Any processes of reality can be digitized. So, the encoding of audio information using computers is performed as follows:
- air vibrations are recorded by sensitive instruments;
- they are converted into an electric current in which the frequency (amplitude) changes accordingly;
- the received current is digitized, that is, it is sampled (it is sometimes said that binary coding of audio information takes place).
The resulting electronic analogue of the original sound stream the better, the higher the sampling frequency at sampling and the coding depth.
In other words, encoding audio information is the process of converting a familiar analogy to a digital signal for many, which is intended for further processing on the corresponding devices. Let's consider in more detail the stages and methods of digitizing sound.
Time frame sampling is the foundation of digitization. According to Kotelnikov’s theorem, an analogue electrical signal can be converted to digital form by reading with a certain step a continuous series of values of its amplitude. The frequency of such readings should be at least twice the limit value of the frequency of the main signal. If it is necessary to digitize an analog “source” with an operating frequency of 0-20 kHz, sampling should be carried out at least 40 thousand times per second (40 kHz). Sampling indicates the number of measurements per second of the original analog signal (sampling, sampling frequency). With the growth of samples, not only the quality but also the volume of the received data stream increases.
Also, encoding of audio information may be performed in other ways. As, for example, digitization by means of non-uniform quantization, sometimes called logarithmic. When using it, the entire amplitude scale is conditionally divided into sections with high and low values. Further coding of audio information occurs by applying a large number of quantization levels in areas with a small amplitude value (and vice versa). However, we note that the total number of levels remains the same as in the homogeneous quantization method (PCM).
A completely different approach is implemented in an alternative coding method. It is called Differential Pulse Code Modulation (DPCM). With this method, quantization is not subjected to the direct amplitude of the signal, but its relative values. As a result, it is possible to achieve a reduction in the volume occupied by the data, since the mechanism for predicting subsequent samples of the original signal is activated.
The coding and processing of audio information described in this paper implies the need to perform an analog-to-digital conversion. This process is carried out using an ADC (analog-to-digital converter). Every owner of a computer equipped with a sound card is faced with the operation of this device every day (in this case, the reverse process takes place - receiving an analog signal from a digital stream).
The functions of the ADC are as follows:
- In limiting the bandwidth of passable frequencies. Using filters, the signal components are cut off, the frequency of which is more than half of the sampling frequency (the reason was described earlier).
- Sampling of amplitude values at certain intervals. As a result, the analog signal is represented by a sequence of unit discharges of various intensities (sampling).
- Replacing the values of the received digits with their nearest values from a fixed set (quantization).
- Conversion of each quantized value by the conditional number of the quantization level (each value has its own serial number). This is the last stage of digitization.