Automatic speech recognition for mobile and personal applications

在这项提案中，我们研究了如何提高自动语音识别(ASR)的性能。自动语音识别将普通语音转换为文本。它允许无人监督的程序通过语音与人交互，从而使人们能够通过电话或PDA(个人数字助理)访问广泛的数据服务。自动语音识别现在只为有限的服务而存在，因为其准确性远远低于人类收听者。如果要使自动语音识别获得广泛的使用，我们需要使其更有效率。这项建议涉及几个有针对性的领域，例如，通过无线信道访问数据库。这项研究涉及智能信号处理，因为我们需要开发算法来根据语音信号的意图来解释它们。考虑到当前ASR中使用的模型的简单性，当与人类所做的复杂的、非线性的处理相比时，这表明仍然有许多方法可以提高ASR。在每个级别(即，特征估计、时间和声学建模、语言建模、决策)，ASR已经做出了妥协，以具有简单和快速的处理，以较低的识别准确率为代价。我们打算检查其中的几个折衷方案，以及接近人类语音识别性能的替代方法。我们工作的一个主要应用是通过语音搜索和访问信息，例如万维网上的大型数据库。在小型便携式设备上打字很笨拙，因此非常希望通过语音访问。在人机交互中，通常使用鼠标+键盘作为机器输入，使用计算机屏幕或打印机作为输出。然而，语音在人类交流中一直具有很高的优先级，早在写作之前就已经开发出来了。虽然在通信带宽上不如文本经济，但语音比文本处理更通用、更容易使用，例如，人们说话比打字更快、更容易。这里提出的研究涉及改进ASR，使其更实用，因此更有用。

In this proposal, we examine ways to improve the performance of automatic speech recognition (ASR), which converts ordinary speech into text.  This allows unsupervised programs to interact with people via voice, thus giving people access to a broad range of data services, over telephones or PDAs (personal digital assistants).  ASR exists now only for limited services, as its accuracy is much lower than that of human listeners.  For ASR to gain widespread use, we need to make it much more efficient.  This proposal addresses several areas of targeted interest, e.g., accessing databases over wireless channels.  The research deals with intelligent signal processing, as we need to develop algorithms to interpret voice signals in terms of their intent.     The simplicity of models used in current ASR, when compared to the complex, nonlinear processing done by humans, suggests that there remain many ways to improve ASR.  At each level (i.e., feature estimation, temporal and acoustic modeling, language modeling, decision making), compromises have been made in ASR to have simple and fast processing, at the expense of lower recognition accuracy.  We intend to examine several of these compromises and alternative ways to approach the performance of human speech recognition. A major application of our work is to render search and access to information via voice, e.g., large databases as found on the World Wide Web.  Typing on small portable devices is awkward; thus access via voice is highly desired.     Human-computer interactions often use a mouse+keyboard as machine input, and a computer screen or printer as output.  Speech, however, has always had a high priority in human communication, developed long before writing.  While less economical in communication bandwidth than text, speech is more universal and easy to use than text processing, e.g., people speak more quickly and easily than typing.  The proposed research here deals with improving ASR to render it more practical and hence more useful.