Yanhong Fan

A 2400 bps vocoder based on mixed excitation linear prediction and channel coding

To obtain high quality synthetic speech at 2400 bps, this paper presents a vocoder based on the Mixed Excitation Linear Prediction (MELP) model. The differences of the vocoder parameters are analyzed, and an unequal error protection channel coding and a parameter substitution method for error frame are proposed to improve the robustness over random error channel. Several channel coding schemes are compared and the optimal one is then selected. Test results show that the proposed speech coding algorithm could provide satisfactory speech quality and strong robustness against channel errors.

Efficient audio data hiding via parallel combinatory spread spectrum

Spread spectrum (SS) modulation is one of the most commonly used methods for data hiding. However, the capacity of SS-based methods is rather limited. To improve the hiding efficiency, this paper presents a method that employs the parallel combinatory spread spectrum (PCSS), in which the hidden data can be transmitted in parallel with a combination of a few pseudo noise sequences. An informed embedding strategy is applied to PCSS data hiding to improve its robustness. Compared with the methods using the techniques of code division multiple access (CDMA) and M-ary SS modulation, the proposed method can provide a much higher achievable capacity with similar decoding complexity. This method is evaluated in an audio data hiding system, and the experimental results show that the embedded data can resist typical signal processing and attacks, even in high-capacity applications.

A 1.8kbps vocoder based on Mixed Excitation Linear Prediction

With the rapid development of communication technology, there is an urgent need for high quality speech coding algorithm at very low bit rate. In the paper, based on the Mixed Excitation Linear Prediction (MELP) and unequal error channel coding, a low bit rate speech coding algorithm is proposed. According to the different importance for each parameter, the relatively significant code stream information is protected by using channel coding with strong error correcting ability to obtain high-quality synthetic speech at 1.8 kbps with 1/3 redundancy for channel coding. Test results show that the proposed algorithm could provide better speech quality and also has strong robustness for channel error.

Informed embedding with selective host rejection for CDMA-based high-payload and robust watermarking

This paper presents an informed embedding method for code division multiple access (CDMA) based robust watermarking. First, a generalized model for CDMA-based watermarking with informed embedding is described, taking into account more detailed host information and possible pre-processing operations on the watermark at both the embedder and the extractor. Then, as special cases of this model, several typical embedding methods are analyzed, including an informed embedding method with selective host rejection which exhibits better robustness for high-payload watermarking. Moreover, an audio watermarking scheme is designed to verify the proposed method. Experimental results show that this scheme can achieve a high payload of up to more than 200 bps (bit per second) while preserving the audio fidelity and the robustness against common signal processing and attacks. Therefore, the proposed model and embedding method provide a general and effective approach to improve the performance of CDMA-based watermarking systems.

Robust audio watermarking based on frequency-domain spread spectrum using CAZAC sequence

This paper presents a robust audio watermarking scheme based on spread spectrum in frequency domain using constant amplitude zero autocorrelation (CAZAC) sequence. First, the watermark bits are modulated by using M-ary phase shift keying (PSK), and then the modulated symbols are spread in phase by using the CAZAC sequence. At the extractor, the embedded information is recovered by despreading the phases of the watermarked audio. Experimental results show that the proposed scheme provides an adjustable payload, and is robust to most of the common signal processing and attacks while preserving the audio fidelity.

Coordinated Transmission with Block Diagonalization in MIMO Broadcast

In this paper, we investigate coordinated multiple-input multiple-output (MIMO) transmission schemes in an interference-limited cellular downlink. We propose a novel block diagonalization (BD) scheme which is combined with zero-forcing (ZF) criterion. In the scheme, multi-user interference is mitigated by BD while interference among spatial data streams for a user is mitigated by ZF. Also an efficient power allocation is proposed. Sum rate is largely increased by combining the novel coordinated scheme with the power allocation. Computer simulations demonstrate the effectiveness of the proposed coordinated scheme and its corresponding power allocation.

400bps High-Quality Speech Coding Algorithm

Low bit rate speech coding is important to speech communications over band-limited or harsh channels. In this paper, based on the mixed excitation linear prediction (MELP) model, we propose a high-quality 400bps low bit rate speech coding algorithm which introduces multi-frame joint vector quantization, adaptive spectral enhancement and multi-band sinusoidal mixed excitation. Efficient parameter quantization schemes are employed on the basis of the super-frame structure. It is verified that the synthesized speech has fairly high intelligibility and naturalness, and the mean opinion score (MOS) is about 2.52.

Adaptive masking system based on Speech Intelligibility Index

This paper presents a speech protection system based on adaptive masking. The system can adaptively adjust the masking noise volume according to the Speech Intelligibility Index (SII) proposed in ANSI S3.5-1997. The volume of the masking noise is minimized by limiting SII to a low level at which the speech to be protected can not be understood. This method can effectively reduce excess masking noise to obtain good acoustic comfort for listeners inside the protection region while preventing eavesdroppers from stealing the speech information. Experimental results show that the effect of the proposed system outperforms the traditional sound masking systems in masking efficiency and flexibility.

A variable-bit-rate speech coding algorithm based on enhanced mixed excitation linear prediction

In order to improve the channel bandwidth utilization of voice communication, a variable bit rate speech coding algorithm based on enhanced mixed excitation linear prediction (MELPe) is proposed in the paper. In voice communication, only about 40% of the time is occupied by talking, whereas the rest is engaged by silence or background noise. In addition, unvoiced frame usually requires less transmission rate than the voiced one in low bit rate speech coding algorithms. Therefore, always using the same coding bit rate for speech coding is a waste of channel resource. In this paper, the input signal is divided into speech and silence by using voice activity detection (VAD) technology. And the speech frames are divided into voiced frame or unvoiced frame. They use different coding rates for speech coding and data transmission. All of the parameters are encoded, transmitted and decoded in voiced frame. Only gain parameters, LSF parameters, pitch parameters and overall voicing are encoded, transmitted and decoded in the unvoiced frame. Furthermore, only the gain parameters and the first level LSF parameters are encoded, transmitted and decoded in the silence frame. When about 40% of the time is occupied by talking, compare with the traditional 2.4 kbps MELPe vocoder, the average coding rate of the proposed variable bit rate vocoder can reach 1.33 kbps. But they can achieve the same quality of synthetic speech. Experimental results show that the proposed method reduces the average coding rate, and the synthetic background noise has good comfort on the subjective sense of hearing.

A new informed spread spectrum embedding for robust audio watermarking

This paper presents a new embedding strategy to extend the performance bound of spread spectrum (SS) based watermarking by introducing more imperceptible distortions measured in the mean square errors (MSE). The potential of the host is sufficiently exploited and utilized to maximize the watermark robustness. This strategy is then realized in audio watermarking by adaptively inverting the host according to the correlation between the host and the modulated watermark sequence. The MSE embedding distortions can reach up to twice of the host power, while the perceptual distortions are effectively controlled. By utilizing more available host, the proposed method exhibits much better robustness than existing SS-based watermarking.