問66
頻出度★★☆
アナログ音声信号をディジタル化する場合, 元のアナログ信号の波形に, より近い波形を復元できる組合せはどれか。
平成21年度 春期 ITパスポート試験 問66
テクノロジ系基礎理論 › 情報に関する理論
This question's explanation and your answers so far.
解説
問 66
Incorrectア
Having many quantization levels is correct as accuracy along the amplitude axis, but because the sampling period is long, the spacing between sample points is wide, and changes in the waveform that occur between one sample and the next are not recorded. No matter how finely the value of each individual point is expressed, if the points that are picked up are themselves coarse, the waveform is blunted along the time axis, and a reconstruction close to the original waveform is not possible.
Incorrectイ
Not only is the sampling period long and the sample points few, but the number of quantization levels is also small, so the amplitude steps are coarse. Both the time-axis condition and the amplitude-axis condition are on the side that lowers accuracy, so among the four combinations this gives the reconstruction that is farthest from the original waveform, in other words the one with the lowest sound quality. The data volume is the smallest, but from the standpoint of the fidelity the question asks about, it is the opposite of what is wanted.
- ウ
正解 The correct answer is ウ. Digitizing an analog audio signal (A/D conversion) is carried out in the order sampling → quantization → encoding. Sampling is the processing along the time axis: the value of the analog waveform is read at fixed intervals. The shorter this sampling period is, the more sample points are read within the same amount of time (that is, the higher the sampling frequency), so even the fine up-and-down movements of the waveform can be captured without being missed. Quantization is the processing along the amplitude (loudness) axis: each value that was read is divided into levels and rounded to the nearest level. The more quantization levels there are, the narrower each single level becomes, and the smaller the deviation caused by rounding (the quantization error) becomes. In other words, the condition that makes the time axis finer is a short sampling period, and the condition that makes the amplitude axis finer is a large number of quantization levels; to reconstruct a waveform closer to the original analog waveform, both must be satisfied at the same time. The only combination in the table that matches this is ウ (short × many), so ウ is the correct answer. Keep in mind, though, that the higher the accuracy, the larger the data volume also becomes.
Incorrectエ
Having a short sampling period is correct as accuracy along the time axis, but because the number of quantization levels is small, the amplitude can only be rounded to coarse levels. Even if the samples are taken finely, the deviation of each individual value (the quantization error) becomes large and the reconstructed waveform breaks down into a staircase shape, so it cannot be said to be closer to the original waveform.
ここがポイント
- A/D conversion has three stages: sampling → quantization → encoding. Sampling divides up the time axis while quantization divides up the amplitude axis; remember them as distinct, since each determines the accuracy along a different axis.
- The conditions for getting closer to the original waveform are a short sampling period (that is, a high sampling frequency) and many quantization levels (that is, a large number of quantization bits). There is a trade-off, however: the more the accuracy of either one is raised, the larger the data volume becomes.
用語