 11-6 each, a multiplier; and
11-7, 11-8 each, an adder. When this low-pass filter is realized by the
operation of the DSP, the following equation
Et(n)=CE.vertline.W.sub.t (n).vertline.+1.8E.sub.t (n-1)-0.81E.sub.t
(N-2)(3)
is executed.
This low-pass filter has a resonance or amplitude type frequency
characteristic which has a peak at .theta.=0, as will be obvious from the
above description and shown in FIG. 12 where the coefficient CE is a
factor to make uniform the respective levels of the notes and can be
obtained, for example, experimentally when required.
FIG. 13 schematically illustrates an envelope signal E.sub.t (n) obtained
from the structure of FIG. 11. The absolute value circuit 11-1 converts
the negative crest values (shown by the broken lines in FIG. 13) to the
corresponding positive crest values, which are then subjected to low-pass
filtering, so that the filter eventually provides the DC component of the
waveform signal .vertline.W.sub.t (n).vertline..
The entire structure of an embodiment
Since the principles of the present invention have been described above,
the specified structure of an embodiment thereof will be described below
more specifically.
FIG. 14 illustrates the entire structure of the embodiment where CPU 1
controlls the entire system. The operation of CPU 1 is controlled by a
program stored in ROM 2 to perform various operations using RAM 3. In a
musical sound detector 4, an acoustic signal which may be a musical
instrument sound, a human voice or a reproduced sound from a tape
recorder, a radio, a television set or a CD player received through a
microphone 41 or a line input LINE IN is appropriately filtered by
low-pass filter 42, converted by A/D converter 43 to digital signal x(n)
using an appropriate sample frequency f.sub.s and then delivered to DSP
(digital signal processor) 44. DSP 44 processes signals using filter
coefficient ROM 45 which stores various coefficients for digital filtering
purposes, and work RAM 46 which stores input waveform signals x(n) and
data for filtering purposes.
The result of the signal processing by DSP 44 is delivered to CPU 1 which
uses those data for providing various control operations. CPU 1 is
connected through a bus to the elements 2-4 as well as to key-board 5,
display 6, printer 7 and musical sound generator 8 for controlling them.
Keyboard 5 has function switches as well as performance keys, the operation
of which can be detected by CPU 1 to allocate musical sounds to be
generated to musical sound generator 8.
Display 6 and printer 7 display and print, respectively, one or more notes
detected by note detector 4 under control of CPU 1. For example, the notes
contained in input sounds on a real time basis may be displayed or may be
compiled, displayed and/or printed as a musical score.
Musical sound generator 8 has a plurality of musical sound generating
channels. Assume that it has 4 channels, for example. The output musical
sound signal from musical sound generator 8 is output from speaker 10
through audio system 9 to which the signals from microphone 41 and/or line
LINE IN are also applied so as to be acoustic outputs when required.
Various types of sound source circuits may be used as the musical sound
generator 8; for example, PCM systems, FM systems, iPD systems, sinusoidal
waveform synthesizing systems, etc. Musical sound generator 8 is able to
generate a musical sound signal having a timbre specified by keyboard 5.
In this case, CPU 1 allocates a note which is to be output in order to
perform a musical sound generation.
For example, CPU 1 allocates the note, detected by note detector 4, to
musical sound generator 8 on a real time basis for generating the
corresponding sound. In that case, the sound can be generated with a pitch
slightly different from that of the original sound or it can be generated
with a transposition.
Alternatively, CPU 1 can sequentially store changes in the notes, detected
by note detector 4, in RAM 3 as sequencer data, and sequentially reads
such sequencer data in response to, for example, a play start command from
keyboard 5 to thereby generate the corresponding musical sound signals
from musical sound generator 8.
CPU 1 may perform various operations in accordance with the contents of the
programs stored in ROM 2.
Structure of DSP
FIG. 15 shows an illustrative structure of DSP 44 which is connected
through interface 441 to CPU 1 and A/D converter 43. Operation ROM 442
controls the operation of DSP 44. Address counter 443 accesses operation
ROM 442 for sequential operations.
CPU 1 instructs DSP 44 to read what operation program it should read from
operation ROM 442 to process signals. The output from operation ROM 442 is
also delivered to decoder 444 which outputs various control signals to
open/close the gates and latches in DSP 44 to perform desired signal
processing operations.
Filter coefficient ROM 45 and work RAM 46 are connected to a bus in DSP 44.
Thus, coefficient data and waveform signals are delivered to DSP 44 or
written into work RAM 46 when required in accordance with the program in
operation ROM 442.
DSP 44 further includes multiplier 445 and adder-subtracter 446 for
calculating purposes and connected to the bus in the form of a 2-input and
1-output configuration. Register group 447 includes a plurality of
registers which stores intermediate data in the course of calculation and
is connected to the input/output terminals of multiplier 445 and
adder-subtracter 446 through the bus.
DSP 44 delivers to address counter 443 a flag signal indicative of the
result of the determination through flag resistor 448 in order to make a
determination in accordance with the results of the calculation from
adder-subtracter 446. An operation signal, etc., read from operation ROM
442 are changed depending on the output of flag register 448.
Note detection
The operation of the present embodiment will now be described. First, note
detection by note detector 4 will be described. FIG. 16 shows a flowchart
indicative of the operation of DSP 44 in accordance with operation ROM
442. When DSP 44 starts note detection in accordance with a command from
CPU 1, it performs initialization (16-1) to clear work RAM 46 mainly.
DSP 44 waits for the termination of A/D conversion at sample periods from
A/D converter 43 (16-2). When it receives a signal indicative of such
determination, it stores the A/D converted input signals into work RAM 46
while sequentially incrementing the address. If a particular area of work
RAM 46 is used as a ring buffer (constituted by connecting the starting
and terminating ends of the area virtually), a limitless input signal can
be stored therein. The input signal will be the signal x(n) (see FIG. 3).
DSP 44 then performs the FIR high-pass filtering operation with H.sub.1
(z) (16-4) in accordance with equation (1) on the input x(n) received this
time as well as the last input x(n-1) and the last but one input x(n-2)
which are read from work RAM 46, using multiplier 445 and adder-subtracter
446 of DSP 44.
Subsequently, DSP 44 performs initial setting t=1 for IIR low-pass
filtering with H.sub.2t (z) for each note (16-5), and then the actual
filtering operation (16-6) in accordance with equation (2) using
multiplier 445 and adder-subtracter 446 of DSP 44 while reading the
respective coefficients CY, 2r cos .theta. and r.sup.2 from filter
coefficient ROM 45. The results W.sub.t (n) of those calculations are
sequentially stored into another particular area of work ROM 46 used as a
ring buffer. By doing so, DSP 44 reads W.sub.t (n-1), W.sub.t (n-2), . . .
, from the buffer for use in operations.
Subsequently, DSP 44 performs IIR low-pass filtering with H.sub.E (z) to
detect the envelope of each note (16-7) by reading the respective
coefficients CE, 1.8, -0.81 from filter coefficient ROM 45 using
multiplier 445 and adder-subtracter 446 of DSP 44 in accordance with
equation (3). Among these operations, DSP 44 also performs calculation of
the absolute value .vertline.W.sub.t (n).vertline. using adder-subtracter
446.
DSP 44 sequentially stores the results of these operations E.sub.t (n) into
another particular area 46 of work RAM 46 used as a ring buffer. By doing
so, DSP 44 can sequentially read E.sub.t (n-1), E.sub.t (n-2), . . . for
use in operations.
Sequentially, it determines whether t=N to see if these detecting
operations have been performed for all the notes (16-8). If the answer is
NO, DSP 44 increments t (16-9) and then performs filtering operations
again at steps 16-6 and 16-7.
When these filtering operations have been performed for all the notes, DSP
44 reports to CPU 1 the respective envelopes E.sub.t (n) where t=1 to N
for the notes (16-10) for preparation for the next A/D conversion (16-2).
Namely, DSP 44 repeatedly performs three different digital filtering
operations sequentially on a time divisional basis for each note at each
interval of sampling to thereby detect the respective envelopes of the
notes and hence the respective levels of the frequency spectra of
frequencies corresponding to the notes on a real time basis.
When CPU 1 reports the termination of the note detecting mode to DSP 44
through operation of keyboard 5, DSP 44 terminates the series of
operations (16-11).
Generation of a detected note by CPU 1
As mentioned above, CPU 1 receives from DSP 44 envelope signals E.sub.t (n)
where t=1 to N for the respective notes and hence the levels of the
frequency spectra of frequencies corresponding to the respective notes at
each period of sampling, so that these signals may be used for various
purposes.
One application directed to the generation of a musical sound signal from
musical sound generator 8 on a real time basis will be described below.
CPU 1 first determines whether it has received an envelope signal from DSP
44 (17-1). If the answer is YES, CPU 1 writes the envelope signal value
into RAM 3 (17-2). The CPU fetches the four greatest received envelopes
and designates them as MAXENV0, 1, 2 and 3 (17-3).
Subsequently, CPU 1 determines whether the greatest envelope value MAXENV0
has exceeded a predetermined threshold value (17-4).
Assume now that input signal x(n) delivered from A/D converter 43 to DSP 44
is set, for example, to a maximum of .+-.100 and that envelope value
E.sub.t (n) delivered from DSP 44 to CPU 1 is, for example, 1000 when only
one note is received for that maximum value. In the case of two input
notes, the respective note inputs are .+-.50, and the envelope value is
500 for the respective notes. Similarly, in the case of four input notes,
the envelope value is 250 for the respective input notes. If this
threshold value is set to an excessively large one, any note of a
plurality of sounds input may not be detected. Therefore, the maximum
envelope value (=1000).div.N(=50)=20 is set as a threshold value where N
is the number of notes to be extracted.
If this threshold value is not exceeded, control passes to step 17-5 where
CPU 1 determines whether there is a sound which is being generated
(whether there is a note signal under generation by using note generator
8). If the answer is NO, CPU 1 prepares for a next input of envelope
E.sub.t (n) where t=1 to N from the DSP 44 (17-1). If note generator 8
generates some note signal, CPU 1 determines that an acoustic input
comprising that note from microphone 41 or line LINE IN has disappeared
and instructs musical sound generator 8 to stop further generation of the
musical sound (17-6).
If the maximum value MAXENV0 of the envelope has exceeded the threshold
value, CPU 1 determines how many musical sound should be generated. To
this end, first, i is set to i=1 (17-7). CPU 1 then checks whether MAXENVi
(hence, in this case, the second largest envelope value MAXENV1) has
exceeded the threshold value (17-8) and is | | |
