diff --git a/Changelog b/Changelog index 58505ed7..3d409594 100644 --- a/Changelog +++ b/Changelog @@ -1,502 +1,525 @@ +=========================================================================== +simon: 0.4.90 +=========================================================================== + +Simon: +====== +- Bugfix: The download of Simon Base Models work again flawlessly (bug: 377968) +- Fix detection of utterid APIs in Pocketsphinx + +cmake: +====== +- drop obsolete modules +- fixed find_package_handle_standard_args usage + + +=========================================================================== +simon: 0.4.80 +=========================================================================== + +First release after 3 or 4 years. Final release notes and changes will be visible +in the 0.5 section above + + =========================================================================== simon: 0.4.1 =========================================================================== General: ======= * Fixing up some strings * Support for Qwt >= 6.1 (bug: 320680) * Better dependency documentation (bug: 314042) Simon: ====== * Better placeholders for the Akonadi plugin (bug: 314422) * Bugfix: Avoid samples filenames that may be too long for the file system * Powertraining: Advance to next page automatically * Bugfix: Server -> Client synchronization: Ensure that all scenarios that we request the client to load are actually synchronized first. * Bufix: Synchronization "hangs" (race condition in the network stream processing) (bug: 314264) * Bugfix: Cleaner timeout handling * Bugfix: Avoid crash on deleting scenarios when those scenarios don't exist anymore (bug: 313417) * Bugfix: Block commands from deactivated scenarios (bug: 314634) * Built-in defaults for prompts, "general" scenario and shadow dict (bug: 320204) SSCd: ==== * Bugfix: Ensure 4 bytes are available before parsing request id Simond: ====== * Bugfix: If a model was repaired during compilation (e.g. phoneme poisoning) link all intermediate inputs to the final result (avoid model re-creation on minor changes) * Bugfix: Full sample synchronization * Bugfix: Enable timeout for possibly malicious requests * Bugfix: Broken HTML in "missing backend" message (bug: 312657) Simonlib: ======== * Bugfix: SPHINX: Support multiple transcriptions per word properly * Bugfix: SPHINX: Fixed encoding issue on WIN32 (bug: 315460) * Bugfix: SPHINX: Proper error message for "too little training data" * Bugfix: SPHINX: Proper error reporting (log) for model compilation (bug: 320202) * Bugfix: SPHINX: Cancel adaption when no grammar was specified * Bugfix: Avoiding double-delete when using multiple recording devices * Bugfix: Better detection of null models =========================================================================== simon: 0.4.0 =========================================================================== General: ======= * Minor string fixes and UI improvements Simon: ====== * Improving the base- and active model information display in the welcome page * Fixing memory leaks in simonvision * Fixing Scenario explosion Simond: ====== * Phoneme poisoning in SPHINX Simonlib: ======== * More robust handling of model input files for both HTK and SPHINX * Bugfixing in the model compilation backends Sam === * SPHINX model adaption =========================================================================== simon: 0.4.0-rc-1 =========================================================================== General: ======= * Updating documentation * Bugfixes in simonvision Afaras: ====== * Handle sample groups in promps Simonlib: ======== * Less strict warnings regarding the sample quality assurance Commands: ======== * Disable the ATSPI plugin for this release * AI: use SimonTTS as output =========================================================================== simon: 0.4.0-beta-2 =========================================================================== Simon: ====== * UI Fix in welcome page * Changing GHNS provider for base models to opendesktop.org Sound: ====== * Fixed race condition in sound input Simond: ====== * Bugfixing SPHINX adaption =========================================================================== simon: 0.4.0-beta-1 =========================================================================== Commands ========= * Adding "Dialog" command plugin: State based dialog plugin with TTS output * Adding "DBus" and "JSON" command plugins * Adding "Akonadi" plugin to schedule Simon commands and to display reminders through dialogs * Introducing multi-stage activation in the Filter plugin * Moving execution of composite commands to separate thread * Shortcuts can now be configured to trigger press, release or both events * Command plugins can now receive arguments through invocation wildcards * Highly experimental: ATSPI plugin prototype Sound ===== * Volume calibration * New sound implementation based on ALSA (Linux) and DirectSound * Multithreaded audio processing Simon ===== * Adding "autorun" commands that are launched when initializing the parent scenario * Expose Simon state through DBus (mainly for Simonoid) * Integrated sample sharing through Voxforge * Introducing Sequitur Grapheme to Phoneme conversion support including Sequitur model generation from within Simon * Adding context-awareness: Activate or deactivate scenarios, sets of training samples and even microphones based on: > Running applications > Currently focused application > Content of a file > DBus information > Face detection > Lip (movement) detection * Scenario hierarchy that is respected by the context layer * Introducing base model containers: SBM (Simon Base Model) with automatic import within Simon. * Completely reworked welcome page and lots of other usability improvements * Added graphical management of local training corpus Simond ====== * Optionally limit write access to a host whitelist * Introduce recognition sharing (multiple clients with same username all receive results from every other client with the same username) and isolated mode (separate recognition result streams for every connected client * Aggressive acoustic model caching * Moving to upstream Julius * Much faster synchronization protocol * Suport for SPHINX: > Model creation with SphinxTrain > Recognition with PocketSphinx Sam === * Playing and re-recording samples from the test set * Better test result breakdown (files, sentences, words), including a graphical overview * Improved test analysis (WER) * Integrated LaTeX reporting * New XML based project files * Batch mode * Add "conservative training" mode: Parses Simond recognition logs Simonlib ======== * Improved event simulation * Introduction of the dialog engine * Adding SimonTTS: A text to speech wrapper framework around Jovie, web services and custom samples * HTK: Configurable GMM generation * HTK: User configurable regression class tree * HTK: GMM auto-detection during adaption * HTK: MFC caching * HTK: Multithreaded model compilation (audio coding, re-estimation) * Automatically detect and handle missing phonemes during model compilation SSC === * Stability improvements - especially on poor connections. * "Locked" mode to hide user details for public accounts Simonoid ======== * New component: Simon plasmoid Shows current recognition status and allows to launch / show Simon Afaras ====== * New component: Automatically Find And Repair Amiss Samples; Helps to identify bad recordings in large corpora =========================================================================== simon: 0.3.0 =========================================================================== General ======= * Up-to-date manuals for all applications * KCMs have been moved to their own library; This fixes problems on e.g. OpenSUSE * Windows' system language is now automatically detected during installation and simon will start in this language automatically * Translating custom installer page to German * Beautifying code through krazy and cppcheck * Global coding style update Translations ============ * General: Added context to many i18n calls with i18nc to help translaters * German: Applications, handbooks * French: Applications * Czech: Applications (90%) Commands ========= * Fixed manual toggle in filter plugin * Fixed calculator output modes * Dictation can now append a configurable text after recognition result (for example a space) Sound ===== * Sample quality assurance: Better warnings and configurable SNR check simon ===== * Vocabulary search is now case insensitive * Adding global keyboard shortcut to activate / deactivate * Bugfix in command configuration * Bugfix: Removing words scheduled for training * Importing training data (prompts) now also imports flac, mp3 and ogg files * Improved the IPA import routine * When selecting the individual files of the base model: Stay in the same folder even on windows * Automatically restart a crashing simond simond ====== * Updating to the latest version of the Julius LVCSR * Performance improvements * Support for binary HMMs for Julius (big performance improvement) * Bugfix in the model check hash * Removing unneeded error messages * Fixing Julius log parsing on Windows ssc === * Offline mode * Fixed potential memory leak sam === * Sorting and filtering of file test results =========================================================================== simon: 0.3-alpha-3 =========================================================================== Sound stack =========== * Fixed crash bug: Closing input device on Linux * SimonSoundInputs are now re-used if a new requester shows up before the last one was deleted even if, at that point, the old input doesn't have any clients anymore ssc === * Model and Type were switched for both microphones and soundcards: Fixed * Calling waitForReadyRead() from different threads on QTcpSocket: Fixed =========================================================================== simon: 0.3-alpha-2 =========================================================================== Sound stack =========== * New sound stack based on QtMultimedia * Allowing the use of multiple devices * Playback with multiple devices simultaniously * Volume calibration Recognition =========== * Implemented our own level based voice activity detection (configurable) Protocol ======== * Julius audio streaming has been replaced with our own implementation using RAW PCM data; Recognition interface now only handels standard WAVE files (easier to replace) SSC / SSCD ========== * Adding support for multiple audio devices * Audio device information is added to the database and can be re- constructed for any sample at any time * Making the connection a bit more stable (handling of connection errors) * New wizard banners simon ===== * Welcome wizard * Handling Greek characters * Automatic stopping / starting a local simond if configured to do so * Configure simond from within simon * New wizard banners sam === * Respecting multiple sound in-/output devices simond ====== * Option to keep all recognition samples Calculator ========== * Adding the option to only output the calculation Build system ============ * Portaudio is now an optional dependency (if activated the command line Julius works with microphone input) =========================================================================== simon: 0.3-alpha-1 =========================================================================== Recognition =========== * simond now sends 10 found sentences from Julius to simon; Also, confidence scores have been added * simon can now ignore recognition results beneath a configurable confidence threshold * simon can now display a "did-you-mean...?"-popup if there are more than one possible recognition results * Improving the recognition * Improving the creation of the model with multiple GMMs, different model type and more * Improving the handling of very, very large models * Implementing model compilation adapter that avoids a lot of basic errors during the compilation process * You can now use static models or adapted base models simon ===== * Implementing package based configuration ("Scenarios") * Implementing a much nicer KDE style welcome page Vocabulary ======== * Better phoneme segmentation during the import process * simon can now import lexica to both the active- and the shadowlexicon (selectable during the import process) * The BOMP can now be imported directly from the internet * Huge performance improvements in the display of the list * simon can now automatically create sentences during the special training * You now don't need to train a word when adding it any more (optional) * The complete vocabulary can now be cleared with the press of a button * Words can now be edited * When words are removed, the user can now move the word to an unused category, or completely remove the word but keep the samples (for a total of 4 deletion modes) Grammar ======= * Displaying examples for each sentence * The text from which to deduce your grammar from can now be entered directly (instead of just loaded from files) Training ======== * simon can now import prompts files * A new "power training" mode speeds up the training process through automatically starting and stopping the recordings * Implementing "Clear training data" * Training texts can now be entered instead of just imported from files Commands ======= * Multiple command plugin instances can now be loaded at once * Implementing keyboard * Implementing calculator * Implementing filter plugin * All triggers can now be configured (including e.g. Buttons in other plugins) * Adding multiple click modes (right click, middle, click, etc.) to the desktop grid * Making plugin fonts configurable * Displaying the complete needed sentence to activate a command in the run dialog ksimond ======= * simon can now be started through the ksimond context menu simond ======= * Adding default user sam === * Adding a new application to the simon suite: sam; The simon acoustic modeller is targeted towards power users who want to manually tweak their model ssc === * Adding new application to the simon suite: ssc; The simon sample collector is a client to the sscd (see below). It is a streamlined application to make collecting a large number of samples as easy as possible. ssc and sscd are more likely to be used for research purposes. sscd === * Adding new application to the simon suite: sscd; The simon sample collector daemon is the server component to ssc (sse above). ssc and sscd are more likely to be used for research purposes.