TY - PAT
T1 - Method and apparatus for processing a signal using a digital processor having a given word length
AU - Linnartz, J.P.M.G.
AU - Kalker, A.A.C.M.
AU - Talstra, J.C.
AU - Staring, A.A.M.
PY - 2005/4/6
Y1 - 2005/4/6
N2 - A digital signal processor such as a watermark detector (12-18) operating with integer arithmetic circuits has a certain accuracy. Each processing step (multiplication, addition) increases the number of bits (the word length). For example, the Fast Fourier Transform (12) having a butterfly structure requires a plurality of such processing steps to be performed. In practical implementations, the processing steps are recursively performed by a single integer arithmetic circuit having a given word length, say N. After each step, the word length of the signal is reduced to the given word length N by rounding, truncation, or some other smart form of quantization. In order to prevent quantization errors, the signal is pre-processed by a pre-processor (19) which reduces the word length and which is invariant with respect to the subsequent process.; This means that if the pre-processor operated with infinite accuracy, it would have no effect on the subsequent process. If such a pre-processor operates with finite accuracy, it will reduce the quantization noise.
AB - A digital signal processor such as a watermark detector (12-18) operating with integer arithmetic circuits has a certain accuracy. Each processing step (multiplication, addition) increases the number of bits (the word length). For example, the Fast Fourier Transform (12) having a butterfly structure requires a plurality of such processing steps to be performed. In practical implementations, the processing steps are recursively performed by a single integer arithmetic circuit having a given word length, say N. After each step, the word length of the signal is reduced to the given word length N by rounding, truncation, or some other smart form of quantization. In order to prevent quantization errors, the signal is pre-processed by a pre-processor (19) which reduces the word length and which is invariant with respect to the subsequent process.; This means that if the pre-processor operated with infinite accuracy, it would have no effect on the subsequent process. If such a pre-processor operates with finite accuracy, it will reduce the quantization noise.
UR - http://v3.espacenet.com/publicationDetails/biblio?DB=EPODOC&adjacent=true&locale=nl_nl&FT=D&date=20050406&CC=EP&NR=1520235A2&KC=A2
M3 - Patent publication
M1 - EP1520235
ER -