| Abstract [eng] |
This study presents „Google“ Lithuanian speech recognition efficiency evaluation research. For the experiment it was chosen method that consists of three parts: first of all to process all voice recordings without adding any noise, secondly process all voice recordings with several different types of noise, modified so as to get some predefined signal to noise ratios (SNR) and at last after one month reprocess all voice recordings without any additional noise and to assess improvements in the quality of the speech recognition. It was chosen WER (Word Error Rate) metrics for speech recognition quality assessment. Analyzing the results of the experiment it was observed that the greatest impact on the quality of speech recognition has a signal-to-noise ratio (SNR) and speech type (most recognizable is isolated words, the worst - spontaneous speech). Meanwhile, characteristics such as the gender of the speaker, smooth speech, speech speed, speech volume does not make any significant influence on speech recognition quality. |
