Music Analysis

An extensive music analysis framework was developed as part of MPS, which facilitates statistical and musicological analyses.

There are a number of motivations for the development of the MPS analysis tool:

• Due to the design of the underlying context-based model, new analysis methods can be developed which benefit from the available context information. Not only explicit information, but also implicit and contextual data can be taken into account and relations between individual context layers can be explored easily.
• Once certain analysis algorithms are developed, they can be applied not only to one musical piece, but also easily to hundreds or thousands of pieces. In this way, interesting results can be obtained by using the advantages of computers. In this way, human researches can be disburdened from laborious manual work.
• By using the processing capabilities of computers, statistical commonalities and distinctive features of musical compositions can be explored.
• MPS does not require the knowledge of programming skills. Users only have to provide MIDI and/or MusicXML files. The system produces output files in CSV format which can be opened with any regular spreadsheet application.
• MPS also generates comprehensible graphical plots and charts from the analysis results, which can be viewed directly in MPS or can be exported in the form of PDF analysis reports.
• By interpreting musical analysis results, insights can be gained into characteristic properties of individual pieces, composers and styles.

Analysis Scopes

To analyze musical compositions, one or more files containing symbolic music data in MIDI or MusicXML format have to be provided. The analysis and export process is configurable. Different analysis scopes are possible. Either a single piece, a collection of pieces or a corpus of pieces can be analyzed.

The first musical analysis scope to be introduced is the single piece scope. It is used for gaining insights about a specific musical composition by performing either a global analysis or a voice-specific analysis of the piece. It is also possible to combine the aforementioned analysis modes. Furthermore musically meaningful sections can manually be marked in the score in order to perform section-wise analysis, which is optionally combinable with voice-wise analysis.

Collection analysis is suitable for comparing multiple musical pieces among each other. For this purpose, analysis results are shown next to each other in combined representations. Corpus analysis goes even one more step further, allowing to compare multiple collections of compositions, e.g. folders containing multiple compositions of individual composers.

Analyzing Music

To start analyzing music, simply select a file or folder and click the button in the toolbar. Another possibility is invoking the context menu by using a right click and choosing Run As → MPS Analysis Run. If a file or folder is analyzed for the first time, a corresponding launch configuration will be opened. Adjust the configuration as desired and then click Apply and/or Run to start the analysis process. See section Configuring Music Analysis for more details.

The progress of the analysis process will be shown in the Progress view, which is located at the bottom of the application by default.

Folders with the name schema <analyzed file or folder>_Analysis will be created for each analyzed file or folder. The results will be stored in these folders in the form of CSV files. These contain column names in the first line/row and analysis result values in the following lines/rows.

Exploring Analysis Results

There are multiple options for viewing and exploring analysis results. One option is to view the raw analysis data directly or in a spreadsheet editor. To view raw analysis data, follow these steps:

• To view a CSV file in raw text format, right-click on a CSV file and choose Open With → Text Editor.
• To view a CSV file in a spreadsheet editor such as Excel, OpenOffice or Numbers, select Open With → System Editor. This will open the application which is associated with the .csv file extension in your operating system.

A more comfortable option is to use the Analysis Result Browser provided by MPS. To open it, right-click an analysis folder and and choose Open Analysis View from the context menu:

The analysis result view will be shown. To initialize the view with data, select an arbitrary folder containing analysis results. Once the view is associated with a data set, CSV files can be selected on the left hand side of the view. MPS will display the raw CSV data and a corresponding plot or graph depending on the content of the selected CSV file.

The generated charts can be exported as file. To export a chart, right-click on the graphic and select Export Chart to File. After specifying the destination, the file will be generated. Currently, the following image formats are supported:

• BMP
• JPEG
• PNG

Configuring Music Analysis

The analysis process can be configured by adjusting a so called run configuration. A run configuration is automatically created when a file or folder is initially analyzed. To change run configurations later, click the black arrow next to the button in the toolbar and click Run Configurations.... The following window will open:

You can either search for an existing run configuration to modify or create a new configuration by clicking the button . The following parameters can be set:

The figure shows the harmonic progression graph of Ludwig van Beethoven’s Piano Sonata No. 14 in C# minor, first movement. The numbers specify the measures in which the corresponding chord change was detected. The colors of the chords encode the consonance or dissonance of the relevant chord (green corresponds to consonant and red to dissonant). The graph reveals which harmonic progressions are only used once and which are used multiple times. The latter are easily identifiable due to comma-separated enumerations of measure numbers in which the corresponding transitions occur.

Analysis launch configurations can be reused, edited and deleted at any time. Refer to section Launch Configurations for further details.

Generating Analysis PDF Reports

Analysis results can also be exported in the form of PDF reports. To invoke a report generation, simply select an analysis data folder of your choice and click the button in the toolbar. Note that you must have a LaTeX environment installed to render analysis reports, which is described in sections LaTeX Installation and LaTeX Configuration in the installation chapter. The PDF will be generated in a directory next the the analysis folder with the name <Analysis Folder Name>_Report.

Analysis Report PDF Settings

To configure the contents of analysis report PDFs, navigate the the corresponding preference page:

1. Open the MPS Preferences
   1. On Windows and Linux: In the menu bar, choose Window → Preferences
   2. On Mac: In the menu bar, choose Music Processing Suite (application menu) → Preferences
2. Navigate to Music Processing Suite → Analysis Reports

The following settings are available:

Generating Progression Graphs

When analyzing music with MPS, progression graphs of the following musical aspects are generated:

• Harmonic Progressions
• Rhythmic Progressions
• Pitch Progressions
• Lyric Progressions

Depending on the setting Progression Graph Label Mode described in the previous section, the edges in the resulting graphs will either have

• Occurrence Counts
• Measure Numbers or
• Markov Model Probabilities

as labels. An example with measure numbers was already presented in the previous section. Here is a harmonic progression graph of Hey Jude by the Beatles with occurrence count labels visualizing how often the corresponding transition was found in the piece:

And this is the same harmonic progression graph as a Markov model with probability labels:

The graphs are exported for each stream separately and can be found in the respective subfolders:

Corresponding PDF files will be generated automatically if GraphViz is installed and configured as described in GraphViz Installation.

If the graphs should be included in analysis report PDF files, the option Export Stream-specific Analysis Results needs to be activated in the preferences as shown in the previous section.

Analysis Features

The following table lists all 117 musical features which can be analyzed with MPS.