Dereje Teferi(PhD)
[email protected]
1
2
3
What is sound?
4
Wave parameters
• Sound travels in waves. Waves have the following characteristics
• Wavelength is distance Infra-sound 0 – 20 Hz
sound travels in one cycle Human hearing range 20 – 20 kHz
• 20 Hz is about 16.5 meters Ultrasound 20 kHz – 1 GHz
• 20 kHz is 1.6 cm Hyper sound 1 GHz – 10 THz
5
Phase angles
6
Fourier Series
The sine wave is the simplest possible wave.
In 1807, Fourier proved that any repeating wave form
could be broken down into a series of sine waves.
The lowest frequency sine wave (in a repeating wave
form) is called the Fundamental.
The frequencies of the other sine waves are always
integer multiples (2x, 3x, 4x, etc) of the fundamental.
These are called Harmonics.
Each harmonic has a different amplitude and phase
7
Adding Harmonics
8
Sound Fields
directed sound
early reflections reverberation
(50-80 msec)
amplitude
time
9
Analog to digital conversion
Sound waves are continuous
One needs analog to digital convertor to store
audio/sound in a computer (or any digital device)
Sampling is required
The sampled data then need to be quantized
10
Quantization
Quantization in general approximates a signal varying
continuously in amplitude by one whose amplitude is
restricted to a prescribed set of discrete values
It is a method to digitize the sampled analog signal
It determines how many discrete digital values should
be used to encode the sampled signals? and
What analog value does each digital value correspond
to?
Commonly used quantization methods are Linear (8 bit
and 16 bit) and logarithmic
11
12
The sample and hold technique
Sample at a specific time intervals
Hold the sample value until the next sample position is
reached
This allows quantization
Then the digitized codes are transmitted via digital devices
as numbers
13
14
Reconstruction of audio from
digitized data
15
Increasing the sampling rate
16
Low Vs. High sampling rate
17
More example
Real music sample at two different sampling rates
(8KHz and 48KHz)
18
Sampling
Information is lost during digitization especially in the
first example where the samples are small in number
This is basically due to the sampling rate
If too much samples are taken, the amount of data will be
huge
The questions now are
How many samples should we take between every cycle?
Should we take the same number of samples for high
frequency as well as low frequency?
19
Nyquist-(Shannon) Theorem
• Nyquist Theorem (written by Harry Nyquist in 1928 and
proved by Claude Shannon in 1949) suggests that Optimal
sampling frequency should al least be twice the highest
frequency of the audio to be sampled
• In mathematical terms:
fs > 2*fm
where fs is sampling frequency (the No of samples taken in one second) and fm is the
maximum frequency in the signal
• Although certain limitations are observed, any arbitrary
wave form can be recorded by taking samples at fixed time
intervals.
• The wave form is played back by outputting the same
samples at the same fixed intervals.
20
• Some example samples for specific applications are listed in
the table below
Format Sampling Rate Bandwidth Frequency Band
Telephony 8 kHz 3.2 kHz 200-3400 Hz
Teleconferencing 16 kHz 7 kHz 50-7000 Hz
Compact Disk 44.1 kHz 20 kHz 20-20,000 Hz
Digital Audio Tape 48 kHz 20 kHz 20-20,000 Hz
21
Sampling rate
22
Digitization and playback
23
Aliasing
Aliasing occurs when the input frequency is greater
than half the sampling frequency.
24
Audio filtering
To reduce aliasing errors, analog to digital convertors
apply low pass filtering on the input signal with a cut
of frequency of (fs/2)
the signal is sampled at sampling frequency fs
Then the sampled signal is quantized into discrete
values
25
26
Sampling rate Vs storage
27
Audio Compression
Compression in general is a mechanism used to reduce the
amount of storage space required by a signal
The signal could be text, audio, image, video or any
combination of them
28
Reading Assignment
Overview of speech codecs
PCM (Pulse Code Modulation)
DPCM: Differential PCM
ADPCM : Adaptive differential PCM
etc
29
30
