B2P3 - Digitisation and lossy compression Flashcards by Don Stuckey

a.c. coefficients

weights the basis functions at (1-7, 1-7) of a DCT table, the results evaluating to the amount of each spatial-frequency in the macroblock;

How well did you know this?

Not at all

Perfectly

Adaptive Multi-Rate (AMR)

a speech coding standard that adapts its bit rate to channel conditions;

How well did you know this?

Not at all

Perfectly

advanced audio coding (AAC)

MPEG-2/4 streamingFormat, with better quality than MP3 but with same bitRate;

How well did you know this?

Not at all

Perfectly

aliasing

where samplingRate < 2*f;

How well did you know this?

Not at all

Perfectly

analogue-to-digital converter (ADC)

takes an analogue signal, samples it, n-bit-quantises it, and returns a digital signal;

How well did you know this?

Not at all

Perfectly

basis function

reps a specific hori-vert spatial-frequency, appearing as an entry in a DCT table;

How well did you know this?

Not at all

Perfectly

B-frames

‘bi-directional prediction’ frames inserted between I-frames and P-frames filling in missing frames in the group of pictures (GOP) format;

How well did you know this?

Not at all

Perfectly

bit rate

number of bits passing a given point in the network per unit time;

How well did you know this?

Not at all

Perfectly

code-excited linear prediction (CELP)

= LPC - U/V_switch + codebook;

How well did you know this?

Not at all

Perfectly

coding efficiency

minimising bitRate while holding videoQuality constant, or vice versa;

How well did you know this?

Not at all

Perfectly

comfort noise (CN)

synthetic background noise for silent periods in voiceComms;

How well did you know this?

Not at all

Perfectly

d.c. coefficient

weights the basis function at (0, 0) of a DCT table, the result evaluating to the average spatial-frequency of the whole macroblock;

How well did you know this?

Not at all

Perfectly

dequantising

dequantise(n-bit_reps) = analogue;

How well did you know this?

Not at all

Perfectly

difference error

pixel-subtracting one video frame from another;

How well did you know this?

Not at all

Perfectly

difference pulse-code modulation (DPCM)

= PCM + predictionOfDataSamples (used in JPEG);

How well did you know this?

Not at all

Perfectly

discontinuous transmission (DTX)

where mobile devices are powered-down when there is no voiceInput;

How well did you know this?

Not at all

Perfectly

discrete cosine transformation (DCT)

DCT(spatialRep) = frequencyRep (used in J/MPEG);

How well did you know this?

Not at all

Perfectly

discrete wavelet transform (DWT)

DWT(spatialRep) = frequencyRep (used in JPEG2000; better performance at low bitRates);

How well did you know this?

Not at all

Perfectly

distortion

= reconstructedSignal - originalSignal;

How well did you know this?

Not at all

Perfectly

encoding

encode(symbols) = bitStrings;

How well did you know this?

Not at all

Perfectly

formants

the vocal tract’s spectral peaks characterising ‘voiced’ speechSegments;

How well did you know this?

Not at all

Perfectly

frequency masking

where high-amplitude sounds cover up low-amplitude sounds at neighbouring frequencies;

How well did you know this?

Not at all

Perfectly

group of pictures (GOP)

concatenating I-, P-, and B-frames for videoCompression (used in MPEG-1/2/4);

How well did you know this?

Not at all

Perfectly

H.264 / advanced video coding (AVC)

an MPEG-4 videoCodingStreaming tech;

How well did you know this?

Not at all

Perfectly

HE-AAC version 2

a tech optimised for low-bitRate audioStreaming apps including digi-radioBroadcasting;

I-frames

'intra-frames' are real reps of world (i.e. they are not fillers nor predictions);

inter-frame compression

stateful video compression;

intra-frame compression

stateless video compression;

inverse DCT

DCT_inv(frequencyRep) = spatialRep;

Joint Photographic Experts Group (JPEG)

an lossy compression standard for images using DCT;

linear predictive coding (LPC)

a sourceCoding method using parameters designed for speechSignals;

low frequency effect (LFE / subwoofer)

where the channel delivers bass-only information (<120 Hz);

macroblock

the formalParameter for the DCT() function, consisting of 8x8 pixels for JPEG and 16x16 pixels for MPEG;

mask-to-noise ratio (MNR)

determines bits per sub-band in MP3 using psychoacoustic model;

motion-JPEG (M-JPEG)

stateless video compression;

motion prediction

frame_n.macroblocks = frame_n-1.macroblocks + motionVectors;

motion vectors (MVs)

used in MPEG video coding to predict camera and object motion;

MPEG audio layer 3 (MP3)

a compressed audio file format;

noise masking

= frequency masking, temporal masking, and perceptual masking;

Ogg Vorbis (OV)

an open-source lossy audio compression format;

perceptual noise substitution (PNS)

where real noise is replaced with randomly generated noise;

perceptual redundancy

where sounds / images are irrelevant as humans do not perceive them;

P-frames

'prediction-frames' do not rep the world and instead are an educated guess based on the previous I-frame;

prediction (residual) error

= P-frame(t)* - I-frame(t); *based on I-frame(t-1);

pulse code modulation (PCM)

PCM(analogueSignal) = encoding ( quantising ( sampling ( analogueSignal ) ) ) = digitalSignal;

quality of service (QoS)

QoS(network) = minThoughput / maxLatency / lowestBER;

quantisation noise (quantisation error)

= analogueSignal - digitalSignal;

quantisation (quantising)

n-bit-quantise(sample(analogueSignal)) = digitalSignal;

rate-distortion (RD) curves

bitsPerSymbol plotted against distortionError;

region-of-interest (ROI) coding

where different areas of an image are coded at different bitRates (used in JPEG2000);

resolution

bitsPerSymbol;

sampling

measuring an analogueSignal at regular intervals;

sampling theorem

IF (samplingRate >= 2*f) THEN (signal can be reconstructed);

scalabililty (subsets)

where video/audio compression has subsets of varying bitsPerSymbol;

statistical redundancy

where repeating patterns are identified and repped with a symbol;

temporal masking

where a high-amplitude sound covers up a subsequent low-amplitude sound;

thresholding

setting a constant such that any values lower than this can be ignored;

unvoiced sounds

where there are no formants; can be repped by Gaussian noise;

vocoder

portmanteau of 'voice' and 'encoder'; compresses speech;

voiced sounds

where there are formants;