UPLB ICS Peak One

The convenience and accessibility of digital storefronts has increased the usage of digital music. However, the quality of digital music can not always be verified by checking encoding and bit rate in the metadata. Bad quality transcodes harm artists who produce high quality music and consumers who download an imperfect product. This study examines the use of spectrogram analysis to classify audio files using a convolutional neural network (CNN). Audio samples were randomly selected from online sources and converted into spectrogram images. These were used to train and evaluate the accuracy of the CNN. Evaluation revealed that the model scored 98.39% in overall accuracy. The model scored the lowest in precision for FLAC and V0 files which means that their spectrograms can be tricky to distinguish. Additionally, a desktop application was developed as an interface for inference. Results suggest that classifying audio files is effective through the use of spectrograms and a CNN.

The Villa Remedios East Homeowners Association (VREHOA) is a private organization with the aim of improving the residents’ quality of life through proper management anduse of organization funds. Not having a single platform for their management needs, record-keeping is done manually through the use of tools such as a spreadsheet. Task assignments are also done via messages through social media, SMS, or verbally. This study aims to provide a Management Reporting System to the VREHOA which will serve as a platform to record their income, expenses, resident complaints, and officer task assignments. Having a single platform for their management needs, the Management Reporting System aims to aid the VREHOA by providing relevant information to be used for decision making.

Efficient and effective communication is important along with the rapid developments in the society and various cross-cultural interactions. A major factor in communication is the speed of language acquisition. This study developed a spaced repetition model that uses tonal features to predict the user recall rate of a Thai word. Data used to develop the model was from an Android application used by participants for 14 days. The developed model was observed to have the best performance among previously implemented models, improving the performance of its base model by 22.7%.

Peanut seeds are susceptible to mold infestation such as Aspergillus, which causes the release of aflatoxins, a carcinogenic substance that can harm human and animal health. The aim of this study is to develop a method for detecting peanut mold using computer vision and artificial intelligence. To isolate the peanut seeds from acquired images for color and texture analysis, marker-based watershed segmentation was applied. RGB, HSV, and Grayscale color spaces are used to extract color features, whereas Gray Level Co-occurrence Matrix properties are used to extract texture features. Selected relevant features were fed into a feedforward backpropagation neural network, which generates values ranging from 0 to 1, with values closer to 1 indicating mold contamination and values near 0 indicating non-contamination. In detecting mold infestation in peanut seeds, the neural network produced an accuracy rate of 89.33%.

Assignment of students to classes is an important task in a university registration or enrollment process and is considered as a subproblem of the more general university timetabling problem which is NP-complete. This study characterizes the assignment of students to classes as the Student Sectioning Problem and proposes a multiagent system framework for solving it. The framework is composed of three types of agents namely scheduler agent, enlister agent, and student agent. The scheduler agent is responsible for creating the initial solution using iterative algorithms augmented with a maximum bipartite matching solver. Student agents communicate with enlister agents to improve the initial solution by performing cancellation and enlistment operations. A prototype system was implemented on top of the JADE multiagent platform using UPLB registration data. In general, the multiagent system framework provides a decentralized and concurrent approach to solving the Student Sectioning Problem and offers a more realistic model for the student registration process.