We present an automatic license plate recognition system that can detect and recognize Bangla license plates from an image of vehicles. It is mandatory for vehicles in Bangladesh to have Bangladesh Road Transport Authority (BRTA) standard license plates attached in front and back of the vehicle. We build a database containing vehicles of these type. From this dataset, detecting the license plates in an input image is the first step of our system. We use a convolutional neural network (CNN) model for this. From that, we take the detected license plate image as a new input for the second and similar CNN model to segment and recognize license plate numbers. We tested our model on 200 images and achieved an accuracy of 99.5%. For speed, we tested our model on Google Colaboratory's Nvidia Tesla K80 GPU and attained a speed of 9 frames per second while detecting and recognizing the license plate numbers in a video.

In this paper, a noise robust formant frequency estimation scheme is developed based on a spectral model matching algorithm. Considering the vocal tract as an autoregressive system, a spectral model of repeated autocorrelation function (RACF) of band-limited speech signal is proposed. It is shown that because of the repeated autocorrelation operation on band-limited signal, the proposed model can exhibit prominent formant characteristics. First from given noisy speech observations, an adaptive band selection criterion is developed. Next, on each resulting band-limited noisy speech signal, a repeated autocorrelation operation is carried out, which not only reduces the effect of noise but also strengthens the dominant poles corresponding to the formant frequencies. Finally, spectrum of the RACF is computed and instead of direct spectral peak picking, a model fitting scheme is introduced to find out model parameters which lead to formant estimation. The proposed algorithm has been tested on natural vowels as well as some naturally spoken sentences in the presence of different environmental noises. It is found that the proposed scheme provides better formant estimation accuracy in comparison to some of the existing methods at low levels of signal-to-noise ratio.