Page 25, line 11: Should be “H=-30/31 log 30/31 -1/31 log 1/31 = 0.206 bits.”
Pages 65-67: The description of Golomb codes is inconsistent. One way of making it better is the following 5 corrections:
Page 65, line -9: “positive” should be “non-negative”
Page 65, line -7: Should be “… alphabet {0, 1, 2, …} with …”
Page 65, line -5: Should be “P(k)=1/2k+1”
Page 66, line 8: Should be “integer n>=0 using …”
Page 67, line 2: Should be “P(n)=pn(1-p)”
Page 71, table 3.22: “Probability” should be “Codeword”.
Page 97, line -1: should be “u(3)=0.312+(0.6-0.312)FX(2)=0.312+0.288*0.82=0.54816”
Page 100, line 17: should be “u(3)=0.312+(0.6-0.312)FX(2)=0.312+0.288*0.82=0.54816”
Page 185, equation 7.21: should be “A(n)=A(n-1)-qe”
Page 238, line 8: should be “[-100, -1), (1, 100] with probability …”
Page 241, figure 9.10: The region for the overload probability should be everything outside [-4*delta, 4*delta].
Page 256, table 9.6, second row of the 8-level gaussian quantizer: The decision boundary should be at 0.5005 and the reconstruction level should be 0.7560
Page 336, formulas 11.36 and 11.37: To get the measures in dB, take the 10-logarithm of the ratios and multiply by 10.
Page 400, formula 13.36: The first element in the matrix should be x00.
Page 402, formula 13.43: The cosine factor for the case when i=0 can be removed, since cos(0)=1.
Page 408, step 4 in the quantization algorithm: Rl should be R1
Page 408, step 4 in the quantization algorithm: After increasing R1, the variance should be divided by 4.
Page 443, figure 14.15: The figure should look like
Page 527, figure 16.12: The caption should be “Frame structure for layer 3.”
