15.01.2013 Views

U. Glaeser

U. Glaeser

U. Glaeser

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

A new standard MPEG-7, named the multimedia content description interface, is aimed to support as<br />

broad range of communications and multimedia applications as possible [19]. MPEG-7 audio merges<br />

five different technologies: the audio description framework (scalable series, low-level descriptors, and<br />

uniform silence segments), musical instrument timbre descriptors, sound recognition tools, spoken<br />

content descriptors, and melody descriptors. In order to describe the low-level audio features, regions<br />

of similarity and dissimilarity within the sound are searched for. This can be done either using samples<br />

taken at regular intervals or segments of samples. The relevant samples are then further manipulated to<br />

form a scalable series, which allows to progressively down-sample the data contained in a series, according<br />

to the application, bandwidth, or storage requirements.<br />

The scope of newest MPEG-21 standard is the integration of technologies enabling transparent and<br />

augmented use of multimedia resources across a wide range of networks and devices to support functions<br />

such as content creation, content production, content distribution, content consumption and usage, content<br />

packaging, intellectual property management and protection, content identification and description, financial<br />

management, user privacy, terminals and network resource abstraction, content representation and<br />

event reporting [20].<br />

MUSICAM as well as MPEG-1 layers I and II coders have the same structure shown in Fig. 27.31. The<br />

input audio signal is transmitted via a 32-band polyphase analysis filter bank (Fig. 27.22(a)) with equally<br />

spaced passbands, according to Eqs. (27.50a–c). All subband filters with impulse responses H i(n), i = 0,<br />

1,…,31, determined by Eq. (27.48), are obtained by modulation of a single prototype lowpass filter with<br />

the impulse response h(n), as is illustrated in Fig. 27.22(c). Their output signals are critically decimated.<br />

For a 48 ksamples/s sampling rate, each subband filter has a passband width of 750 Hz. Although these<br />

filters are highly overlapping, they can guarantee a perfect (or at least a nearly perfect) signal reconstruction<br />

(via the synthesis filter bank in Fig. 27.22(b)) due to the power complementarity. For instance, at<br />

multiples of 750 Hz the respective filters, i.e., those with neighboring passbands exhibit a 3 dB attenuation.<br />

Samples of subband signal components are quantized with a number of uniform midtread quantizers<br />

with 3, 5, 7,…,65,535 possible levels. Blocks of samples are formed (e.g., blocks of 12 samples in layer I)<br />

and divided by a scalefactor s sf selected in such a way that the sample with the largest magnitude is scaled<br />

to 1. By this means a quite large overall dynamic range of approximately 126 dB is reached. Proper<br />

quantizers are selected with the dynamic bit allocation algorithm described in section 27.9, controlled<br />

PCM<br />

audio<br />

Encoded<br />

bit stream<br />

FIGURE 27.31 Structure of MUSICAM and MPEG-1 layer I and II coders: (a) encoder, (b) decoder.<br />

© 2002 by CRC Press LLC<br />

Analysis<br />

filter bank<br />

(time-to-frequency<br />

mapping)<br />

FFT<br />

Bit stream<br />

unpacking<br />

_ _<br />

|<br />

|<br />

|<br />

|<br />

|<br />

Scaling,<br />

bit allocation,<br />

and<br />

quantization<br />

Psychoacoustic<br />

perception<br />

model<br />

(a)<br />

Dequantization<br />

and<br />

descaling<br />

Ancillary data<br />

(if included)<br />

(b)<br />

|<br />

|<br />

|<br />

|<br />

|<br />

Bit stream<br />

formatting<br />

Ancillary data<br />

(optional)<br />

Sythesis<br />

filter bank<br />

(frequency-to-time<br />

mapping)<br />

Encoded<br />

bit stream<br />

PCM audio<br />

(reconstructed)

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!