CosmoNote can play audio recordings of piano music interpretations, show a map of the notes pitches in time, display when the sustain pedal was pressed, and more. This is because CosmoNote uses time series data as input.
This optional lesson gives some examples of other types of data amenable to the same kinds of treatment in the hopes that Citizen Science projects in other fields might be inspired by our approach.
A time series is a collection of data points that are organized according to usually equally spaced points in time. This type of data is said to be discrete because there is a finite number of values that are represented (e.g., integers from 1 to 10). The following figure presents an example of aligned time-series data from an audio recording and corresponding musical features. Even though the curves have a different number of individual points (samples), aligning them allows us to see a relationship between them.
Examples of Time-series data from a musical excerpt of the Chopin Ballade Nº2. a. Tempo b. Loudness c. Audio recording d. Close-up of audio data samples
Different mathematical techniques can help us understand the properties of time-based data. Statistical analysis can provide meaningful information to identify patterns and make inferences about the data. Forecasting can be used to predict future data points based on the previous data.
Numerous techniques exist to analyze time series data depending on its characteristics. The following figure shows a non-exhaustive list of methods used to perform this analysis.
Classification of time-series analysis methods
We will briefly mention one important time-series problem that is relevant to music structures annotations: