Digital processing of speech signals

When digital images are shared over an open access network such as the internet, facebook, WhatsApp, and other social media, then the security of these images are required. The digital watermarking is one approach for the security of images (e.g. copyright protection, ownership authentication). In most of the watermarking approaches, secret information such as owner binary logos and texts are used for protection of images. These days, biometric watermarks such as human speech signals are preferred for protection of images. In this paper, a watermarking technique based on various signal processing transforms is proposed and implemented for the security of image using human speech signal. In this technique, the first discrete cosine transform (DCT) and then singular value decomposition (SVD) are applied on the watermark speech signal to get its hybrid coefficients which are inserted into hybrid coefficients of the cover image to get a watermarked image. These hybrid coefficients of cover image are first generated using discrete wavelet transform (DWT) and then fast discrete curvelet transform (FDCuT) is applied on it. The performance of techniques is tested for standard speech database such as TIMIT in terms of imperceptibility, robustness and payload capacity. The experimental results and comparison show that the proposed watermarking technique performs better than the existing watermarking techniques available in the literature. This technique may also be used for security of speech signal against spoof attack.

Hybrid domain watermarking technique for copyright protection of images using speech watermarks

A combination of high efficiency cryptosystem algorithms based on MOBS (Modified Overlapped Block Shuffling) and a HCS (Hybrid Chaotic System) are proposed to secure the transfer of an on-the-go audio signal (i.e. speech). As online voice-based communication is becoming increasingly more common, the need to protect these signals from an unauthorized third party becomes more significant. In natural speech, the values and positions for neighboring samples are highly correlated. The proposed algorithm breaks the correlation by using MOBS and encrypting with a HCS. First, the original speech signal is converted from a one dimensional into a two dimensional array. Then, it is divided into overlapped squared blocks followed by a subsequent permutation. Third, each permutated block is dynamically shuffled using Henon and Arnold transformation. And finally a Hybrid Chaotic Map (Henon and Arnold Cat map) is employed to generate a key matrix to encrypt the shuffled blocks, thus producing ciphered speech ready for transmission. Experimental results show that the cryptosystem can successfully encrypt/decrypt with symmetric keys. Experimental analyses like: Signal-to-Noise-Ratio (SNR), Segmental Signal-to-Noise-Ratio (segSNR), Correlation Coefficient Analysis (CCA), Log-Likelihood Ratio (LLR), Residual Deviation, key space and key-sensitivity reveal the effectiveness of the proposed technique for secure speech communication.

/pdf/modify-speech-cryptosystem-based-on-shuffling-overlapping-2at4e1eq7v.pdf

Modify Speech Cryptosystem Based on Shuffling Overlapping Blocks Technique

In﻿this﻿article,﻿a﻿watermarking﻿scheme﻿using﻿Curvelet﻿Transform﻿with﻿a﻿combination﻿of﻿compressive﻿ sensing﻿(CS)﻿theory﻿is﻿proposed﻿for﻿the﻿protection﻿of﻿a﻿digital﻿audio﻿signal.﻿The﻿curvelet﻿coefficients﻿ of﻿the﻿host﻿audio﻿signal﻿are﻿modified﻿according﻿to﻿compressive﻿sensing﻿(CS)﻿measurements﻿of﻿the﻿ watermarked﻿data.﻿The﻿CS﻿measurements﻿of﻿watermark﻿data﻿is﻿generated﻿using﻿CS﻿theory﻿processes﻿ and﻿sparse﻿coefficients﻿ (wavelet﻿coefficients﻿of﻿DCT﻿coefficients).﻿The﻿proposed﻿scheme﻿can﻿be﻿ employed﻿for﻿both﻿audio﻿and﻿speech﻿watermarking.﻿The﻿gray﻿scale﻿watermark﻿image﻿is﻿inserted﻿into﻿ the﻿host﻿digital﻿audio﻿signal﻿when﻿the﻿proposed﻿scheme﻿is﻿used﻿for﻿audio﻿watermarking.﻿The﻿speech﻿ signal﻿is﻿inserted﻿into﻿the﻿host﻿digital﻿audio﻿signal﻿when﻿the﻿proposed﻿scheme﻿is﻿employed﻿for﻿speech﻿ watermarking.﻿The﻿experimental﻿results﻿show﻿that﻿proposed﻿scheme﻿performs﻿better﻿than﻿the﻿existing﻿ watermarking﻿schemes﻿in﻿terms﻿of﻿perceptual﻿transparency. KEyWoRDS Audio Signal, Compressive Sensing (CS), Discrete Cosine Transform (DCT), Discrete Wavelet Transform (DWT), Fast Discrete Curvelet Transform (FDCuT), Speech Signal, Watermarking

Watermarking Scheme with CS Encryption for Security and Piracy of Digital Audio Signals

Normalization and envelope generation of signals are essential pre-processing steps in many signal processing applications. With the emergence of high-performance computing and the need for hardware accelerators in digital signal processing, research for hardware efficient solutions has become a necessity. Interpolation and signal normalization algorithms, though have been implemented in software systems extensively, however, lack the research in hardware system solutions. In this paper, an envelope-based signal normalization technique has been presented along with a detailed discussion on the usage of cubic hermite interpolation. The hardware accelerator architecture for cubic hermite interpolation is presented, and the same is implemented on a field-programmable gate array (FPGA). It is found that the envelope method of normalization which uses interpolation performs better in terms of speed, hardware re-usability and complexity. Using cubic hermite interpolation, an architecture for hardware accelerator for the signal normalization is implemented on FPGA. The hardware accelerator is tested using data obtained from frequency-modulated continuous wave radar. The MATLAB simulation results for both interpolation and normalization are compared with the FPGA implementation results of the same. An average root mean square error and correlation of $$1.67 \times 10^{-4}$$ and 99.78% for interpolation and $$2.1 \times 10^{-4}$$ and 99.71% for normalization are obtained.

A Novel and Efficient Hardware Accelerator Architecture for Signal Normalization

Internet is one of the most valuable resources for information communication and retrievals. Most multimedia signals today are in digital formats. The digital data can be duplicated and edited with great ease which has led to a need for data integrity and protection of digital data. The security requirements such as integrity or data authentication can be met by implementing security measures using digital watermarking techniques. In this paper a blind speech watermarking algorithm that embeds the watermark signal data in the musical (sequence) host signal by using frequency masking is used. A different logarithmic approach is proposed. In this regard a logarithmic function is first applied to watermark data. Then the transformed signal is embedded to the converted version of host signal which is obtained by applying Fast Fourier transform method. Finally using inverse Fast Fourier Transform and antilogarithmic function watermark signal is retrieved.

/pdf/secure-transmission-of-password-using-speech-watermarking-3z1mrz66k1.pdf

Secure Transmission of Password Using Speech Watermarking

In most security domain transmission of secret and confidential voice data requires that the identity of the speaker should not get revealed but at the same time data contents in speech signal should be interpreted. Speech and speaker normalization techniques help to solve this issue. This paper proposes speaker normalization technique that attempts to alter the characteristics of speaker's voice so that it would be difficult for an eavesdropper to identify the speaker. Intended listener can retrieve the particular information. Evaluations of subjective and objective tests confirm the effectiveness of the proposed method and because of this method security enhances.

Normalization techniques for hiding speakers identity

The secure transmission of speech information is a significant issue faced by many security professionals and individuals. By applying voice-encryption technique any kind of encrypted sensitive speech data such as password can be transmitted. But this has the serious disadvantage that by means of cryptanalysis attack encrypted data can be compromised. Increasing the strength of encryption/decryption results in an associated increased in the cost. Additional techniques like stenography and digital watermarking can be used to conceal information in an undetectable way in audio data. However this watermarked audio data has to be send through unreliable media and an eavesdropper might get hold of secret message and can also determine the identity of a speaker who is sending the information since human voice contains information based on its characteristics such as frequency, pitch, and energy. This paper proposes Normalized Speech Watermarking technique. Speech signal is normalized to hide the identity of the speaker who is sending the information and then speech watermarking technique is applied on this normalized signal that contains the message (password) so that what information is transmitted should not be unauthorizedly revealed.

Security issues in speech watermarking for information transmission

The recent advances in information technology have simplified voice data transmission but brings with it major security threats. Voice data transmission not only conveys confidential information but also convey information about the speaker. Most of the security domain complements secrecy about speaker identity. This paper presents a method that deals with the subject of hiding speaker identity. Alteration of acoustic features mainly pitch, and energy of speakers voice is necessary for making it difficult for adversary to infer about the speaker. Speaker normalization approach is used to accomplish this. An evaluation of subjective and objective tests confirms that the normalized signal allows speech waveform to sound intelligible without revealing information about the speaker. This technique can be used in medical, in military and in those domains in which maintaining secrecy about the speaker is a major security concern.

Hiding speaker characteristics for security

This paper addresses two security challenges faced by many while transmitting confidential voice message. First, how to protect confidential message content against unauthorized access and second, how to protect information about the speakers personality. Presented work provides solution to these issues using hybrid approach. The former issue is handled by speech watermarking while speaker normalization deals with latter. The method initially normalizes speech signal containing secret information by modifying the speaker voice acoustic characteristics which is further processed to incorporate additional security measure and then it is embedded in high frequency components of the host signal using the human auditory system (HAS) properties and concept of blind, invisible watermarking approach. Reverse practice is followed to get the intended confidential message. This scheme has been evaluated and results of test shows that the approach does not degrade the quality of host signal and ensures reasonably excellent security. Technique can be used in any domain where secure transmission of the information is important.

Rupa Patel

Papers

Secure Transmission of Password Using Speech Watermarking

Normalization techniques for hiding speakers identity

Security issues in speech watermarking for information transmission

Hiding speaker characteristics for security

Two Level Security Implementation for Voice Data Transmission