As shown in FIG. 17B, the speech signals 1706 are represented as a continuously varying acoustic waveform that can be transmitted, recorded, and manipulated. In one embodiment, the speech signals 1706 may comprise acoustic samples, e.g., a discrete signal x[n], i.e., a sequence of real or integer numbers corresponding to the signal samples sampled uniformly at a sampling rate. In one embodiment, the sampling rate for the speech signals 1706 may be 16 kHz, while in other embodiments, the sampling rate (e.g., for “telephone speech” coding) may be 8 kHz. In one embodiment, the speech signals 1706 may be represented as a number of amplitude-altered signals representing the outputs of an array of narrow frequency bands (“Fi”). The speech signals 1706 include high-frequency components that are altered by a lower frequency, slowly evolving temporal envelope 1728 as shown in FIG. 17B.
The temporal envelope 1728 shown in FIG. 17B represents changes in the amplitude of the speech signals 1706. The temporal envelope 1728 is the slow variation in the amplitude (e.g., 1732) of the speech signals 1706 over time (e.g., 1730). Temporal information at the output of each frequency band Fi of the speech signals 1706 may be separated into temporal fine structure (the rapid oscillations close to the center frequency) and the temporal envelope 1728 (the slower amplitude changes) shown in FIG. 17B. The temporal envelope 1728 evolves more slowly in time than the temporal fine structure and requires a lower sampling rate to be represented. The temporal envelope 1728 is a one-dimensional signal and can be represented with a single time-varying waveform.
