Enhancement of speech signals is required in many situations in which the signal is to be communicated or stored. Speech enhancement is required when either the signal or its receiver is degraded. For example, hearing impaired individuals require enhancement of perfectly normal speech to fit their individual hearing capabilities. Speech signals produced in a room generate reverberations, which may be quite noticeable when a hands-free single channel telephone system is used and binaural listening is not possible. A speech coder may be designed for clean speech signals while its input signal may be noisy. Similarly, a speech recognition system may be operated in an environment different from that it was designed to work in.
The rapid growth of broadband communications is providing connections for high-quality voice communications. Scalable audio/speech coding has become increasingly popular recently because it
can efficiently accommodate the bandwidth fluctuation. A scalable audio/speech bitstream typically consists of a base layer plus a number of enhancement layers. It is possible to use only a subset of the layers to decode the audio with lower sampling resolutions and/or quality. In streaming applications, the layers in a scalable bitstream are selectively delivered to adapt to network bandwidth fluctuation and packet loss level. When the available bandwidth is low or packet loss ratio is high, only the base layer is transmitted.
For more information
- De-reverberation of Speech Signals
- Error Concealment for Speech Transmission
- Dual Channel Noise Estimation for Speech Enhancement
- Speech Enhancement and Speech Intelligibility
- Standard Methods of Voice Activity Detection (VAD)
- The Applications of Voice Activity Detection (VAD)
- Battlefield Voice Activated Transmission (VOX)
- VoIP Stack Featuring Voice Quality Enhancement (VQE)
- Echo Reduction in Coding
