The article considers the pre-processing voice signals for voice recognition systems based on the use of artificial neural networks. Based segmentation preprocessing is put in the speech signal according to a phonetic transcription of language, in order to reduce the amount of data supplied to the input of the neural network, which considerably improves its input data sensitivity. Application of numerical methods in processing will reduce acoustic noise impact on the speech signal segmentation, which will more accurately identify the areas of classification. Simulation results of the speech signal partition into components are shown, i.e. the selection of phonemes which will be the voice message classification.
29 October 2026 in Eastern-European Journal of Enterprise Technologies
The object of the study is preprocessing, detection, motion planning, sorting, and target placement of Data Matrix images in boxes using a robotic manipulator with six degrees of freedom. The problem to be solved is that motion planning for a robotic manipulator, sorting operations by the manipulator and recognition and preprocessing of Data Matrix images are usually considered as separate tasks. Therefore, the influence of the Data Matrix image quality on the efficiency of sorting by a robotic manipulator is ambiguous. The proposed model comprises three main levels: data acquisition, software development, and planning and execution. A cascaded Gaussian filtering method was employed to process Data Matrix images distorted by Gaussian noise. The robotic arm's operation was simulated in the ROS 2 (USA) environment using RViz (USA) and MoveIt (USA) tools. As a result of the two-stage filtering, the root mean square error (RMSE) decreased across all noise levels studied: from 21.19 to 19.55 at +10 dB, from 36.36 to 33.62 at +5 dB, from 55.78 to 52.46 at 0 dB, from 72.07 to 68.57 at −5 dB, and from 82.64 to 79.11 at −10 dB. As a result of filtering, the peak signal-to-noise ratio (PSNR) increased by 0.38–0.70 dB. The time required for trajectory planning generally ranged from 0.02 to 0.04 seconds, while the length of the movement trajectories varied between 0.55 and 1.05 meters. In most cases, the root-mean-square error of angular deviation did not exceed 0.10°, and the displacement error ranged from 0.9 to 1.5 mm. The results demonstrated that image filtering enhances the stability of visual data. These metrics indicate the high efficiency of the proposed approach. The quality of image preprocessing was evaluated based on the integrity of the binary code structure, alongside metrics such as root-mean-square error (RMSE) and peak signal-to-noise ratio (PSNR). The proposed approach can be applied in warehouse and logistics systems where it is necessary to monitor noise levels in camera-captured images and reliably identify objects via codes prior to their movement and sorting