Atsushi Murashima

A post-processing technique to improve coding quality of CELP under background noise

This paper proposes a novel post-processing technique to improve the coding quality of CELP under background noise. It adaptively smoothes both the spectral envelope and the energy of the estimated excitation signal to reduce their temporal fluctuations, which cause the perceptual degradation. The excitation signal is calculated using the synthesized signal and the spectral parameters given from the decoder. Thus, the proposed post-processing is performed separately from the decoder. The smoothing is applied only in non-speech periods and the smoothing strength is controlled depending on the characteristics of the synthesized signal to avoid the degradation in speech and non-stationary noise periods. Subjective test results show that the proposed post-processing improves degradation mean opinion score (DMOS) by 0.2 to 0.4 for noisy speech signals, which are coded by the GSM adaptive multi-rate (AMR) codec.

Multi-rate wideband speech/channel codec based on MPEG-4/CELP for ETSI/GSM full-rate channel

This paper proposes a wideband multi-rate speech and channel codec based on the MPEG-4/CELP for the ETSI/GSM full-rate channel. In order to improve coding performance under mobile environments, such as channel error and background noise, the proposed codec operates at three bit allocations between speech and channel coding with a constant gross bit-rate of 22.8 kbit/s. The speech coding bit-rates are 10.9, 12.1 and 15.9 kbit/s. It achieves high speech quality under any channel condition by switching the bit allocations and also for noisy speech by using the highest bit-rate. The preliminary subjective evaluation tests show the speech quality is improved by switching the bit allocation under error conditions. It is also comparable of superior to ITU-T Recommendation G.722 48 kbit/s for carrier-to-interference ratios (C/I) higher than 10 dB. The codec at 15.9 kbit/s also gives comparable speech quality to G.722 at 48 kbit/s under background noise conditions.

Low Complexity Real-Time Video Transcoders for Video Upload and Retrieval Applications

This paper proposes low complexity real-time video transcoders for video upload and retrieval applications. The proposed transcoders consist of both picture type conversion and syntax conversion in the compressed domain, and can highly reduce the computation amount in comparison with that of the conventional tandem transcoding method, in which decoding and re-encoding are necessary. The proposed syntax conversion can provide loss-less conversion, while the conventional tandem method is lossy conversion. Timestamp adjustment method is also proposed to avoid audio visual de-synchronization problem due to picture type conversion, when the stream is transmitted via constant bandwidth network. Complexity evaluation results based on DSP implementation confirm that the complexity of the proposed transcoders is very low and is less than 1/3 of that of the conventional tandem transcoder.