Presentation on theme: "Echo cancellation Ian Hung 2B Computer Engineering University of Waterloo August 17th, 2001."— Presentation transcript:
echo cancellation Ian Hung 2B Computer Engineering University of Waterloo August 17th, 2001
Presentation Overview Squelching Echo in a Digital Cellular Network A. Background and Significance B. Echo Basics C. Process of Echo Cancellation D. Echo Control E. Evaluating Audio Quality F. Future Trends This presentation will explore the different methods of handling echo inherent within a digital cellular network.
A. Background and Significance
Wireless phones have gained enormous popularity and are becoming essential communication tools that impact our lives from day-to-day personal tasks to how business is performed. The key is to maintain customer loyalty by improving network quality thus generating additional revenue. Speech quality is the benchmark at which the quality of the network is assessed. For this reason, a considerable amount of research in the removal of inherent hybrid and acoustic echo with the aim of improving perceived voice quality has been conducted.
B. Echo Basics
Hybrid Echo Acoustic Echo Other Echo Sources Communications Network General Overview Cell Site Mobile Switching Center (MSC) Local Telephone Exchange Long Distance Toll Office 4 Wire 2 Wire Due to the cost of cabling, two- wire circuits are used to connect the telephone to the local telephone exchange rather than four-wire trunks.
Hybrid Device 2 Wire Port B. Echo Basics Hybrid Echo 4 Wire Receive Port 4 Wire Send Port Echo Hybrid Echo Acoustic Echo Other Echo Sources Hybrid transformers are used to interface this two-wire circuit with a four-wire station. This conversion does not work perfectly and results in electrical energy leaking back to the talker in the form of echo. When there is little round-trip delay (28ms), it creates a positive sense that the call is live by adding sidetone.
Intrusive echo results when the total network delay exceeds 36ms. The amount of echo depends on how well the hybrid matches both circuits. This is measured as echo return loss ( ERL ). High ERL value = weak signal being reflected. B. Echo Basics Hybrid Echo Acoustic Echo Other Echo Sources
Hybrid Echo Acoustic Echo Other Echo Sources Acoustic Echo Audio reflected from window Audio reflected from dash Audio reflected from door This form of echo originates from handsets and is produced by poor isolation between the microphone and speaker. Multipath echo is transmitted back to the distant end and heard as echo. B. Echo Basics
Hybrid Echo Acoustic Echo Other Echo Sources B. Echo Basics Background noise is generated when the mobile is operated in hands-free mode. This noise is passed through the vocoder causing further distortion in speech. Delay in either speech compression or transmission equipment further induce echo and degrade voice quality. …especially true when the call is processed through mixed network infrastructures including copper wire, fiber optic lines, microwave connections, international gateways, and satellite transmission
C. Process of Echo Cancellation
Hybrid Echo Canceller Send to Far End Line-Side Receive from Far End Longhaul Endpath Drop-Side S-inS-out R-inR-out ERLE Echo Canceller Implementation in the Digital Wireless Network
C. Process of Echo Cancellation Hybrid S-inS-out R-in R-out Echo Canceller – Schematic Diagram Convolution Processor H-Register X-Register Subtractor Double-Talk Detector NLP Echo Estimate Error Signal Residual Echo Contents within either registers are multiplied together, a process called convolution, producing the echo estimate that is fed to the Subtractor. The endpath is mapped Hybrid's impulse response must be learned- known as adaptation
C. Process of Echo Cancellation Hybrid S-inS-out R-in R-out Echo Canceller – Schematic Diagram Convolution Processor H-Register X-Register Subtractor Double-Talk Detector NLP Echo Estimate Error Signal Residual Echo Subtractor eliminates the echo estimate from the convolution processor H-register is adjusted accordingly, refining the estimate of the impulse response, in a process called convergence
C. Process of Echo Cancellation Hybrid S-inS-out R-in R-out Echo Canceller – Schematic Diagram Convolution Processor H-Register X-Register Subtractor Double-Talk Detector NLP Echo Estimate Error Signal Residual Echo Error signal from the Subtractor is ignored causing the contents within the H-register to freeze Double-talk occurs when both near and far-end callers are speaking at the same time. Near-end speech during double- talk can distort the error signal and confuse the adaptation process.
C. Process of Echo Cancellation Hybrid S-inS-out R-in R-out Echo Canceller – Schematic Diagram Convolution Processor H-Register X-Register Subtractor Double-Talk Detector NLP Echo Estimate Error Signal Residual Echo NLP: Reduces the residual echo to inaudible levels based on the adaptive suppression threshold Signals above the threshold are allowed to pass where as signals below the threshold are removed Nonlinear echoes: clipped speech signals speech compression poor quality speakerphones pulse code modulation (PCM) …challenge convolution processor to develop an accurate echo estimate.
D. Echo Control
Handling Hybrid Echo Handling Acoustic Echo Handling Other Complex Echo D. Echo Control Hybrid Echo Canceller Implementation in the Digital Wireless Network Cell Site Mobile Switching Center (MSC) Public Switched Telephone Network (PSTN) 4 Wire 2 Wire4 Wire Echo Canceller When a call is made from the mobile user to the PSTN, the caller will experience echo-free speech, while the receiver may experience poor speech quality. Therefore, the echo canceller must work both ways to accommodate users from either ends.
D. Echo Control Handling Hybrid Echo Handling Acoustic Echo Handling Other Complex Echo D. Echo Control Echo Canceller Implementation in the Mobile Handset Echo Echo Canceller To MSC (no Echo) From MSC Microphone (Speech w/Echo) Loudspeaker During conversation, the audio picture dynamically changes, which causes the canceller to adapt continually. EC should be capable of receiving 270 ms of delay. Any less and the remaining echo would be beyond the ability for the canceller to remove them. Acoustic echo return loss enhancement (AERLE), the amount of attenuation applied to unfavorable echo, has a value of 65 dB as the minimum requirement with the non-linear processor enabled.
D. Echo Control Handling Hybrid Echo Handling Acoustic Echo Handling Other Complex Echo D. Echo Control Back-to-Back Implementation in the Digital Wireless Network Due to strong price pressures, most handsets do not control echo very well. In some cases, such handsets are known to produce a terminal compiling loss of 24 dB. Performance is dependent upon both network-side EC integration, and mobile-side handset design Echo Cancelled Voice Sample Echo Cancelled Hybrid 2 Wire4 Wire
E. Evaluating Audio Quality
D. Echo ControlE. Evaluating Audio Quality A. Total Echo Removal 1. No echo at the beginning of the call 2. No Echo throughout the call - call is stable without bursts or distortion 3. Residual Echo (low-level) is absent throughout the call B. Double-Talk Clarity 1. No echo or scratching noises throughout the call 2. Speech is clearly audible during double-talk 3. No clipping at the beginning and end of speech *Occasionally, echo cancellers may be fooled due to high background noise. If double-talk detection is handled incorrectly, near-end speech, perceived to be echo, may be attenuated. C. Background Audio Transparency On contrary, background audio transparency is the ability for an echo canceller to allow background noise to be heard. Poor DSP = audio fading in and out Better DSP = consistent background audio. Noise injection to fill-in for background audio is occasionally used. Replacing these intelligible background sounds with random noise can be irritating.
F. Future Trends
D. Echo ControlF. Future Trends In the pursuit for greater call clarity, echo cancellers have been the center of recent development, and have now become highly sophisticated transmission equipment at the heart of highly complex networks. Effective implementation of echo cancellation across the network holds the key to improving call quality and ultimately, strengthen customer satisfaction, boost revenues, and reduce subscriber churn.