Representing Digital Media (Audio)

Transducer:

- converting audio (sound waves) into an electrical signal; sometimes immediately digitising it

- Microphone: converts sound to electrical energy

- Speakers: convert electrical energy into sound

Ways of creating Digital Audio

1. Sampling
2. Synthesizing

Sampling

Analog sound from real world (like a voice) into an electronic device that will store it as a digital signal

Synthesizing

Electronic circuits synthesising (creating) digital audio

Sine Wave as function of time

F(t)= a sin (2 π f (t-0)) *zero has a line through it

F(t)= a sin (2 pi f (t-0))

A = amplitude
F = frequency
- Number of cycles per second → affects pitch of sound
T = time
- Independent variable → measured in unit of time
0 = phase: time shift

Amplitude

related to intensity of sound
Example: raising T.V volume raises amplitude

Frequency

elated to perceived pitch;

Phase

can depend on relative location of the sound

Can periodic functions be expressed as sum of sine waves?; Non-periodic?

Yes!; Make the period as long as the signal

Hearing Intensity Limits

- threshold of hearing: RMS sound pressure of 20 upa
- DB-SPL (sound pressure level) is a relative
logarithmic unit relative to the 20 uPa value above
DB-SPL = 20log(sound/threshold)

Hearing Frequency Limits

“16 Hz - 20 kHz”

Sampling Rate and Bit Rate *Testable

Digitising sound involves taking samples (measurements) at a fixed rate (sampling rate) and recording them

what happens when an audio sample its too fast? Too slow?

Too fast —> large file
Quickly uses up storage space
Too slow -> inaccurate
If a fast moving signal is sampled slowly→ it will
not accurately represent the signal

Sampling Period / Interval

time between one sample and the next

Sampling Frequency / Rate

Are specified in hertz (Hz)

Nyquist-Shannon (*Testable): Sample interval ≤ half period

Has a maximum (no min)
Example: highest frequency is 400 Hz, the required
sample interval should be 1/400 seconds

Nyquist-Shannon (*Testable): Sampling-rate (frequency) ≥ 2f

Has a minimum (no max)
Example: highest frequency is 400 Hz, so the required sampling frequency should be 800 Hz

Recall:

- Quantization: is representing a quantity with the number of bits available
- We want to represent as many levels of the
signal as possible & we want the resolution
- Resolution: smallest interval between levels

Bit Depth

- sample recorded using a fixed number of bits

- Higher bit depth means a more accurate sample

Can be used to record physical processes like blood pressure, heartbeat, motion due to walking or running

8 bits

Common in telephony

16 bits

Most high-quality sound (usually used for Music Audio)

24+ bits

- Higher quality/dynamic range
- Often used before or during sound processing /editing (mastering)

Bit rate

- Sample Frequency x Bit Depth
- Bit rate is the number of bits per second
- Higher bit rate = better quality = larger file size
Example: CD-quality audio
-16 bits bit depth (per channel, stereo has 2 channels)
- 44.1 kHz sampling rate
- 16 x 44.1k x 2 = 1.4Mbits/sec

lossless audio compression

Similar to text compression techniques

Lossy audio compression

1. Remove imperceptible sounds (using psychoacoustics models)
e.g., too close to a louder sound either in
frequency or in time
2. Reduce bit rate (less accurate reproduction of the original)
- Codecs
- Compression and decompression algorithms
or audio (and video)

FLAC (on test)

- Lossless
- Using a combination of run-length and Huffman coding
- Compression ratio around 62% (just know low compression ratio)

MP3 (on test)

(moving Pictures Experts Group, Audio Layer 3)
- Lossy, with a compression ratio around 13% (just know - low compression ratio)
-Uses psychoacoustics, Huffman encoding, lower bit rates

ACC (on test)

- Lossy,
- Compression ratio around 14% (just know low compression ratio)
- Uses psychoacoustics, Huffman encoding, and lower bit rates

Level of Compression:

Low compression ratio means heavily compressed
High compression ratio means light compression