Voice Training Machine
6.111 Final Project
By: Masood Qazi
Zhongying Zhou
Introduction
American Idol is a trademark of FOX
Introduction
Problem
z
Causes of singing off tune
- Cannot hear ‘own’ voice
- Shortness of breath
- Faulty memory of tune due to background
instrumentals
z
Conventional solution
- Practice with tune to commit to memory
Relies subjectively on singer’s distinction of pitch
New Solution
z
Visual outputs quantifies degree the singer is
off-key
z
Audio outputs allow user to hear their voice
with the synthesized note in real time
z
Pause feature allows user to perfect singing
note by note
Project Description: Notes
k=11
f A4 = 440 Hz
G#
k=0
A
A#
G
f n = f A4
• 2 n /12
k = n mod 12
B
F#
F
C
E
C#
D#
k=6
D
+
Audio
Interface
fforward
pause
tempo_sel [1:0]
Display
RGB [23:0]
Note Decision
reset_sync (global)
Music
measure[7:0]
Reader
music_note[3:0]
rewind_sync
fforward_sync
FSM
Pulsify
pause_sync
data[7:0]
enable
Divider
VGA
addr [12:0]
pause_p
Timer
line_count[9:0]
clk
pixel_count[9:0]
rewind
&
expired
reset
audio_out[15:0]
fft_out[7:0]
fft_frame
singer_note[3:0]
en48k
time_interval
[3:0]
DCM
x3, 180
EN
Select
&
Add
voice[15:0]
en48k
tone[15:0]
music_note[3:0]
FFT
Debounce/Synchronize
headphone
audio_out[15:0]
bitcount[7:0]
clk12p288
D/A
tone[15:0]
Synthesizer
start_timer
mic.
voice_on
tone_on
tonev_up/dn
outv_up/dn
micv_up/dn
voice[15:0]
A/D
ROM
Memory
~64kb
vga signals
AC97: A/D, D/A, 48kHz
clk
count
1
2
3
4
5
6
7
8
9
10
11
serial_in
parallel_out
4’bX
4’b1010
4’b0110
busy
FFT
Block
Diagram
FALL2PU
fft_frame
xk_re[7:0]
SQUARE
xk_im[7:0]
FFT
& SUM
fft_out[7:0]
en48k
voice[15:0]
xn_index[9:0]
start
xn_re[15:0]
xk_index[9:0]
CONTROL
FSM
MAX
fft_frame
FSM
data[15:0]
addr[9:0]
RAM
singer_note[3:0]
Synthesizer
music_note[3:0]
z
z
z
note
to
period
period[15:0]
cyclic
counter
count[15:0]
Explore different waveforms
Possible filtering
Avoid 3rd harmonic
single period
waveform
tone[15:0]
Music Reading / Display
Memory
Memory – Play
Memory – Fast Forward
Memory – Pause
Music Reader
pause_p=0 & expired=1
& rewind=0 & fforward=0
rewind=1
rewind=1
Play
fforward=0
fforward=1
Rewind
rewind=0
pause_p=1
pause_p=1
Pause
pause_p=0
fforward=1
Fforward
Display – Block Diagram
top_y
music_note [3:0]
top_x
Display
d_music
rgb_m [23:0]
tone [15:9]
singer_note [3:0]
d_singer
rgb_s [23:0]
voice [15:9]
blank
d_fft
rgb_f [23:0]
To All
new_data
l_count [9:0]
p_count [9:0]
fft_out [8:2]
RGB [23:0]
Display – Monitor
Conclusion
z
We hope to achieve:
–
–
–
Less noise pollution
Provide authentic feedback to singer about
his/her ability
Introduce a greater degree of objectivity in
competitive singing