Automatic Speech Recognition Research Paper

Assignment
ECE 3266_Digital Audio Signal Processing

Trimester 1 2016/2017

ID 1111113540
ABDULRAHMAN SAEED

Introduction:
In real life the need of smart assistant application is one of advantage of smartphone. Audio Recognition application is useful because it helps in making the life easier, in such a way of communication between you and software to do what is it has been ordered. one of these application Siri, Cortana, and Google’s. all of these application is build in application on OS system and work as intelligent personal assistant. Automatic Speech Recognition (ASR) is being used to satisfy this task. ASR is the process of converting an acoustic signal to a set of word. This technology perspective, speech recognition

To begin application, you have to type a question in the search box, or select the microphone and talk to Cortana. several techniques. I am concerning in this repot about voice command.
1- ASR (Automatic speech recognition) to transcribe human speech (in this case, short utterances of commands, questions, or dictations) into text.

 It has two stages preprocessing and recognition. In preprocessing stage, we prepare the waveform for recognition using:
• Digitize and represent waveforms: in this stage we use we take pure audio signal then convert it to digital signal using DSP In general, we take the pure audio signal then by using low pass filtering which is necessary before A/D conversion to bandlimited the signal to ensure the absence of aliasing. After that we sample signal, sampling is representing a continuously varying signal as a periodic sequence of values. Third stage quantization, is representing a waveform value by one of a finite set of values. Fourth stage coding which is concerns assigning a binary number to each value.

• Feature extraction (10 ms frame)
Important Feature: Mel-frequency cepstral coefficients (MFCC) developed based on how human hears sound.
 Recognition: Identify what the user has

(example: Today's weather report "The weather will be sunny”)
It is done based on Analysis-Synthesis Method which is a speech waves of recorded human voice are transformed into parameter sequences and is stored in a memory. In Synthesizer by concatenated parameters to produce desired messages. This method allowing us to control the parameters, smoothing the pitch and spectral change at connections can be performed. Microsoft had chosen this technique base on Intelligibility of this method is high, although the naturalness of synthesized speech is slightly degraded. There is better method in TTS on naturalness of synthesized like Waveform coding but it used only memory as main hardware as Cortana need a processor to handle multitask such as third party connection (Google