問60
頻出度★★☆
ある音をコンピュータのファイルにデータとして記録するとき, 符号化ビット数を 8 ビットとしている。符号化ビット数を 16 ビットに変更し, 同じ音を同じサンプリング周波数で記録したときの説明として, 適切なものはどれか。
平成22年度 春期 ITパスポート試験 問60
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 60
- ア
正解 The correct answer is ア. Digitizing sound is done in the order sampling → quantization → encoding, and the number of encoding bits (quantization bits) decides how many levels of numeric values are used to express the amplitude at each point in time taken by sampling. With 8 bits, 2 to the 8th power = 256 levels can be expressed, and with 16 bits, 2 to the 16th power = 65,536 levels can be expressed, so the scale for the amplitude becomes 256 times finer. The finer the scale is, the smaller the difference from the true amplitude value (quantization error, quantization noise) becomes, so the loudness of the original sound can be reproduced more faithfully. This question does not change the sampling frequency, so the fineness in the time direction stays as it is and only the accuracy in the amplitude direction (the vertical axis) improves; that is the effect of this change. Therefore “The amplitude of the sound can be recorded more finely.” is appropriate.
Incorrectイ
The amount of audio data is obtained roughly as “sampling frequency × number of encoding bits × recording time × number of channels”. If the sampling frequency and the recording time are the same, then doubling the number of encoding bits from 8 bits to 16 bits also increases the amount of data to about twice as much. Saying that it becomes smaller has the direction of the change reversed.
Incorrectウ
Because the data per sample increases from 8 bits to 16 bits, the amount of data handled when processing sound of the same length also becomes about twice as much. Since the data to be processed increases, the load on the CPU moves rather in the direction of increasing, and there is no reason for it to decrease.
Incorrectエ
What decides the upper limit of the frequency that can be recorded is the sampling frequency. According to the sampling theorem, what can be correctly recorded and reproduced goes only up to frequencies below one half of the sampling frequency. In this question the sampling frequency is not changed, so the upper limit does not change. The number of encoding bits is a value that decides the fineness of the amplitude axis, not of the time axis, and it has nothing to do with the upper limit of the frequency.
ここがポイント
- Sampling decides the fineness of the time axis, while quantization and the number of encoding bits decide the fineness of the amplitude axis. Raising the sampling frequency makes the upper limit of the recordable frequency higher (sampling theorem: the upper limit is one half of the sampling frequency), and increasing the number of encoding bits raises the number of amplitude levels to 2 to the power of the bit count, which makes the quantization error smaller. Not mixing up these two roles is the core of this question.
- Amount of audio data = sampling frequency × number of encoding bits × recording time × number of channels. Keep in mind this proportional relationship: operations that raise the sound quality (raising the sampling frequency, increasing the number of encoding bits) always come with an increase in the amount of data, and the processing load does not go down either.
用語