Perceptual models

Before introducing the method for calculating the internal representation the psychoacoustic fundamentals of the perceptual model is explained in the next chapter. 

Using either the time domain input signal or the output of the analysis filter bank, an estimate of the actual (time dependent) masked threshold is computed using rules known from psychoacoustics. This is called the perceptual model of the perceptual encoding system. 

In practice, the calculation of the PE requires an analysis filter bank and a perceptual model. The PE is defined as... 

The main task of the perceptual model in a perceptual encoding system is to deliver accurate estimates of the allowed noise (just masked threshold) according to the time/frequency resolution of the coding system. Additional tasks include... 

To solve these tasks, perceptual models often work directly on the time domain input data thus allowing a time and/or frequency resolution of the model which is better than the time and/or frequency resolution of the main filter bank of the perceptual coding system. 

The data are quantized with possible modifications to the quantization step sizes according to a perceptual model. A count of how many bits are used for each component can only be done after the process is completed. 

Block companding. This method is also known under the name block floating point . A number of values, ordered either in time domain (successive samples) or in frequency domain (adjacent frequency lines) are normalized to a maximum absolute value. The normalization factor is called the scalefactor (or, in some cases, exponent). All values within one block are then quantized with a quantization step size selected according to the number of bits allocated for this block. A bit allocation algorithm is necessary to derive the number of bits allocated for each block from the perceptual model. In some cases, a simple bit allocation scheme without an explicit perceptual model (but still obeying masking rules) is used. 

Perceptual Models. It seems that the search for more accurate psychoacoustic models will not be over for some time to come. Progress at very low bit-rates and for variable rate coding depends on the availability of better perceptual models. One area with promising results is the application of nonlinear models as proposed in [Baumgarte et al., 1995],... 

Jot, 1992b] Jot, J. M. (1992b). Etude et realisation d un spatialisateur de sons par modeles physiques et perceptifs (Design and implementation of a sound spatializer based on physical and perceptual models, in French). PhD thesis, Telecom Paris. 

Petersen and Boll, 1981] Petersen, T. L. and Boll, S. F. (1981). Acoustic noise suppression in the context of a perceptual model. In Proc. IEEE Int. Conf. Acoust., Speech, Signal Processing, volume 1086-1088. 

The perceptual model as developed in this chapter is used to map the input and output of the audio device onto internal representations that are as close as possible to the internal representations used by the subject to judge the quality of the audio device. It is shown that the difference in internal representation can form the basis of a perceptual audio quality measure (PAQM) that has a high correlation with the subjectively perceived audio quality. Furthermore it is shown that with a simple cognitive module that interprets the difference in internal representation the correlation between objective and subjective results is always above 0.9 for both wideband music and telephone-band speech signals. For the measurement of the quality of telephone-band speech codecs a simplified version of the PAQM, the perceptual speech quality measure (PSQM), is presented. 

As discussed above, the model of hearing built into a perceptual coding system forms the heart of the algorithm. A lot of systems (like MPEG Audio, see below) just define the transmission format, thus allowing changes and improvements to the perceptual model even after a standard is fixed and a lot of decoders are deployed at the customers. 

The tonality index v(/,co) denotes the final result of the tonality estimation and can be applied to a perceptual model as for example the one described below. 

MPEG-1 perceptual model 2. As an example for actual perceptual models we give a description of the perceptual model 2 as it is described in the informative annex of MPEG-1 audio [MPEG, 1992],... 

As described above, the pre-echo control in the perceptual model can lead to a PE signalling a bit-rate demand which is increased by a possibly large factor. 

Perceptual Model. A psychoacoustic model similar to IS 11172-3 psychoacoustic model II is employed. 

I. J, Cox and M, L, Miller, A review of watermarking and the importance of perceptual modeling, Proceedings, SPIE Human Vision and Electronic Imaging II, volume SPIE Vol. 3016, February 1997. 

Fig. 5 Hierarchy of knowledge manifold interactions as described in text. Top Minimal input from naive category surface forces complete modeling at level of perception - surface to identify dog . Middle More complex category surface (more experienced) allows less perceptual modeling (more abstract) to reach the state of dog . Bottom Category influence in the content of a safe or dangerous environment influences minimal data stream to generate two different cognates, either a bear (dangerous) or a dog (safe)...

The cognitive submodel architecture allows for the development of higher-order functions of cognition. For example, Shiveley et al. 1995 has developed and demonstrated a situation awareness function that combines characteristics of working memory, long-term memory search, and perceptual models to develop an abstraction termed situation, which can then be used to guide behavior or to serve as a measure of adequacy of information in the environment and in the crew s knowledge store to meet task demands. 

The data sets available at the outset of an impact assessment are mostly of the first type. However, the environmental impact assessor will be guided to a certain extent in the selection of data sets by knowledge of the physical, biological, social and/or economic systems they are studying. Conversely, however, the data sources available within a region will influence the nature of the perceptual models used in the assessment. Where there are few data, the analysis wiU not include much detail. Supplementary data collected during the impact assessment should preferably be of the second type. The data should be sufficient to enable the prediction of an impact to be made within specified confidence Umits. The amoimt to be collected, the frequency, precision, accuracy, and type are dependent upon the known variability of the element in space and time. Where the variability is unknown, it must be determined by a pilot study. 

Grogan, T.A. 1992. Image evaluation with a contour-based perceptual model. Human Vision, Visual Processing, and Digital Display III, ed. B.E. Rogowitz, pp. 188-197. Proc. SPIE 1666. 

---