Slicing by Amplitude Envelope in the Time Domain

Amplitude Envelope is often used as a shorthand for a variety of ways that a signal's energy profile can be estimated. Read more about envelopes.

By and large, using amplitude envelopes tends to work best for slicing at an 'event' time-scale, for instance notes or drum hits, as with slicing by onsets. However, if you have a sound with longer passages that are very clearly distinguished by their amplitude, such as periods of relative silence punctuated by activity, it can be useful to get larger slices as well.

Estimating an envelope can be very computationally cheap and often produces quite acceptable results, so is a useful thing to try first. Many automatic slicing tools in DAWs and other software use this method, with the main difference being how much control we are offered about how to decide where to slice. The very simplest approach is to use a single threshold, and when this is crossed a slice point is generated.

However, such a simple approach is only useful in very simple and controlled circumstances. If our sound has overall dynamic changes, it is hard or impossible to determine a single threshold value that works for the whole thing, which can lead to over- or under-detecting slices. Furthermore, the signal can 'dance' around the threshold point, and produce troublesome multiple hits. One way to try and deal with the first problem can be to use two envelopes, one that reacts quickly, and one that reacts more slowly. We can then use a second threshold that compares the activity between these two envelopes to make something that is more robust to signals with changes in dynamics. A way to deal with the second problem can be to impose time limits on how quickly our detector is allowed to change between 'on' and 'off' states around the threshold, similar to a 'hold' control on a noise gate.