| Abstract [eng] |
The researches of recognition of speech signals are relevant for the development of informational technologies. More and more mechanisms controlled by voice commands which make everyday life easier for both disable and healthy people are used. In the surrounding around we face various noise which makes the perception of speech signals more difficult. Noise negatively affects the technologies which are related to transmission of speech, for example telephonic conversations. The technologies of speech recognition by machines are far from the speech perception of human beings. In order to improve the recognition of speech signals of machines, the principle of speech perception of human beings needs to be analyzed. In this work we will research the speech perception of human beings at different frequencies and affecting at different level of noise. The research of perception of speech signals was carried by using the cut syllables from occasional words from the speech of different newscasters. The cut syllables were normalized to 0dB level before summing up with noise. Stimulating various real situations white noise of different levels was selected: 1. 0dB 2. -5dB 3. -10dB 4. -13dB Later in the work the filter of FIR type of strips of different frequencies was used for each syllable: 1. from 0 to 1000 Hz 2. from 1000 to 2000 Hz 3. from 2000 to 3000 Hz 4. from 3000 to 4000 Hz. During the experiment it was noticed that the recognition of women’s voice is better in strips of higher frequencies of 3_4 kHz, while the recognition of men’s voice is better at medium frequencies of 1_2 kHz and of 2_3 kHz. Then 6 participants of the research who were divided into two age ranges: from 24-27 and from 47-51 listened the syllables and it was noticed that the results of participants of the research of the senior age were worse. During the experiment the influence of noise for the perception of sounds was identified, it was noticed that the more the level of noise is coming to the 0dB the more the recognition of signal is becoming worse. |
