c@63: c@63: QM Vamp Plugins c@63: =============== c@63: c@63: Vamp audio feature extraction plugins from Queen Mary, University of London. c@80: Version 1.5. c@63: c@63: For more information about Vamp plugins, see http://www.vamp-plugins.org/ c@63: and http://www.sonicvisualiser.org/ . c@63: c@63: c@63: To Install c@63: ========== c@63: c@63: This package contains plugins for the Apple OS/X operating system, c@63: compatible with both PPC and Intel hardware. c@63: c@63: To install them, copy the files c@63: c@63: qm-vamp-plugins.dylib and c@63: qm-vamp-plugins.cat c@63: c@63: to either c@63: c@63: /Library/Audio/Plug-Ins/Vamp/ (for plugins available to all users) or c@63: $HOME/Library/Audio/Plug-Ins/Vamp/ (for plugins available to you only). c@63: c@63: c@63: License c@63: ======= c@63: c@63: These plugins are provided in binary form only. You may install and c@63: use the plugin binaries without fee for any purpose commercial or c@63: non-commercial. You may redistribute the plugin binaries provided you c@63: do so without fee and you retain this README file with your c@63: distribution. You may not bundle these plugins with a commercial c@63: product or distribute them on commercial terms. If you wish to c@63: arrange commercial licensing terms, please contact the Centre for c@63: Digital Music at Queen Mary, University of London. c@63: c@63: Copyright (c) 2006-2008 Queen Mary, University of London. All rights c@63: reserved except as described above. c@63: c@63: c@80: About This Release c@80: ================== c@63: c@80: This is a bugfix release only. The plugins provided are unchanged c@80: from 1.4. c@63: c@63: c@63: Plugins Included c@63: ================ c@63: c@63: This plugin set includes the following plugins: c@63: c@63: * Note onset detector c@63: c@63: * Beat tracker and tempo estimator c@63: c@63: * Key estimator and tonal change detector c@63: c@63: * Segmenter, to divide a track into a consistent sequence of segments c@63: c@63: * Timbral and rhythmic similarity between audio tracks c@63: c@63: * Chromagram, constant-Q spectrogram, and MFCC calculation plugins c@63: c@63: More details about the plugins follow. c@63: c@63: c@63: Note Onset Detector c@63: ------------------- c@63: c@63: Identifier: qm-onsetdetector c@63: Authors: Chris Duxbury, Juan Pablo Bello and Christian Landone c@63: Category: Time > Onsets c@63: c@63: References: C. Duxbury, J. P. Bello, M. Davies and M. Sandler. c@63: Complex domain Onset Detection for Musical Signals. c@63: In Proceedings of the 6th Conference on Digital Audio c@63: Effects (DAFx-03). London, UK. September 2003. c@63: c@63: D. Stowell and M. D. Plumbley. c@63: Adaptive whitening for improved real-time audio onset c@63: detection. c@63: In Proceedings of the International Computer Music c@63: Conference (ICMC'07), August 2007. c@63: c@63: D. Barry, D. Fitzgerald, E. Coyle and B. Lawlor. c@63: Drum Source Separation using Percussive Feature c@63: Detection and Spectral Modulation. c@63: ISSC 2005 c@63: c@63: The Note Onset Detector plugin analyses a single channel of audio and c@63: estimates the locations of note onsets within the music. c@63: c@63: It calculates an onset likelihood function for each spectral frame, c@63: and picks peaks in a smoothed version of this function. The plugin is c@63: non-causal, returning all results at the end of processing. c@63: c@63: It has three outputs: the note onset positions, the onset detection c@63: function used in estimating onset positions, and a smoothed version of c@63: the detection function that is used in the peak-picking phase. c@63: c@63: c@63: Tempo and Beat Tracker c@63: ---------------------- c@63: c@63: Identifier: qm-tempotracker c@63: Authors: Matthew Davies and Christian Landone c@63: Category: Time > Tempo c@63: c@63: References: M. E. P. Davies and M. D. Plumbley. c@63: Context-dependent beat tracking of musical audio. c@63: In IEEE Transactions on Audio, Speech and Language c@63: Processing. Vol. 15, No. 3, pp1009-1020, 2007. c@63: c@63: M. E. P. Davies and M. D. Plumbley. c@63: Beat Tracking With A Two State Model. c@63: In Proceedings of the IEEE International Conference c@63: on Acoustics, Speech and Signal Processing (ICASSP 2005), c@63: Vol. 3, pp241-244 Philadelphia, USA, March 19-23, 2005. c@63: c@63: The Tempo and Beat Tracker plugin analyses a single channel of audio c@63: and estimates the locations of metrical beats and the resulting tempo. c@63: c@63: It has three outputs: the beat positions, an ongoing estimate of tempo c@63: where available, and the onset detection function used in estimating c@63: beat positions. c@63: c@63: c@63: Key Detector c@63: ------------ c@63: c@63: Identifier: qm-keydetector c@63: Authors: Katy Noland and Christian Landone c@63: Category: Key and Tonality c@63: c@63: References: K. Noland and M. Sandler. c@63: Signal Processing Parameters for Tonality Estimation. c@63: In Proceedings of Audio Engineering Society 122nd c@63: Convention, Vienna, 2007. c@63: c@63: The Key Detector plugin analyses a single channel of audio and c@63: continuously estimates the key of the music. c@63: c@63: It has four outputs: the tonic pitch of the key; a major or minor mode c@63: flag; the key (combining the tonic and major/minor into a single c@63: value); and a key strength plot which reports the degree to which the c@63: chroma vector extracted from each input block correlates to the stored c@63: key profiles for each major and minor key. The key profiles are drawn c@63: from analysis of Book I of the Well Tempered Klavier by J S Bach, c@63: recorded at A=440 equal temperament. c@63: c@63: The outputs have the values: c@63: c@63: Tonic pitch: C = 1, C#/Db = 2, ..., B = 12 c@63: c@63: Major/minor mode: major = 0, minor = 1 c@63: c@63: Key: C major = 1, C#/Db major = 2, ..., B major = 12 c@63: C minor = 13, C#/Db minor = 14, ..., B minor = 24 c@63: c@63: Key Strength Plot: 25 separate bins per feature, separated into 1-12 c@63: (major from C) and 14-25 (minor from C). Bin 13 is unused, not c@63: for superstitious reasons but simply so as to delimit the major c@63: and minor areas if they are displayed on a single plot by the c@63: plugin host. Higher bin values show increased correlation with c@63: the key profile for that key. c@63: c@63: The outputs are also labelled with pitch or key as text. c@63: c@63: c@63: Tonal Change c@63: ------------ c@63: c@63: Identifier: qm-tonalchange c@63: Authors: Chris Harte and Martin Gasser c@63: Category: Key and Tonality c@63: c@63: References: C. A. Harte, M. Gasser, and M. Sandler. c@63: Detecting harmonic change in musical audio. c@63: In Proceedings of the 1st ACM workshop on Audio and Music c@63: Computing Multimedia, Santa Barbara, 2006. c@63: c@63: C. A. Harte and M. Sandler. c@63: Automatic chord identification using a quantised chromagram. c@63: In Proceedings of the 118th Convention of the Audio c@63: Engineering Society, Barcelona, Spain, May 28-31 2005. c@63: c@63: The Tonal Change plugin analyses a single channel of audio, detecting c@63: harmonic changes such as chord boundaries. c@63: c@63: It has three outputs: a representation of the musical content in a c@63: six-dimensional tonal space onto which the algorithm maps 12-bin c@63: chroma vectors extracted from the audio; a function representing the c@63: estimated likelihood of a tonal change occurring in each spectral c@63: frame; and the resulting estimated positions of tonal changes. c@63: c@63: c@63: Segmenter c@63: --------- c@63: c@63: Identifier: qm-segmenter c@63: Authors: Mark Levy c@63: Category: Classification c@63: c@63: References: M. Levy and M. Sandler. c@63: Structural segmentation of musical audio by constrained c@63: clustering. c@63: IEEE Transactions on Audio, Speech, and Language Processing, c@63: February 2008. c@63: c@63: The Segmenter plugin divides a single channel of music up into c@63: structurally consistent segments. Its single output contains a c@63: numeric value (the segment type) for each moment at which a new c@63: segment starts. c@63: c@63: For music with clearly tonally distinguishable sections such as verse, c@63: chorus, etc., the segments with the same type may be expected to be c@63: similar to one another in some structural sense (e.g. repetitions of c@63: the chorus). c@63: c@63: The type of feature used in segmentation can be selected using the c@63: Feature Type parameter. The default Hybrid (Constant-Q) is generally c@63: effective for modern studio recordings, while the Chromatic option may c@63: be preferable for live, acoustic, or older recordings, in which c@63: repeated sections may be less consistent in sound. Also available is c@63: a timbral (MFCC) feature, which is more likely to result in c@63: classification by instrumentation rather than musical content. c@63: c@63: Note that this plugin does a substantial amount of processing after c@63: receiving all of the input audio data, before it produces any results. c@63: c@63: c@63: Similarity c@63: ---------- c@63: c@63: Identifier: qm-similarity c@63: Authors: Mark Levy, Kurt Jacobson and Chris Cannam c@63: Category: Classification c@63: c@63: References: M. Levy and M. Sandler. c@63: Lightweight measures for timbral similarity of musical audio. c@63: In Proceedings of the 1st ACM workshop on Audio and Music c@63: Computing Multimedia, Santa Barbara, 2006. c@63: c@63: K. Jacobson. c@63: A Multifaceted Approach to Music Similarity. c@63: In Proceedings of the Seventh International Conference on c@63: Music Information Retrieval (ISMIR), 2006. c@63: c@63: The Similarity plugin treats each channel of its audio input as a c@63: separate "track", and estimates how similar the tracks are to one c@63: another using a selectable similarity measure. c@63: c@63: The plugin also returns the intermediate data used as a basis of the c@63: similarity measure; it can therefore be used on a single channel of c@63: input (with the resulting intermediate data then being applied in some c@63: other similarity or clustering algorithm, for example) if desired, as c@63: well as with multiple inputs. c@63: c@63: The underlying audio features used for the similarity measure can be c@63: selected using the Feature Type parameter. The available features are c@63: Timbre (in which the distance between tracks is a symmetrised c@63: Kullback-Leibler divergence between Gaussian-modelled MFCC means and c@63: variances across each track); Chroma (KL divergence of mean chroma c@63: histogram); Rhythm (cosine distance between "beat spectrum" measures c@63: derived from a short sampled section of the track); and combined c@63: "Timbre and Rhythm" and "Chroma and Rhythm". c@63: c@63: The plugin has six outputs: a matrix of the distances between input c@63: channels; a vector containing the distances between the first input c@63: channel and each of the input channels; a pair of vectors containing c@63: the indices of the input channels in the order of their similarity to c@63: the first input channel, and the distances between the first input c@63: channel and each of those channels; the means of the underlying c@63: feature bins (MFCCs or chroma); the variances of the underlying c@63: feature bins; and the beat spectra used for the rhythmic feature. c@63: c@63: Because Vamp does not have the capability to return features in matrix c@63: form explicitly, the matrix output is returned as a series of vector c@63: features timestamped at one-second intervals. Likewise, the c@63: underlying feature outputs contain one vector feature per input c@63: channel, timestamped at one-second intervals (so the feature for the c@63: first channel is at time 0, and so on). Examining the features that c@63: the plugin actually returns, when run on some test data, may make this c@63: arrangement more clear. c@63: c@63: Note that the underlying feature values are only returned if the c@63: relevant feature type is selected. That is, the means and variances c@63: outputs are valid provided the pure rhythm feature is not selected; c@63: the beat spectra output is valid provided rhythm is included in the c@63: selected feature type. c@63: c@63: c@63: Constant-Q Spectrogram c@63: ---------------------- c@63: c@63: Identifier: qm-constantq c@63: Authors: Christian Landone c@63: Category: Visualisation c@63: c@63: References: J. Brown. c@63: Calculation of a constant Q spectral transform. c@63: Journal of the Acoustical Society of America, 89(1): c@63: 425-434, 1991. c@63: c@63: The Constant-Q Spectrogram plugin calculates a spectrogram based on a c@63: short-time windowed constant Q spectral transform. This is a c@63: spectrogram in which the ratio of centre frequency to resolution is c@63: constant for each frequency bin. The frequency bins correspond to the c@63: frequencies of "musical notes" rather than being linearly spaced in c@63: frequency as they are for the conventional DFT spectrogram. c@63: c@63: The pitch range and the number of frequency bins per octave may be c@63: adjusted using the plugin's parameters. Note that the plugin's c@63: preferred step and block sizes depend on these parameters, and the c@63: plugin will not accept any other block size. c@63: c@63: c@63: Chromagram c@63: ---------- c@63: c@63: Identifier: qm-chromagram c@63: Authors: Christian Landone c@63: Category: Visualisation c@63: c@63: The Chromagram plugin calculates a constant Q spectral transform (as c@63: above) and then wraps the frequency bin values into a single octave, c@63: with each bin containing the sum of the magnitudes from the c@63: corresponding bin in all octaves. The number of values in each c@63: feature vector returned by the plugin is therefore the same as the c@63: number of bins per octave configured for the underlying constant Q c@63: transform. c@63: c@63: The pitch range and the number of frequency bins per octave for the c@63: transform may be adjusted using the plugin's parameters. Note that c@63: the plugin's preferred step and block sizes depend on these c@63: parameters, and the plugin will not accept any other block size. c@63: c@63: c@63: Mel-Frequency Cepstral Coefficients c@63: ----------------------------------- c@63: c@63: Identifier: qm-mfcc c@63: Authors: Nicolas Chetry and Chris Cannam c@63: Category: Low Level Features c@63: c@63: References: B. Logan. c@63: Mel-Frequency Cepstral Coefficients for Music Modeling. c@63: In Proceedings of the First International Symposium on Music c@63: Information Retrieval (ISMIR), 2000. c@63: c@63: The Mel-Frequency Cepstral Coefficients plugin calculates MFCCs from a c@63: single channel of audio, returning one MFCC vector from each process c@63: call. It also returns the overall means of the coefficient values c@63: across the length of the audio input, as a separate output at the end c@63: of processing. c@63: