MP3 Digital Sound

MP3 stands for MPEG 1 Layer 3. MPEG is a compression type for digital data. MP3 is a variation of this format that allows sound files to be compressed by 90 percent without major degradation of the quality of the sound. The compressed audio file takes up so much less storage space than on a regular compact disc or tape that it has become very convenient for transfer on the Internet.

MP3 是 MPEG 1 层 3 的缩写。MPEG 是一种数字数据的压缩类型。MP3 是这种格式的一种变体，它允许声音文件在不显著降低音质的情况下压缩到原来的 90%。压缩后的音频文件占用的存储空间远小于普通光盘或磁带，因此在因特网上传输变得非常方便。

The real possibility for sound compression without audible loss comes from the fact that the sampling for CDs contains far more than the necessary data. Sixteen-bit digital sampling at 44.1 kHz gives you a staggering amount of information. From the audio CD you get about 1.4 million bits per second of information, much more information than your ears can process. To create the MP3 signal, the information from the CD format is divided into frequency subbands. Then the signal in each subband is examined in the process of encoding to decide how many bits to allocate to it. The process employs a "psychoacoustic model" to decide which subbands will be recorded most accurately and which will be discriminated against. The idea is that only that which can realistically be heard by the ear is kept.

声音无损压缩的现实可能性来自于CD采样中包含的数据远多于必要数据。在44.1 kHz下以16位进行采样，可以得到大量信息。从音频CD中可以得到每秒约1400万位的信息，这比人耳能处理的信息量要多得多。为了生成MP3信号，CD格式的信息被分成频率子带。在编码过程中，每个子带的信号会被检查，以决定分配给它的位数。这个过程采用“心理听觉模型”来决定哪些子带会被准确记录，哪些子带会被忽略。其想法是只保留那些实际上能被耳朵听到的内容。

The favorite visual metaphor is the "polar bear in the snow storm". Against a dark mountain on a clear day, you would have to paint the polar bear with great definition. But if the polar bear is in a snowstorm, you don't have to provide as much detail, because you are not going to see much detail anyway. By analogy, if a sound in a particular subband is going to be masked out by other subbands so that you won't hear it anyway, you might as well save the bits you were going to use to record it. The "psychoacoustic model" makes judgements about which sounds were going to be masked out.

最常用的视觉隐喻是‘雪中北极熊’。在晴朗的天空下，如果背景是黑色的山峰，你必须用鲜明的笔触来描绘北极熊。但如果北极熊处于一场雪暴中，你就不需要画那么多细节，因为无论如何你都看不到太多细节。类比地，如果某个特定频段的声音会被其他频段掩盖，以至于你根本听不到，那么你就可以省略用来记录它的位元。‘心理听觉模型’会判断哪些声音会被其他声音掩盖。

Some model is applied in the encoding of the high-resolution digital sound image to MP3, and that model is inevitably going to take out some audible information. You can improve the model by encoding at a higher bit rate, because you are putting in more information. Typical current bit rates are 128, 160, 192, 256 and 320 kbps. Tests show that the accuracy increases significantly up to 256 kbps with some current decoders, so 256 kbps is perhaps a good comparison standard. At 256,000 bits of information per second, you have reduced the 1.4 Mbps to about 18% - compression by more than five to one. Of course you can get ten to one at 128 kbps, but you can't expect to get it without noticeable loss of sound quality.

在将高分辨率数字声音图像编码为MP3时，应用了某种模型，该模型不可避免地会丢失一些可听信息。你可以通过提高编码的比特率来改进该模型，因为你在输入更多的信息。当前典型的比特率是128, 160, 192, 256和320 kbps。测试显示，使用某些当前解码器时，准确性在256 kbps时显著提高，因此256 kbps可能是一个良好的比较标准。在每秒256,000比特的信息量下，你将1.4 Mbps压缩到约18%，压缩比超过五倍。当然，你可以在128 kbps时得到十倍的压缩，但你不能期望在没有明显声音质量损失的情况下实现这一点。