QLAW: An Improved Quantization-Based Local Audio Watermarking Scheme Using Inter-Frame Correlation

With the rapid development of the Internet, audio distribution has become more convenient with increasing copyright infringement. To address this problem, this paper proposes a quantization-based local audio watermarking scheme using inter-frame correlation, integrating machine learning techniques and traditional methods. To obtain the time-frequency spectrogram of the audio signal, a Short-Time Fourier Transform (STFT) is first applied to the audio signal. Then, Main Energy Region Extractor (MERE) is proposed to extract the main energy region of the spectrogram. Based on the main energy region, the Stable Frequency and Energy Region Extractor is conducted to find the local feature region for embedding. After segmenting the local feature embedding region into several frames, Adjacent Frame Extraction Process (AFEP) is conducted to select the adjacent frame. Then, Discrete Cosine Transform (DCT) is applied to each embedding frame and its adjacent frame to extract their corresponding frequency domain coefficients. To improve robustness, mid-frequency DCT coefficients are alternately selected to embed the watermark. By adjusting the difference between the embedding frame and its corresponding adjacent frame in a predefined range, the local watermark is embedded. Experimental results show that the proposed scheme outperforms existing schemes in inaudibility and robustness, achieving an average Signal-to-Noise Ratio (SNR) above 25 dB and a lower Bit Error Rate (BER) under various attacks.