Process of selectively protecting information bits against transmission errors

Process for coding a first sequence of information bits, wherein there is provided at least a first protection class for M bits of the first sequence and a second protection class for M' bits of the first sequence, M and M' denoting two integers at least equal to 1, said process comprising the steps of:

• forming a second sequence of bits exhibiting redundancy from the first sequence, in such a way that the second sequence includes at least a first redundant subset consisting of L bits which represent the M bits of the first class of the first sequence with M<L, and a second subset consisting of L' bits which depend on the M' bits of the second class of the first sequence;
• forming a third sequence of bits by applying a differential coding of the form d.sub.n =c.sub.n .sym.d.sub.f(n) to the second sequence, where c.sub.n and d.sub.n denote the bits of rank n of the second and of the third sequence respectively, f(n) denotes an integer at most equal to n-1, and .sym. denotes the exclusive OR operation;
• and
• delivering a signal representing the bits of the third sequence in a specified order,
• and wherein the structure of the second sequence and the function f are chosen such that for each bit c.sub.n ' belonging to the second subset of the second sequence, there is at least one bit c.sub.n belonging to the first subset of the second sequence and such that n'=f(n) or f(n')=n or f(n).

Process for correcting a sequence of likelihood data r.sub.n obtained from a signal received from a sender device via a transmission channel, the sign of each likelihood data item r.sub.n representing an estimate, having a reliability which increases with the absolute value of r.sub.n, of a value of a respective bit d.sub.n of a coded sequence of bits resulting from a differential coding of a second sequence of bits by the sender device, the second sequence of bits including at least one redundant subset consisting of L bits which represent M information bits to be communicated with 0.ltoreq.M<L, wherein said differential coding is of the form d.sub.n =c.sub.n .sym.d.sub.f(n), c.sub.n and d.sub.n denoting the bits of rank n of the second sequence and of the coded sequence respectively, f(n) denoting an integer at most equal to n-1, and .sym. denoting the exclusive OR operation, the correction process including the steps of:

• calculating respective first estimates of the bits of the redundant subset, each first estimate of a bit c.sub.n of the redundant subset depending on the sign of r.sub.n .multidot.r.sub.f(n) ;
• detecting the bits of the redundant subset having erroneous first estimates due to errors resulting from the transmission channel, by exploiting the redundancy present in the redundant subset;
• and
• for each bit c.sub.n of the redundant subset whose first estimate has been detected as erroneous, modifying the sign of that of the two likelihood data items r.sub.n and r.sub.f(n) which has the smaller absolute value.