Bernoulli versus Markov: Investigation of state transition regime in switching-state acoustic models
نویسندگان
چکیده
In this paper, a new acoustic model called time-inhomogeneous hidden Bernoulli model (TI-HBM) is introduced as an alternative to hidden Markov model (HMM) in continuous speech recognition. Contrary to HMM, the state transition process in TI-HBM is not a Markov process, rather it is an independent (generalized Bernoulli) process. This difference leads to elimination of dynamic programming at state-level in TI-HBM decoding process. Thus, the computational complexity of TI-HBM for probability evaluation and state estimation is OðNLÞ (instead of OðNLÞ in the HMM case, where N and L are number of states and sequence length respectively). As a new framework for phone duration modeling, TI-HBM is able to model acoustic-unit duration (e.g. phone duration) by using a built-in parameter named survival probability. Similar to the HMM case, three essential problems in TI-HBM have been solved. An EM-algorithm-based method has been proposed for training TI-HBM parameters. Experiments in phone recognition for Persian (Farsi) spoken language show that the TI-HBM has some advantages over HMM (e.g. more simplicity and increased speed in recognition phase), and also outperforms HMM in terms of phone recognition accuracy. & 2008 Elsevier B.V. All rights reserved.
منابع مشابه
Estimating Stock Price in Energy Market Including Oil, Gas, and Coal: The Comparison of Linear and Non-Linear Two-State Markov Regime Switching Models
A common method to study the dynamic behavior of macroeconomic variables is using linear time series models; however, they are unable to explain nonlinear behavior of the series. Given the dependency between stock market and derivatives, the behavior of the underlying asset price can be modeled using Markov switching process properties and the economic regime significance. In this paper, a two-...
متن کاملFads Models with Markov Switching Hetroskedasticity: decomposing Tehran Stock Exchange return into Permanent and Transitory Components
Stochastic behavior of stock returns is very important for investors and policy makers in the stock market. In this paper, the stochastic behavior of the return index of Tehran Stock Exchange (TEDPIX) is examined using unobserved component Markov switching model (UC-MS) for the 3/27/2010 until 8/3/2015 period. In this model, stock returns are decomposed into two components; a permanent componen...
متن کاملMonetary Fundamental-Based Exchange Rate Model in Iran: Applying a MS-TVTP Approach
T he main purpose of this article is to analyze exchange rate behavior based on monetary fundamentals in the context of Iranian economy over the period 1990:2 to 2014:3. To do so, two monetary exchange rate models is investigated, the first by regarding interest rate differential as a monetary variable, and the second one regardless of interest rate differential as a monetary variabl...
متن کاملThe Effects of Oil Price Shocks on Transitional Dynamics of Business Cycles in Iran: Markov Switching Model with Time Varying Transition Probabilities (MS-TVTP)
The business cycles are one of the most important economic indicators that they show the changes in economic activities during time. The study of business cycles is important because the understanding fluctuations in GDP and effective factors on these fluctuations help policy makers to plan better and more efficient. The main purpose of this paper is to investigate the effects of oil price shoc...
متن کاملAsymmetric Effects of Government Spending on Economic Growth Over the Business Cycle: Application of Markov Switching Models
This paper is investigated four subject with uses iranian economic data and using the Markov-Switching model during the period (1369: 3-1393: 4), So that: (a) were Examined impact of the positive and negative Fiscal shocks on Iran economic growth ( B) the Hypothesis impact of negative shocks is greater than a positive shock was tested. (C) were tested the impact of government expenditure (f...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید
ثبت ناماگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید
ورودعنوان ژورنال:
- Signal Processing
دوره 89 شماره
صفحات -
تاریخ انتشار 2009