peterf@2: Changes
peterf@2: =======
peterf@2: 
peterf@2: ##v0.3.1
peterf@2: 
peterf@2: Bug fixes
peterf@2:     
peterf@2:   - Fixed bug #117: `librosa.segment.agglomerative` now returns a numpy.ndarray instead of a list
peterf@2:   - Fixed bug #115: off-by-one error in `librosa.core.load` with fixed duration
peterf@2:   - Fixed numerical underflow errors in `librosa.decompose.hpss`
peterf@2:   - Fixed bug #104: `librosa.decompose.hpss` failed with silent, complex-valued input
peterf@2:   - Fixed bug #103: `librosa.feature.estimate_tuning` fails when no bins exceed the threshold
peterf@2: 
peterf@2: Features
peterf@2: 
peterf@2:   - New function `librosa.core.get_duration()` computes the duration of an audio signal 
peterf@2:     or spectrogram-like input matrix
peterf@2:   - `librosa.util.pad_center` now accepts multi-dimensional input
peterf@2: 
peterf@2: Other changes
peterf@2: 
peterf@2:   - Adopted the ISC license
peterf@2:   - Python 3 compatibility via futurize
peterf@2:   - Fixed issue #102: segment.agglomerative no longer depends on the deprecated 
peterf@2:     Ward module of sklearn; it now depends on the newer Agglomerative module.
peterf@2:   - Issue #108: set character encoding on all source files
peterf@2:   - Added dtype persistence for resample, stft, istft, and effects functions
peterf@2: 
peterf@2: ##v0.3.0
peterf@2: 
peterf@2: Bug fixes
peterf@2: 
peterf@2:   - Fixed numpy array indices to force integer values
peterf@2:   - `librosa.util.frame` now warns if the input data is non-contiguous
peterf@2:   - Fixed a formatting error in `librosa.display.time_ticks()`
peterf@2:   - Added a warning if `scikits.samplerate` is not detected
peterf@2: 
peterf@2: Features
peterf@2: 
peterf@2:   - New module `librosa.chord` for training chord recognition models
peterf@2:   - Parabolic interpolation piptracking `librosa.feature.piptrack()`
peterf@2:   - `librosa.localmax()` now supports multi-dimensional slicing
peterf@2:   - New example scripts
peterf@2:   - Improved documentation
peterf@2:   - Added the `librosa.util.FeatureExtractor` class, which allows librosa functions
peterf@2:     to act as feature extraction stages in `sklearn`
peterf@2:   - New module `librosa.effects` for time-domain audio processing
peterf@2:   - Added demo notebooks for the `librosa.effects` and `librosa.util.FeatureExtractor`
peterf@2:   - Added a full-track audio example, `librosa.util.example_audio_file()`
peterf@2:   - Added peak-frequency sorting of basis elements in `librosa.decompose.decompose()`
peterf@2: 
peterf@2: Other changes
peterf@2: 
peterf@2:   - Spectrogram frames are now centered, rather than left-aligned.  This removes the
peterf@2:     need for window correction in `librosa.frames_to_time()`
peterf@2:   - Accelerated constant-Q transform `librosa.cqt()`
peterf@2:   - PEP8 compliance
peterf@2:   - Removed normalization from `librosa.feature.logfsgram()`
peterf@2:   - Efficiency improvements by ensuring memory contiguity 
peterf@2:   - `librosa.logamplitude()` now supports functional reference power, in addition
peterf@2:     to scalar values
peterf@2:   - Improved `librosa.feature.delta()`
peterf@2:   - Additional padding options to `librosa.feature.stack_memory()`
peterf@2:   - `librosa.cqt` and `librosa.feature.logfsgram` now use the same parameter
peterf@2:     formats `(fmin, n_bins, bins_per_octave)`.
peterf@2:   - Updated demo notebook(s) to IPython 2.0
peterf@2:   - Moved `perceptual_weighting()` from `librosa.feature` into `librosa.core`
peterf@2:   - Moved `stack_memory()` from `librosa.segment` into `librosa.feature`
peterf@2:   - Standardized `librosa.output.annotation` input format to match `mir_eval`
peterf@2:   - Standardized variable names (e.g., `onset_envelope`).
peterf@2: 
peterf@2: ##v0.2.1
peterf@2: 
peterf@2: Bug fixes
peterf@2: 
peterf@2:   - fixed an off-by-one error in `librosa.onset.onset_strength()`
peterf@2:   - fixed a sign-flip error in `librosa.output.write_wav()`
peterf@2:   - removed all mutable object default parameters
peterf@2: 
peterf@2: Features
peterf@2: 
peterf@2:   - added option `centering` to `librosa.onset.onset_strength()` to resolve frame-centering issues with sliding window STFT
peterf@2:   - added frame-center correction to `librosa.core.frames_to_time()` and `librosa.core.time_to_frames()`
peterf@2:   - added `librosa.util.pad_center()`
peterf@2:   - added `librosa.output.annotation()`
peterf@2:   - added `librosa.output.times_csv()`
peterf@2:   - accelerated `librosa.core.stft()` and `ifgram()`
peterf@2:   - added `librosa.util.frame` for in-place signal framing
peterf@2:   - `librosa.beat.beat_track` now supports user-supplied tempo
peterf@2:   - added `librosa.util.normalize()`
peterf@2:   - added `librosa.util.find_files()`
peterf@2:   - added `librosa.util.axis_sort()`
peterf@2:   - new module: `librosa.util()`
peterf@2:   - `librosa.filters.constant_q` now support padding
peterf@2:   - added boolean input support for `librosa.display.cmap()`
peterf@2:   - speedup in `librosa.core.cqt()`
peterf@2: 
peterf@2: Other changes
peterf@2: 
peterf@2:   - optimized default parameters for `librosa.onset.onset_detect`
peterf@2:   - set `librosa.filters.mel` parameter `n_mels=128` by default
peterf@2:   - `librosa.feature.chromagram()` and `logfsgram()` now use power instead of energy
peterf@2:   - `librosa.display.specshow()` with `y_axis='chroma'` now labels as `pitch class`
peterf@2:   - set `librosa.core.cqt` parameter `resolution=2` by default
peterf@2:   - set `librosa.feature.chromagram` parameter `octwidth=2` by default
peterf@2: 
peterf@2: ## v0.2.0
peterf@2: 
peterf@2: Bug fixes
peterf@2: 
peterf@2:   - fixed default `librosa.core.stft, istft, ifgram` to match specification
peterf@2:   - fixed a float->int bug in peak_pick
peterf@2:   - better memory efficiency
peterf@2:   - `librosa.segment.recurrence_matrix` corrects for width suppression
peterf@2:   - fixed a divide-by-0 error in the beat tracker
peterf@2:   - fixed a bug in tempo estimation with short windows
peterf@2:   - `librosa.feature.sync` now supports 1d arrays
peterf@2:   - fixed a bug in beat trimming
peterf@2:   - fixed a bug in `librosa.core.stft` when calculating window size
peterf@2:   - fixed `librosa.core.resample` to support stereo signals
peterf@2: 
peterf@2: Features
peterf@2: 
peterf@2:   - added filters option to cqt
peterf@2:   - added window function support to istft
peterf@2:   - added an IPython notebook demo
peterf@2:   - added `librosa.features.delta` for computing temporal difference features
peterf@2:   - new `examples` scripts:  tuning, hpss
peterf@2:   - added optional trimming to `librosa.segment.stack_memory`
peterf@2:   - `librosa.onset.onset_strength` now takes generic spectrogram function `feature` 
peterf@2:   - compute reference power directly in `librosa.core.logamplitude`
peterf@2:   - color-blind-friendly default color maps in `librosa.display.cmap`
peterf@2:   - `librosa.core.onset_strength` now accepts an aggregator
peterf@2:   - added `librosa.feature.perceptual_weighting`
peterf@2:   - added tuning estimation to `librosa.feature.chromagram`
peterf@2:   - added `librosa.core.A_weighting`
peterf@2:   - vectorized frequency converters
peterf@2:   - added `librosa.core.cqt_frequencies` to get CQT frequencies
peterf@2:   - `librosa.core.cqt` basic constant-Q transform implementation
peterf@2:   - `librosa.filters.cq_to_chroma` to convert log-frequency to chroma
peterf@2:   - added `librosa.core.fft_frequencies`
peterf@2:   - `librosa.decompose.hpss` can now return masking matrices
peterf@2:   - added reversal for `librosa.segment.structure_feature`
peterf@2:   - added `librosa.core.time_to_frames`
peterf@2:   - added cent notation to `librosa.core.midi_to_note`
peterf@2:   - added time-series or spectrogram input options to `chromagram`, `logfsgram`, `melspectrogram`, and `mfcc`
peterf@2:   - new module: `librosa.display`
peterf@2:   - `librosa.output.segment_csv` => `librosa.output.frames_csv`
peterf@2:   - migrated frequency converters to `librosa.core`
peterf@2:   - new module: `librosa.filters`
peterf@2:   - `librosa.decompose.hpss` now supports complex-valued STFT matrices
peterf@2:   - `librosa.decompose.decompose()` supports `sklearn` decomposition objects
peterf@2:   - added `librosa.core.phase_vocoder`
peterf@2:   - new module: `librosa.onset`; migrated onset strength from `librosa.beat`
peterf@2:   - added `librosa.core.pick_peaks`
peterf@2:   - `librosa.core.load()` supports offset and duration parameters
peterf@2:   - `librosa.core.magphase()` to separate magnitude and phase from a complex matrix
peterf@2:   - new module: `librosa.segment`
peterf@2: 
peterf@2: Other changes
peterf@2: 
peterf@2:   - `onset_estimate_bpm => estimate_tempo`
peterf@2:   - removed `n_fft` from `librosa.core.istft()`
peterf@2:   - `librosa.core.mel_frequencies` returns `n_mels` values by default
peterf@2:   - changed default `librosa.decompose.hpss` window to 31
peterf@2:   - disabled onset de-trending by default in `librosa.onset.onset_strength`
peterf@2:   - added complex-value warning to `librosa.display.specshow`
peterf@2:   - broke compatibilty with `ifgram.m`; `librosa.core.ifgram` now matches `stft`
peterf@2:   - changed default beat tracker settings
peterf@2:   - migrated `hpss` into `librosa.decompose`
peterf@2:   - changed default `librosa.decompose.hpss` power parameter to `2.0`
peterf@2:   - `librosa.core.load()` now returns single-precision by default
peterf@2:   - standardized `n_fft=2048`, `hop_length=512` for most functions
peterf@2:   - refactored tempo estimator
peterf@2: 
peterf@2: ## v0.1.0
peterf@2: 
peterf@2: Initial public release.