Search

Caption and scene texts found in images contain valuable information. These texts can be used for many applications to answer questions like what, when, where, and by who to give context to the images. So, automatic text detection enhances the user’s understanding of the media content. In Ethiopia, most street posts and promotional boards are written in multilingual characterssuch as Latin (English, Afaan Oromo etc.) and Ethiopic (Amharic, Tigrigna etc.). In this work, we have studied Ethiopic and Latinmultilingual text detection from images for both caption and scene texts. After the images are pre-processed, maximally stable extremal region (MSER) algorithm, aspect ratio and stroke width transform (SWT) algorithm are used to extract text regions, respectively. Then texture features are computed using local binary patterns (LBP) from the extracted regions. Finally, the support vector machine (SVM) is used to classify text region vs nontext using the computed LBP features. We prepared a new multilingual Ethiopic and Latin script image dataset to evaluate our method.

A language identification (LID) system is an approach in which machines can determine the language and identify it from relatively brief audio spoken samples. Very few attempts have been made on LID Systems for Ethiopian languages. The importance of LID is increasing due to the development of telecommunication infrastructures. Using an LID, service calls from customers can be forwarded to a person who knows the language. Therefore, an LID system involving four Ethiopian languages (Amharic, Oromiffa, Guragegna and Tigregna) is done using Gaussian mixture models (GMM). A dataset consisted of recordings from seven different speakers of each language was prepared and after preprocessing the data, the features are extracted using Mel frequency cepstral coefficients (MFCC) and classification is done using GMM. The performance of the LID system was tested with scenarios by taking two, three and four languages at a time. The LID system is also tested for utterance and speaker dependence performances. The average accuracy of utterance dependent LID test for the four languages was about 93%, the utterance independent test for the four languages was about 70% while the speaker independent test, being tested on utterance dependent scenario only, for the four languages was nearly 91%.Keywords: Accuracy, GMM, LID, Language Identification System, MFCC, Utterance

With the rapid increase of computing resources within the last decade, several real-time applications such as sound editing and processing are becoming more popular. Virtual Studio Technology (VST) was introduced by Steinberg Media Technologies GmbH in 1996 which provided a platform for sound effect developers to implement their own effect plugins. In 1999 a standard framework for software sequencers, so called VST instruments (VSTi), was introduced in which music instruments are simulated in the virtual studio. To design an instrument as part of the virtual studio, a series of steps from audio sample collection and recording to integrating the samples using a plug-in to the VST is required. To the best of our knowledge, there is no VSTi plugins developed for Ethiopic music instruments so far. Therefore, we have effectively used the structure of the VSTi API and its Software Development Kit (SDK) tools to design and implement a plugin for Kirar (Ethiopian music instrument) and effectively integrated and tested. Our plugin was successfully recognized and operated in the Cubase5 and FL studio Digital AudioWorkstations (DAWs) as a VSTi instrument which initiates its own graphical user interface (GUI) with full mouse and keyboard controls having 3 level Musical Instruments Digital Interface (MIDI) mapping.Keywords: DSP, Kirar, MIDI, VSTi