以PDA為平台之語音辨識 應用系統開發

Speech recognition application system on Personal Digital Assistant

孫益君(Yi-Chun Sun)

指導教授 ： 杜筑奎

中原大學/工學院/資訊工程研究所/碩士(2004年)

https://doi.org/10.6840/cycu200400159

摘要

本研究主要在建立一套語音辨識環境。其研究平台主要是透過PDA(個人數位助理)系統，將語音訊號經由特徵參數處理過程以求得辨識語音的特徵參數。其次，將語音特徵參數輸入至類神經網路辨識模型，以便完成辨識語音處理及執行各種語音指令與功能。語音特徵參數來源，使用語音LPC演算法與倒頻譜特徵參數，語音特徵參數共截取280個特徵值。系統辨識核心，為使用類神經倒傳遞網路架構。其輸入層神經元數為280、隱藏層神經元數為20，而輸出層神經元數為4。透過類神經辨識網路模型獲得語音辨識結果，並將結果傳送至遠端伺服器來控制相關設備。而本研究之語音辨識率為92%，其辨識速度平均為2秒半左右。

當伺服器接收到PDA指令後，便根據所下達指令做相應的處理及外部裝置操作。同時，將所截取的影像畫面做初步的分析及處理，並將影像畫面回傳並顯示至遠端的PDA之中，使得執行者能夠清楚的瞭解到所下達的指令結果與目前裝置設備的狀態。當伺服器在執行外部裝置時，其使用RS-232為主要的傳送通道。透過RS-232傳送通道，來啟動或關閉外部裝置設備。同時，也配合所設計之紅外線及無線電發射與接收介面板，使得本研究系統的應用領域範圍更加的延伸與擴展。

關鍵字

DSP ； 語音辨識 ； 類神經網路 ； PDA

並列摘要

This paper presents a new approach to design a speech recognition system. The user can use speech voice to operate equipments in remote place. The first step of the procedure is to analysis the speech signal to get speech features. The second step is to establish a neural network, for the purpose of speech recognition.

The model of neural network method uses Black Propagation Network (BPN). The speech features extraction uses the methods of Linear Predictive Coefficient (LPC) and Cepstrum. The Black Propagation Network import the 280 numbers of speech feature to train the neural model until the neural model has be convergent. Then, user input a speech command to the system. The neural model will recognize the command.

The system is established on Personal Digital Assistant (PDA) and connected with computer system as its remote server. When user input speech command to PDA, the PDA will recognize speech and execute the commands. Then, the remote server will take action on equipment according to the command of the speech. The remote server has a basic ability of image analysis, and can transfer the image, which take from remote field, back to the command side.

When server interface with it peripherals, the RS-232 serial data transfer standard is used, via this channel the command signals are implement to control the remote equipment. In order to enhance the ability of application an Infra-Red and a Radio wave transmitter receiver interface card are equipped.

並列關鍵字

PDA ； Neural Network ； speech ； DSP

參考文獻

1.Emmanuel C. Ifeachor, Barrie W. Jervis, “Digital Signal Processing A Practical Approach”, 1993.

2.Paul A. Lynn, Wolfgang Fuerst, “Introductory Digital Single Processing With Computer Applications”, 1989.

3.O'Shaughnessy D.” Interacting with computers by voice: automatic speech recognition and synthesis”, Proceedings of the IEEE, Volume: 91, Issue: 9, Sept. 2003.

4.Morgan N. and Bourlard H., “Continuous speech recongnition”, IEEE Signal Processing Mag., vol. 12, pp. 25-42, May 1995.

5.---,”Neural networks for statistical recognition of continuous speech”, Proc. IEEE, vol. 12, pp. 25-42, May 1995.