10.14489/vkit.2017.09.pp.024-031 |
DOI: 10.14489/vkit.2017.09.pp.024-031 Корсун О. Н., Михайлов Е. И. Аннотация. Рассмотрена проблема выделения полезного речевого сигнала из потока «речь–фон» и «речь–шум». Предложен алгоритм детектирования речевых команд, основанный на использовании спектральной параметризации речевого сигнала, множественной регрессии и методов статистического анализа, в том числе, контрольных карт. Приведены результаты тестирования алгоритма на речевой базе, включающей в себя 20 различных изолированных слов. Ключевые слова: детектирование речевых команд; множественная регрессия; метод контрольных карт.
Korsun O. N., Mikhaylov E. I. Abstract. The speech technologies are being developed intensively in the recent years, especially the automatic speech recognition as an additional channel between human interface and technical devices. One of the most important parts of speech recognition is speech signal detection from non-speech signals. Widespread methods, like amplitude detection, have low accuracy in noisy conditions and at low signal/noise ratio. The article is devoted to a problem of improving accuracy of detecting a useful speech signal from the stream «speech–background» and «speech–noise». A new approach is proposed for this purpose. This new approach is based on the joint use of spectral parametrization, linear multiple regression and methods of statistical analysis, including control charts. It is assumed that basic features of the speech signal distinguishing it from the background, noise and other non-speech signals are contained in a small number of different words (reference words). For example, group of three words was selected in this article. The speech signal is decomposed by a basis of reference words. The estimation of the speech signal level is calculated as the sum of squares of the projections onto the vectors of reference words, using sliding window. In the absence of a speech signal in the sliding window, the value of the estimation is small, because noises and non-speech signals are weakly correlated with speech. When a speech signal appears in the sliding window, the value of the estimation increases rapidly, which is a diagnostic sign of the word. The advantage of the proposed algorithm is that it detects almost any words that do not match with the selected basis of reference words. The practical importance of the problem is that detected words are used for purposes of automatic speech recognition. This article presents the detection algorithm based on proposed methods. The experimental results of approbation of the developed algorithm on the base, which includes twenty different isolated words, are also discussed in the article. Keywords: Detection of speech commands; Multiple regression; Method of control charts.
РусО. Н. Корсун (ФГУП «Государственный научно-исследовательский институт авиационных систем» ГНЦ РФ, Москва, Россия)
EngO. N. Korsun (State Research Institute of Aviation Systems State Scientific Center of Russian Federation, Moscow, Russia)
Рус1. Бондарос Ю. Г., Маковкин К. А., Чучупал В. Я. Система распознавания команд речевого интерфейса пилота для интегрированной модульной авионики // Вестник компьютерных и информационных технологий. 2007. № 4. С. 2 – 13. Eng1. Bondaros Iu. G., Makovkin K. A., Chuchupal V. Ia. (2007). The system of recognition of voice commands in pilot's interface for integrated modular avionics. Vestnik komp'iuternykh i informatsionnykh tekhnologii, (4), pp. 2-13. [in Russian language]
РусСтатью можно приобрести в электронном виде (PDF формат). Стоимость статьи 350 руб. (в том числе НДС 18%). После оформления заказа, в течение нескольких дней, на указанный вами e-mail придут счет и квитанция для оплаты в банке. После поступления денег на счет издательства, вам будет выслан электронный вариант статьи. Для заказа скопируйте doi статьи: 10.14489/vkit.2017.09.pp.024-031 Отправляя форму вы даете согласие на обработку персональных данных. . EngThis article is available in electronic format (PDF). The cost of a single article is 350 rubles. (including VAT 18%). After you place an order within a few days, you will receive following documents to your specified e-mail: account on payment and receipt to pay in the bank. After depositing your payment on our bank account we send you file of the article by e-mail. To order articles please copy the article doi: 14489/vkit.2017.09.pp.024-031 and fill out the .
|