Scribd
Upload
48 views27 pages

Multimedia Systems: Sreeraj K. P. Asst. Professor, Dec, Rset

The document summarizes two approaches to implementing analysis-synthesis filter banks: 1) Direct implementation in the time domain using overlapped FIR filters which requires alias correction in the frequency domain. 2) Windowing input samples and transforming to the frequency domain (DCT/DST) which requires time domain alias correction. It then describes the theory of polyphase filters which decompose a filter bank into a prototype lowpass filter modulated by sinusoids at the center frequencies of the desired bands, allowing implementation using DCT/DST.

Uploaded by

sujith_mathew
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
48 views27 pages

Multimedia Systems: Sreeraj K. P. Asst. Professor, Dec, Rset

The document summarizes two approaches to implementing analysis-synthesis filter banks: 1) Direct implementation in the time domain using overlapped FIR filters which requires alias correction in the frequency domain. 2) Windowing input samples and transforming to the frequency domain (DCT/DST) which requires time domain alias correction. It then describes the theory of polyphase filters which decompose a filter bank into a prototype lowpass filter modulated by sinusoids at the center frequencies of the desired bands, allowing implementation using DCT/DST.

Uploaded by

sujith_mathew
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
You are on page 1/ 27

MULTIMEDIA SYSTEMS Lecture 25

Sreeraj K. P. Asst. Professor, DEC, RSET

Polyphase filter implementation

Introduction
Two approaches to analysis synthesis filter design
Direct implementation of filter banks in time domain (through FIR filters) with overlapped frequency domain characteristics.
require frequency domain alias correction by the proper design of adjacent filter bank characteristics.

Window the input samples and transforms those through DCT/DST.


require time domain alias correction.

Theory of Polyphase filters


Design of a multiband filter requires alias cancellation between adjacent bands and that the filter shapes be controlled such that the transition bands of adjacent filters add to produce a flat response. This can be formed by first designing a low pass prototype filter with a controlled transition band frequency response. The filter bank can then be composed by multiplying the impulse response of the prototype low pass filter with a sinusoid having frequencies equal to the centre frequencies of the desired filters.

Theory of Polyphase filters


A bank of M filters to be synthesized is shown

Theory of Polyphase filters


Real bandpass filters are composed of two complex filters Fi (z) and Gi (z) located respectively at the positive and negative centre frequencies. If the prototype lowpass filter is a FIR filter with impulse response h(n) and z- transform H(z) then

Theory of Polyphase filters


We obtain a band of filers with i=0,1.M-1 and the corresponding composite filters are

where ai, bi, ci and di are complex constants

Theory of Polyphase filters


the analysis and the synthesis filter banks is shown as

If ai = bi*

Theory of Polyphase filters


hi(n) contains odd number of half cycles of sinusoids in 2M points If the input samples are given by x(n), n=0,1,., then filtered output

L : length of filter tap Substituting equation hi(n) in si(n) we obtain DCT/DST of the input samples, multiplied by the prototype low pass filters impulse.

Theory of Polyphase filters


The response of each band i is a modulation of the prototype response with a cosine term to shift the low pass response to the appropriate band. Hence, these are called polyphase filters.

Theory of Polyphase filters


The input samples are first multiplied by lowpass prototype filter h(n). Blocks of 2M products of the multiplications are accumulated with the sign of alternate blocks negated. These 2M values are then multiplied by M sinusoids to generate the M output values.

Theory of Polyphase filters

A typical response of low-pass prototype analysis filter h(n)

Windowing function c(n)

Polyphase analysis filter for MPEG-1 audio

Polyphase implementation of analysis filter bank

Polyphase analysis filter for MPEG-1 audio

The audio signal is shifted into a 512 samples X buffer, 32 samples at a time. The content of X buffer are multiplied by the C-window function c(n) and the results are stored into the Z-buffer. The Z-buffer contents are divided into eight 64-element vectors (taking M=32), which are summed to form a 64-element Y-vector. The Y-vector is transformed using MDCT to yield the 32-subband samples.

Polyphase synthecsis filter for MPEG-1 audio

Polyphase implementation of synthesis filter bank

Polyphase synthecsis filter for MPEG-1 audio

The 32 subband samples are transformed back to the 64 element V vector, using inverse MDCT (IMDCT). The V-vector is pushed into a FIFO which stores the last 16 V vectors. A U-vector is created from the alternate 32 component blocks and a window (called Dwindow) is applied to U to produce the Wvector, which is divided into 16 vectors, each having 32 values. These 16 vectors are added together to obtain 32 sample output.

Psychoacoustic Models

Psychoacoustic model classification

Model 1:
is computationally simple. has high accuracy at high bit rate.

Model 2:
is computationally complex. has high accuracy at low bit rate.

Psychoacoustic model classification Essential philosophies of both the models:


Compute Fourier power spectrum of the signal. (512 point FFT for layer 1 & 2 / 1024 point FFT for layer 3). Map the spectrum into critical band domain. Distinguish between the tonal and nontonal components. Calculate the masking function. Map these functions back to the subband domain.

Psychoacoustic model I
The auditory spectrum is approximated by a list of tonal and non-tonal components. Tonal components are selected by identifying the maxima in the spectrum whose height is greatest in the neighbourhood. All the remaining spectral lines are used for calculating the non-tonal components. They are grouped into critical bands. Within each critical band, a non-tonal component is represented. Then, the list of tonal and noise components are decimated by eliminating those components which are below the auditory threshold or are less than one half of a critical band width from a neighbouring component.

Psychoacoustic model I
To compute the masking effect of a tonal or non-tonal component on the neighbouring spectral frequencies, the strength of the component is summed with two terms called the masking index and the masking function.
Masking index: An attenuation term which depends on the critical band rate of the component and whether it is tonal or non-tonal. Masking function: An attenuation factor which depends on
Displacement of the component from neighbouring frequency. The component signal strength.

Psychoacoustic model I
Tonal masking index

Non- tonal masking index

Psychoacoustic model I

Psychoacoustic model I

Psychoacoustic model I
For a tonal component j, at critical band rate z(j), the masking threshold LTtm(j,i) at critical band rate z(i) is given by
LTtm(j,i) = Xtm(j)+avtm(z(j))+vf[z(i)z(j), Xtm(j)]
Xtm(j) : the strength of tonal component at frequency j avtm(j) : the tonal masking index at the critical band rate z(j), vf (i,j) : the masking function
i representing displacement j representing signal strength.

Psychoacoustic model I
For non-tonal components the masking index can be calculated as:
LTnm (j, i) = Xnm(j)+avnm(j)+vf[z(i) - z(j), Xnm(j)]

The global masking thresholds are computed for all spectral frequencies by adding the masking thresholds computed above, for all the neighbouring tonal & nontonal components, with the threshold of hearing. The minimum masking threshold function is determined for each sub-band from the minimum of all the global masking thresholds contributing to that sub-band. Signal to mask ratio (SMR) is computed.

Psychoacoustic model II
It does not make a distinction between the tonal and non-tonal components. Spectral data is transformed into a partition domain. 1024 point FFT computation is used. Tonality is decided by the unpredictability of the spectrum with time.

You might also like