This paper investigates the effectiveness of quantization techniques for speaker verification models on resource-constrained Android devices. Tiny-SV-CNN and EdgeSpeakerCNN-AAM were trained on VoxCeleb1, evaluated using the VoxCeleb1-O protocol, and deployed on a Samsung A04 smartphone with TensorFlow Lite. Experimental results show that INT8 significantly reduces the model size of Tiny-SV-CNN, while FP16 preserves the accuracy of EdgeSpeakerCNN-AAM with nearly 50% model size reduction. The study provides practical guidelines for selecting appropriate deployment formats for edge AI applications.
| Mualliflar | Nurimov, Parakhat |
|---|---|
| Jurnal | Al-Farg'oniy avlodlari |
| Nashr sanasi | 2026-07-19 |
| Son | 3 |
| Betlar | 44-48 |
| Til | Rus |
Speaker Verification, Android, TensorFlow Lite, kvantlash, FP16, INT8, Edge AI, Tiny-SV-CNN, AAM-Softmax, Speaker Verification,Android,TensorFlow Lite,quantization,FP16,INT8,Edge AI,Tiny-SV-CNN,AAM-Softmax
. In this study, the structural changes, phase composition, and hardness of gray cast iron samples containing different amounts of alloying elements were analyzed. The results showed that increasing copper content…
This study investigates the problem of modeling the deformation of square thin plates under the influence of their own weight. The main objective is to conduct a comparative analysis of the accuracy and efficiency of…
The spatial distribution of slope angle is an important geotechnical indicator for assessing open-pit wall stability. This study proposes a method for deriving slope fields from a global digital elevation model and…
Uzbek phraseological units are figurative and structurally fixed, so machine-translation and NLP systems often mishandle them. On 150 idioms sampled from a 4,408-unit corpus, each was annotated for structural type, a…
This paper proposes an adaptive hybrid machine learning model for remote fault detection in wireless data transmission network elements, fault-severity and root-cause assessment, and automated response generation. The…
This article uses the finite difference method to study the stress-strain state of a rectangular plate with a central horizontal crack under plane stress conditions. An equivalent isotropic linear elastic model of…
Imbalance corrections are known to harm calibration without improving discrimination in logistic risk models, but this has not been examined where a regressor predicts a continuous marker and categories follow from…
This paper investigates the mathematical framework for managing VM migration in cloud environments. Pre-copy, post-copy and hybrid approaches are formalized, an automatic migration model is developed using a MDP. The…
This paper proposes a concept for intelligent energy management of a distillation column based on an Adaptive Neuro-Fuzzy Inference System (ANFIS) that combines the interpretability of fuzzy logic with the learning…
The paper presents a systematic study of stochastic mathematical models and simulation methods applied to evaluating the performance of computing machines. Three queuing models have been formalized and verified — M/M/1…