Toponym Speech Recognition System For Cargo Sorting

As an efficient human-computer interaction technology,speech recognition technology have been widely applied to the fields of house,automotive,mobile equipment and industrial control.Recent years,the domestic logistics industry compete intensely.Nevertheless,the logistics site still use the way of scanning code for goods sorting.This result in a low efficiency.there is an urgent need to develop new human-computer interaction technology for goods sorting.This paper proposed a toponym speech recognition system for goods sorting.This system can simplify the goods sorting steps and improve sorting efficiency.The toponym speech modeling system,toponym speech recognition system and the embedded toponym speech recognition system were designed in this paper to realize the recognition of toponym speech for goods sorting in logistics site.The main work of this paper includes three parts:1)Frame design of toponym recognition system.In the hardware of the system,the hardware platform of the system is Tiny4412 development board,it can recognize the voice from the microphone,translate the result into control instruction and send int to machine to control it's operation.In the software of the system,the program was designed by MVC models.This can divide the software into three different parts to reduce the contact between different parts.It's good for the development and maintenance of software.2)The research and improvement of algorithms used in speech recognition system.Firstly,this paper introduced preprocessing of speech signals,including processing of pre emphasis,frames,and window.Then proposed the endpoint detecting using the detect methods based on short-time energy and zero crossing rate,and improved it to increase the accuracy of the system.Then studied the method of extracting MFCC matrix to extract feature from speech signal.The amplitude entropy of the signal was also added into MFCC matrix to form mix characteristics parameters,which can better show the feature of signals.The cepstral mean normalization(CMN)techniques was used to reduce the influence of different speakers.Finally,the modeling methods was studied to build Hidden Markov Model(HMM),and realized the recognition of toponym signals.3)The realize of toponym speech recognition system.This paper designed three systems,including modeling system of toponym signals,toponym speech recognition system andembedded toponym speech recognition system.Firstly,a modeling system of toponym signals and a human-computer desktop were designed to make it easier to build HMM.Then a toponym speech recognition system and was designed based on the recognition methods.This system can realize toponym speech recognition stably on computer.Finally,the speech recognition was moved to embedded board to form embedded toponym speech recognition.The system is easy to take,making its application wider.