12.5 - Low Power Speech Enhancement David Halupka Ph.D. Candidate Electronics Group June 24 th, 2005.

Slides:



Advertisements
Similar presentations
Artificial passenger.
Advertisements

NCCR-MICS Project MP3 on Btnode. Main Idea Btnode designed as clever « sensor » Btnode designed as clever « sensor » Goal : Use it as audio sensor (AudioNode)
P A C I F I C C O N C O U R S E D R. S U I T E L O S A N G E L E S, C A P | F | E I.
Masters Presentation at Griffith University Master of Computer and Information Engineering Magnus Nilsson
RAPID Robotic Arm empowering People wIth Disabilities Contact person: Market: Assistive.
Digital Systems: Introductory Concepts Wen-Hung Liao, Ph.D.
King Chung, Ph.D., CCC-A, FAAA Amplification and Communication Research Lab ACRL.
Demo Time Slots: Authors: Gabriel Kopin Eugene Kozorovitsky Advisor: Dr. Rahul Mangharman Contributors: Sidharrth Deliwala ESE 441/442 Senior Design 2007.
ASIC vs. FPGA – A Comparisson Hardware-Software Codesign Voin Legourski.
RoboTechTronix Ryan Fonnesbeck (CS and CE) Brian Clay (CE) Justin Hansen (CE)
KnoteBox Joe Kramer, Leo Ovanesyan, Jimmy Thompon.
Real-Time Speech Recognition Thang Pham Advisor: Shane Cotter.
So far: Historical overview of speech technology  basic components/goals for systems Quick review of DSP fundamentals Quick overview of pattern recognition.
Energy Models David Holmer Energy Model  Captures the effect of the limited energy reserves of mobile devices (i.e. batteries)  Models.
A Voice Based Command and control System for Emergency Applications William H. Lenharth, Ph.D. Project54 / ECE Dept.-UNH.
Institute of Electronics, National Chiao Tung University VLSI Signal Processing Lab A 242mW, 10mm2 H.264/AVC High Profile Encoder H.264 High Profile Encoder.
SECTION 4B CPUs Used in Personal Computers. This lesson introduces: A Look Inside the Processor Microcomputer Processors Parallel Processing Extending.
Bringing your technology to life…
Introduction to Automatic Speech Recognition
Sensory Aids for Persons with Auditory Impairments Damian Gordon Cook and Hussey, Chapter 9.
Information and Communication Technology Fundamentals Credits Hours: 2+1 Instructor: Ayesha Bint Saleem.
ACOE2551 Microprocessors Data Converters Analog to Digital Converters (ADC) –Convert an analog quantity (voltage, current) into a digital code Digital.
Notes on ICASSP 2004 Arthur Chan May 24, This Presentation (5 pages)  Brief note of ICASSP 2004  NIST RT 04 Evaluation results  Other interesting.
Microphone Integration – Can Improve ARS Accuracy? Tom Houy
A 30-GS/sec Track and Hold Amplifier in 0.13-µm CMOS Technology
Input Devices.  Identify audio and video input devices  List the function of the respective devices.
ELEC 423 Digital Signal Processing Prof. Siripong Potisuk.
Group Members: Sam Marlin, Jonathan Brown Faculty Adviser: Tom Miller.
EFC Media Center Training Class 2A. Sound Input Levels Microphone Level Signals Line Level Signals Speaker Level Signals Balanced Inputs Stereo Inputs.
Computers in Police Cruisers Article in Pervasive Computing FIRST RESPONSE Authors: Andrew L. Kun, W. Thomas Miller III, and William H. Lenharth ECE in.
Chapter 4 Applying Technologies for Effective Instruction Perry C. Hanavan.
Recognition of Speech Using Representation in High-Dimensional Spaces University of Washington, Seattle, WA AT&T Labs (Retd), Florham Park, NJ Bishnu Atal.
Timo Haapsaari Laboratory of Acoustics and Audio Signal Processing April 10, 2007 Two-Way Acoustic Window using Wave Field Synthesis.
1 Robust Endpoint Detection and Energy Normalization for Real-Time Speech and Speaker Recognition Qi Li, Senior Member, IEEE, Jinsong Zheng, Augustine.
1 Basic MOS Device Physics. 2 Why Analog Circuits? DSP algorithms were predicted to replace all analog blocks with the flexibility in silicon implementation.
1 Reconfigurable Acceleration of Microphone Array Algorithms for Speech Enhancement Ka Fai Cedric Yiu, Yao Lu, Xiaoxiang Shi The Hong Kong Polytechnic.
Parts of a Computer - Introduction
E.g.: MS-DOS interface. DIR C: /W /A:D will list all the directories in the root directory of drive C in wide list format. Disadvantage is that commands.
Kevin KleinLinxuan Yang. Motivation Develop a means of communication to increase the safety factor for an individual involved in various motorsports Safe.
Audio Location Accurate Low-Cost Location Sensing James Scott Intel Research Cambridge Boris Dragovic Intern in 2004 at Intel Research Cambridge Studying.
Robust Entropy-based Endpoint Detection for Speech Recognition in Noisy Environments 張智星
Why purchase an editing suite? Give students experience on modern video equipment Professional-quality editing studio have come within reach of schools.
Interfacing and I/O Peripherals Dr John Cowell phones off (please)
Implementing algorithms for advanced communication systems -- My bag of tricks Sridhar Rajagopal Electrical and Computer Engineering This work is supported.
Conference Phone Manager V2 (CPM)
Dual-Use Wideband Microphone System
1 Status Report on ADC LPC Clermont-Ferrand Laurent ROYER, Samuel MANEN.
PHASE-BASED DUAL-MICROPHONE SPEECH ENHANCEMENT USING A PRIOR SPEECH MODEL Guangji Shi, M.A.Sc. Ph.D. Candidate University of Toronto Research Supervisor:
A Reconfigurable FPGA Architecture for DSP Transforms Subramanian Rama Vishnu Vijayaraghavan.
Speech Enhancement using Excitation Source Information B. Yegnanarayana, S.R. Mahadeva Prasanna & K. Sreenivasa Rao Department of Computer Science & Engineering.
Power Management System Hardware Milestone: 0 Software Milestone: 0 Requirements: Interface with the existing bicycle battery voltage (35-40V typical)
Motorola presents in collaboration with CNEL Introduction  Motivation: The limitation of traditional narrowband transmission channel  Advantage: Phone.
Introducing Networks and the Internet Mrs. Wilson Rocky Point High School.
UNIT-IV. Introduction Speech signal is generated from a system. Generation is via excitation of system. Speech travels through various media. Nature of.
A DSP based on on-line UPS R.Padamaja G.Mamatha Reddy EEE EEE S.V.C.E S.V.C.E BY.
Maria Cinque, Michele Crudele, Giulio Iannello Università Campus Bio-Medico di Roma Hospital Information System for Students The results of the HISS project:
Microprocessor Presentation Md. Enamul Haque Id:
What is the difference between PSD and CCD sensor technology?
BioLock (Biometric Home Entry System)
Performance of Computer Vision
Augmented von Neumann Processors
Fondazione Istituto Italiano di Tecnologia, Genoa, Italy
Equipment & Environment
Human–computer interfaces
Introducing Networks and the Internet
Anne Pratoomtong ECE734, Spring2002
Digital Systems: Introductory Concepts
DAISY Friend or Foe? Your Wearable Devices Reveal Your Personal PIN
Berkeley Institute of Design
Presentation transcript:

Low Power Speech Enhancement David Halupka Ph.D. Candidate Electronics Group June 24 th, 2005

University of Toronto 2 of 6 Motivation Today’s recognition systems can achieve a 95%+ recognition accuracy after extensive training Research systems: same accuracy with no training Typically: 10% accuracy in the presence of noise, reverberations, and conflicting conversations Humans are equipped to deal with noisy environments  Two ears → let us localize and focus on a single speaker Complex noise: one sensor doesn’t cut it  Multiple microphones → superhuman noise filtering

June 24th, 2005 University of Toronto 3 of 6 Step 1: Sound Localization d x+τν x t t m 2 (t) m 1 (t) Time-Based Cross-Correlation

June 24th, 2005 University of Toronto 4 of 6 Step 2: Speech Enhancement

June 24th, 2005 University of Toronto 5 of 6 A Hard Case for Hardware Localization is a exhaustive linear search  Gradient search, etc. not applicable Each time delay must be checked  Each likelihood can be evaluated in parallel 1 GHz Intel Pentium III needed just for real- time localization → consumes 35 W Speech interface is beneficial for handheld devices, but battery life is limited.  Palm M100 → 150 mW

June 24th, 2005 University of Toronto 6 of 6 Results – 0.18 μm CMOS Die Size: 2.51 mm x 2.51 mm Power Utilization: 29 mW Die Size: 1.51 mm x 1.38 mm Power Utilization: 3.45 mW FPGA: 184 mW DSP: 650 mW