Rowena Cristina L. Guevara

FMO slice group maps using spatial and temporal indicators for H.264 wireless video transmission

We present a new method, the STI-FMO, that combines spatial and temporal indicators of MB importance to generate slice group maps for H.264 wireless video transmission. Starting from an initial MB-to-slice group mapping we propose a distortion-from-propagation measure that can be used to generate a slice group map that offers less distortion than the initial map from an error concealment point of view. The new method exploits the location of the MB within a slice group, the distortion measure of the MB in the same slice and the probability of error of a MB estimated from the bit-count information to compute a distortion-from-propagation measure. The new method considers the effect of the MBs location within a slice group as added information to assess its importance. Experiments indicate the effectiveness of the proposed method.

Global word shape processing using polar-radii graphs for offline handwriting recognition

The result of investigating the capability of the polar-radii graph (PRG) of the word contour as a global feature for offline handwriting recognition is presented in this paper. The PRG is combined with the local geometric features to improve the accuracy of the handwriting recognition system. The handwriting recognition system using local features and an HMM recognizer obtained a recognition rate of 58% for a 20-word vocabulary using the demo version of the publicly available IAM database. Using the PRG of the word contour as a global feature and an MLP classifier, the recognition rate is 78%. Using the linear confidence accumulation (LCA) method to combine the results of the recognizers based on the local and global features, an over-all recognition accuracy of 72% was obtained.

Polar-radii graphs: a geometric-invariant, shape-based approach to handwritten character recognition

The problem of variability in handwriting due to geometric distortion of letters by scale, location and rotation is addressed using the polar-radii graph (PRG) scheme. The PRG of a handwritten character is a functional representation that can be used to describe and regenerate the character with accuracy. It shows the relationship between the relative distances from the characters centroid to points along the outer boundary of the character image as a function of angle. A handwriting recognizer was developed using PRGs as features and was tested using both single and double character handwriting obtained from 500 different writers. The handwritten texts are limited to at most two characters that may or may not be connected. They can be any of the 52 upper and lower case characters of the English alphabet. The test for 26,000 single handwritten characters yielded a recognition rate of 90.22%. The test for 1,500 double characters yielded a recognition rate of 87.08%.

Developing a children's Filipino speech corpus for application in automatic detection of reading miscues and disfluencies

Recognizing the potential benefit that the current speech processing technology offers to improve childrens literacy, researchers in the past few years have devoted their efforts in developing reading miscue detectors (RMDs) and automated reading tutors (ARTs). A primary challenge however in developing speech technologies for children may be the unavailability of a dedicated childrens speech corpus that can be used for system design and test. In the past few years, childrens speech corpora have been developed for languages such as English, Dutch, Chinese Mandarin, Italian, German and Swedish. But since Filipino has features and orthography that are distinct from other languages, the focus of this study is the development of a childrens Filipino speech corpus (CFSC). In this paper, we present the CFSC design, reading text, data collection procedure and speech transcription method. We also performed initial analysis of the reading miscues and disfluencies found in the CFSC. The results of the miscue analysis suggest possible ways for modeling the reading miscues and possible methods for detecting them. Among these methods are acoustic model likelihood calculation and analysis of duration-based prosodic features. The CFSC presented in this study will be used for the development of an RMD and an ART for Filipino.

Implementation of sliding mode control for current sharing in fixed frequency voltage regulator modules

Voltage regulator modules (VRMs) have to maintain perfect current sharing between modules in both static and dynamic loading. In this study, sliding mode control was applied to both the voltage loop and the current share loop of the two-phase VRM with the aim of providing an effective current sharing scheme. Unlike other sliding mode control implementation that uses infinite switching frequency in its control method, fixed switching frequency was used. Derivation for the fixed frequency sliding mode control was done in the analog domain and transformed into the digital domain. A simulation model using PSIM was developed for both the analog and the digital control implementations. Results show that current sharing is achieved in both the analog and the digital control.

Feature set for Philippine Gong Music classification by indigenous group

In this study, the feature set which brought about the highest classification accuracy for sorting Philippine Gong Music clips by indigenous group was sought. The features reflected Timbre, Loudness, Rhythm and Melody-and-Pitch. Two classifiers were used: Support Vector Machines and Neural Networks. Sequential Feature Selection was used to optimize the feature set. The highest accuracy achieved was 90.83% when the combination of SVM, 30s clips and the full Timbre feature set (64 features) was used. K-means clustering was also done to find similarities among the gong styles of the different groups.

Real-time implementation of wideband sinusoidal speech coder on ADSP-21065L

This paper discusses the real-time implementation of the wideband sinusoidal (WS) speech coder using the ADSP-21065L EZ-Kit Lite™ evaluation board. To achieve a real-time implementation, optimization strategies were applied to the speech coding algorithms and data collection and speech output blocks were implemented as interrupt service routines (ISRs). The entire WS speech coder was programmed in assembly. This implementation uses 7.35% of the DSPs processing power and 38.94% of the DSPs program and data memory. Listening test results showed that the output speech has acceptable quality and intelligibility, with WS speech coder configurations obtaining MOS greater than 3.0 and DRT scores of higher than 80%.

Improved Sorting Algorithm for Explicit FMO Macroblock Classification in H.264

An improved sorting algorithm is presented that classifies macroblocks into different slice groups for FMO in H.264. The classification process reinforces the slice structured coding nature of H.264 by making a conscious effort to minimize the variations in the statistics of the macroblock parameters of the resulting slice group maps. Results show that modest improvements can be achieved in terms of PSNR and the number of undecodable macroblocks by using the proposed improved sorting algorithm as compared to an interleave sorting algorithm where the MB is assigned to different slice groups following the order of the sorted list.

MIDI implementation of a kulintang modal synthesizer using the VST2.4 standard

Philippine Indigenous Music is slowly disappearing partly because of insufficient accessibility to Philippine indigenous musical instruments. This project aims to give musicians access to the sounds of the Philippine instrument known as the kulintang through a synthesizer plug-in. This paper ventures into the implementation of a Virtual Studio Technology Instrument (VSTi) plug-in containing kulintang sounds analyzed using the Modal Distribution and synthesized using Sum of Sinusoid (SOS) Synthesis. The VSTi plug-in was implemented in C++ using the VST2.4 SDK (Software Development Kit). The plug-in is controlled by Musical Instrument Digital Interface (MIDI) events. The Filipino Kulintang MIDI (FKM1.0) was proposed as an initial MIDI mapping standard for Filipino kulintang instruments. A VST Dynamic Link Library (DLL) containing the kulintang synthesizer was successfully ported and controlled by a Digital Audio Workstation (DAW). Twenty three tuning presets with eight gongs each were included in the library.

Prosody modification in Filipino speech synthesis using dynamic time warping

Prosody is composed of two components: microprosody and macroprosody. Microprosody is solely influenced by individual speech sounds while macroprosody is subject to the speakers choice of intonation. The paper deals with the latter and describes the result of using dynamic time warping (DTW) for changing the macroprosody of speech segments in a concatenative synthesizer. Prerecorded Filipino words uttered in isolation are stored in a corpus. When a text is typed, the utterances of the words are searched in the corpus. The acoustical features of macroprosody are extracted from prerecorded utterances of selected Filipino sentences and modified through DTW. The acoustic features are embedded in the speech segments by using the TD-PSOLA. This synthesis process achieves an average MOS of 2.64 in the acceptability test. In a separate test procedure, results showed 98% of the synthesized Filipino sentences were accurately distinguished as either declarative or interrogative sentences.