aoTuV_technical.txt 5.28 KB
Newer Older
j's avatar
j committed
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27
aoTuV beta5.5 technical information


The differences from the aoTuV beta 5...

  1.  The frequency domain width of M6 was revised.

  2.  For q-1/-2, a pre-echo reduction code was applied (M3).  In addition,
  the M3 code was improved.

  3.  The floor setup parameters in the low bit-rate was changed.

  4.  The part including bug of noise normalization was rewritten.

  5.  libvorbis 1.2.0 were merged. Furthermore, a revision of Bug #300 and
  #1229 was applied.

  6.  The ATH curve of the high frequency area (more than 32kHz) was revised. 

...and I tune up many parameters. 


2008/03/30
Aoyumi

----------------------------------------------------------------------------

j's avatar
j committed
28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46
aoTuV beta5 technical information


The differences from the aoTuV beta 4.51...

  1.   In order to work noise normalization correctly, the value of tone 
  masking is compressed under certain conditions. A floor curve approaches 
  the original audio spectrum line as a result.  [New M4] 

  2.   In order to cut down the bitrate, a part of M2 code of beta4 was 
  deleted. 

@3.   The ATH curve was changed slightly. 

  4.   Pre-echo decreased slightly. This is based on change of M3 code. 

  5.   In order to decrease artifact by collapse of audio energy, noise 
  normalization processing of point stereo was extended. [32/44.1/48kHz only] 

j's avatar
j committed
47 48
  6.   The threshold of channel coupling (lossless <=> point) is changed 
  dynamically. This is a method based on a certain idea. [M6] 
j's avatar
j committed
49 50 51 52 53 54 55 56 57

...and I tune up many parameters. 


2006/10/24
Aoyumi

----------------------------------------------------------------------------

j's avatar
j committed
58 59 60 61 62
aoTuV beta4.51 technical information


The differences from the aoTuV beta 4.5...

j's avatar
j committed
63 64
  1.  The changed part of "scale.h" of Xiph.Org SVN was reflected in aoTuV. 
  This will solve the problem of the compile by gcc4. 
j's avatar
j committed
65

j's avatar
j committed
66 67 68
  2.  A part of portion added by beta4.5 had caused memory access violation. 
  Depending on environment, a problem arises in qualiy-1/-2. This problem was 
  corrected. 
j's avatar
j committed
69 70 71 72 73 74 75 76 77 78 79 80


2005/11/17
Aoyumi

----------------------------------------------------------------------------

aoTuV beta4.5 technical information


The differences from the aoTuV beta 4...

j's avatar
j committed
81 82 83 84
  1.  M4 code of beta3 was deleted. M5 is added as what is replaced with it. 
  The advantage of this method is that the result stabilized more is obtained. 
  Instead, the change width of the bit rate becomes large rather than before. 
  [32/44.1/48kHz only] 
j's avatar
j committed
85

j's avatar
j committed
86 87 88
  2.  New M4 was added. There is this for the same purpose as M1. However, 
  it is tuned up only in the low bit rate region according to work and 
  individual parameters. [32/44.1/48kHz only] 
j's avatar
j committed
89

j's avatar
j committed
90 91
  3.  M2 code was extended. This decreases a specific noise problem.  
  [32/44.1/48kHz only] 
j's avatar
j committed
92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172

...and I mainly tune up tone/noise masking and noise normalization parameters. 

The above change is applied below quality3. 


2005/11/05
Aoyumi

----------------------------------------------------------------------------

aoTuV beta4 technical information


The differences from the aoTuV beta 3...

  1.  Adoption of new hypot (point stereo). This works in general better than
  dipole/round hypot. And when residue is likely to decrease greatly at the
  time of a reverse phase, added code is made to increase the ratio of lossless.

  2.  Change of ATH curve. This is not a big change. 

  3.  Residue used for threshold calculation of channel coupling is changed.
  This one seems to be good selection. 

  4.  HF reduction code of beta3 was deleted. Now, this is not required. 

  5.  The energy deficit of M4 code of beta3 -- correction of a bug Now.

  6.  The code for pulling downward floor of the frequency range which can
  be easy to be heard was added (New M2). This works by the low bit rate and
  has a comparatively good trade-off. 

  7.  Change of threshold calculation of noise normalization. Moreover, it
  takes phase information into consideration. 


...and I mainly tune up tone/noise masking and noise compander parameters. 


2005/06/18
Aoyumi

----------------------------------------------------------------------------

aoTuV beta3 technical information


The differences from the aoTuV beta 2 and Vorbis 1.1...

  1. Dynamic impulse block noise control(q0-10). [32/44.1/48kHz only] 
  Conspicuous pre-echo is reduced.

  2. New ATH curve. 

  3. In order to cancel that a small sound becomes unstable, the noise level
  of low frequency is lowered under specific conditions. (Only below q3) 
  There is a greatest effect on a piano etc. [32/44.1/48kHz only]

  4. The rate of lossless stereo is increased in specific conditions. This
  improves many of conspicuous problems which a point stereo causes. (e.g 
  drumus's hiss/voice, guitar and brass's distortion) [32/44.1/48kHz only]

  5. The action of the noise normalization of point stereo is corrected. It
  was set to one of the causes of distortion in a long tone (e.g 
  strings/oboe/high tone voice). 

  6. The HF reduction code was not working normally. This is corrected and
  it retuned up according to the present condition. [32/44.1/48kHz only]

  7. The q-2 mode was added in all sampling frequencies. And nominal bitrate
  of q-1 are changed into a more nearly average value. Although nominal 
  bitrate is more large at a low sampling frequency, average bitrate hardly
  differs from the former version. 


...and I mainly tune up tone/noise masking parameters. 


2004/11/21
Aoyumi